# Changelog

All notable changes to CodeGraph are documented here. Each entry also ships as
a [GitHub Release](https://github.com/colbymchenry/codegraph/releases) tagged
`vX.Y.Z`, which is where most people will look.

Each release opens with a short **Highlights** list — the handful of things most
users will notice — followed by the full notes.

This project follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/)
and adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [Unreleased]

### Highlights

- **`codegraph ui` — your graph in a browser.** A local, read-only viewer for the project you already indexed: your code with its callers and callees in the margin, a map of the whole repository, and a strip that shows how one symbol reaches another.
- **See your app the way its users meet it.** A Screens tab draws every screen and the navigation between them, for Expo Router, React Router, Next.js, TanStack Router, Vue Router / Nuxt and SvelteKit apps.
- **See what happens from a screen or an endpoint.** A Steps tab draws what one action sets in motion — the handlers it fires, the state it writes, the calls that leave your code, and every way it can answer — with the condition on each arrow.
- **APIs too, and across tiers.** Endpoints in Express, NestJS, Fastify, Koa, Hono, FastAPI, Flask, Django, Spring, ASP.NET, Vapor and Gin, with a page's `fetch` following through to the route that serves it, a queued job to its consumer, an event to its handler.
- **Read a handler in the order its code runs.** The same picture laid out by when things happen rather than by distance, so a reply sits below the token it carries. Where the code chooses, the condition is said once and each arrow answers it.
- **Find what nothing reaches, and where everything starts.** Dead code and entry points as screens of their own, a type hierarchy showing what is built on what, and trails you can name, save and come back to after the code has changed.
- **Conditions and arguments for more languages.** What a call passes, and the `if` / `switch` / `try` it runs under, are now read for Python, Java, Kotlin, C#, Go and C as well as JavaScript, TypeScript and Swift.
- **Upgrading:** re-index your projects after this release — several of the new readings rest on edges that are written while indexing.

### New Features

- **Codex and Astra read project guidance from `AGENTS.md`.** The canonical agent guide now lives in `AGENTS.md` (with a nested `docs/AGENTS.md` for long validation notes); `CLAUDE.md` is a thin `@AGENTS.md` wrapper for Claude Code. Codex/Astra no longer miss the old CLAUDE-only instructions.

- **A big screen's picture stops wrapping into a column.** How wide a screen's lines run before they wrap was worked out with a formula, and the formula was wrong for the way these pictures are actually drawn: a part of a screen spends lines on its own structure — a step that fires things gets a line to itself, and what it fires starts another — so estimating the lines from the boxes alone badly undercounted them, and one screen's 98 boxes wrapped into a 4,356px column. Laying a picture out is cheap and exact, so the widths are now simply tried and the one that comes out closest to the shape of a window is kept. Across one app's 51 screens the tallest picture went from 4,356px to 3,796px, total height fell 8%, and — because a shorter picture is also a picture whose lines have less far to go — lines running over other boxes fell by a third and lines crossing each other went from 13 to 5.

- **A screen's picture stops being mostly empty.** The parts of a screen were tiled a row at a time, each row as tall as its tallest member — so one short region beside a tall one left the whole rest of that row blank. Measured across one app: the median screen's canvas was only 55% picture and 45% nothing, and its busiest screen was 44% — 4,860px tall to hold about 2,160px of content, all of which a reader has to scroll through. Each part now goes as high as it can and then as far left as it can, so a short one tucks under another short one instead of waiting for the tall one beside it. Reading order is unchanged: parts are still placed in the screen's own source order, and an earlier one is never pushed below a later one. That screen is now 3,584px instead of 4,860px, and close to square rather than a ribbon.

- **The screen's own line into each part of itself stops sweeping across the picture.** A screen's picture holds back the screen's whole fan-out and draws one line into each region instead — but on a screen with ten regions those regions tile into bands, so the line into a region two bands down had to travel the height of the whole picture to get there. Measured across one app's 51 screens: those lines were 17% of everything drawn and caused **79% of every remaining crossing**. Two fixes. The line now lands on the box nearest the region's top-left that the screen actually leads to, rather than on whichever box the walk happened to meet first — which, once a screen is drawn in clusters, could be lines down inside the region. And when the region is still too far for a line to be followed, the link is said in words like any other, with no exception for the screen itself. Across those 51 screens, lines crossing each other went from 47 to **10**, with no screen above 10; the busiest screen went from 19 to 2.

- **A busy screen's picture no longer draws itself as a web.** On a screen whose whole body is one component, the region grouping had nothing to divide — 98 of 100 boxes landed in a single column — and the picture came back as a 5,600px ribbon whose 113 lines crossed each other 652 times, each one running over five other boxes' names. Three things are fixed. A screen is now laid out in **clusters**: a step, then the steps it sets in motion on the line under it, stepped in — so a box and the thing it fires are one line apart instead of seven, which is what the crossings were made of. A region's lines run to a width its size earns rather than a fixed one, so a big screen comes out about as wide as it is tall instead of a narrow ribbon. And a link whose two boxes end up too far apart to follow is now **said in words at both ends** — `→ resumeInference` under the box that leads there, `← CaptureView` under the box it arrives at — rather than drawn as a line across the whole picture; select either box and every one of its real lines draws, exactly as before. On the screen that prompted this, the lines drawn at rest cross each other **once**, down from 652, with half of them still drawn as real lines. Steps that fire nothing still share a line, so a flat screen is unchanged. Nothing needs a re-index.

- **A region that holds a cycle no longer stretches the picture.** Rows inside a region were settled by relaxation, which never converges on a graph that has a loop in it: one real screen's 65 boxes were pushed to rows 294-301 while the rest sat at rows 0-2, and the order they were drawn in had nothing to do with what leads to what. Cycle-closing links are now set aside before the rows are settled, as the in-order reading already did — the loop is still drawn, it just cannot stretch the picture.

- **A busy screen's picture is laid out by the parts of the screen.** A screen is a set of handlers with no order between them, so on a hub screen the old rows-by-distance collapsed into one enormous row — the main screen of one app put 89 boxes side by side on a canvas over 28,000px wide, every line a near-horizontal sweep across all of it. The Steps tab now groups a screen's picture by region — the component that owns each handler, named in a small caption over its boxes — with each region a column where a step sits above what it sets in motion, tiled in the screen's own source order. At rest the picture hides only two things: the screen's own fan-out — one line into each region stands in for it — and lines that point back up; every other line draws where it leads, between two regions included, and selecting a step brings out its whole story in the side panel, link by link. A box nothing points at is the screen's own doing — run on render or mount, or from a binding written inline — the key says so, and selecting it lights its line from the screen with what fires it. The same app's widest screen now lays out under 3,500px with every line local, and the whole picture fits on screen when it opens. Endpoints, handlers and the in-order reading are untouched, and nothing needs a re-index: the regions come from the same walk that draws the steps.

- **Where the code chooses, the picture says so once.** A helper that ends `return (await hasSeenWelcome(id)) ? '/home/' : '/welcome/'` sends the app to one of two screens, but the Steps picture drew that as two separate arrows, each carrying the whole condition with one of them negated and both cut off at the same forty characters — and before you clicked anything, neither arrow was labelled at all, so nothing said it was a choice. Now sibling arrows out of one box that are the arms of one `if`, `switch` or ternary are drawn as the choice they are: the condition is written once under the box that decides it, and each arrow out says only which way it is — `yes`, `no`, or a case's own value. They are the only arrows labelled before you select anything, so the picture reads at a glance without becoming a wall of text. A one-sided guard — an early exit, an `if` with only one side drawn — still carries its condition on the arrow, and an arrow that is reached whether or not the condition holds never claims a side. Nothing needs a re-index: the decision is read from the source at request time.

- **A dialog's buttons fire what they run.** `Alert.prompt('Add Folder', …, [{ onPress: (name) => createBackgroundFolder(name) }])` is two facts: the prompt is a call that leaves the index, and its button fires the handler. The handler's line now arrives from the dialog's own box — with the condition on it — instead of from the screen, so the confirm-then-act chains a mobile app is full of read as chains: the delete alert leads to the delete, which leads to the request it sends. The same holds for anything bound inside the arguments of a call that leaves the index. Nothing needs a re-index: it is read from the source at request time.

- **The Steps tab draws a handler in the order its code runs.** The picture of what an endpoint sets in motion put the lookup, the token signing, the 200 and the 401 side by side, because each is one step from the anchor — true, and not how the code reads. Now a handler, an endpoint or any function opens as the same picture laid out by *when* things happen: a line means **and then**, so the 200 sits below the token signing it is built from and the 401 branches off the check that chose it. Where the code forks both ways, a small box asks the condition once and each line out of it answers — `yes`, `no`, a case's value — so the two arms of one `if` read as one decision instead of two lines that each carry the whole predicate, one of them negated; selecting a step next to the decision lights the whole fork through it. A lone guard — an early exit, an `if` with one drawn side — keeps its condition on the line, and an arm that answers the request, returns or throws simply has nothing leaving it. A call written inside another call's arguments happens first, so the token is signed before the reply that carries it. A helper is drawn where it is called (`via generateToken`), a body that repeats says so (`for each item of items`), and work registered to run later (`later · then`) or started at once (`together · Promise.all`) says that rather than pretending to be a sequence. A screen still opens as before — its handlers fire on events and have no order between them — and either reading is one click, or one `&view=order` / `&view=tree` in the link, away. Nothing to re-index: it is read from the source at request time, and where the conditions cannot be read the picture is a plain sequence rather than an invented structure.

- **A Next.js app lands on the Screens tab like a mobile app.** App Router pages (`app/(group)/blog/[slug]/page.tsx` → `/blog/:slug`) and Pages Router pages are screens bound to the component they export; `<Link href>`, an internal `<a href>`, `router.push` / `router.replace` (`next/navigation` and `next/router`), `redirect()` / `permanentRedirect()` in a server action or a page, and the middleware's `NextResponse.redirect(new URL('/login', req.url))` are the transitions between them — each attributed back to the page it starts on with the plumbing folded and the condition on the arrow, a link written in markup drawn dashed as an inferred hop. `app/api/**/route.ts` exports (`GET`, `POST`, …) are endpoints bound to their functions, `pages/api/*` handlers are `ANY /api/…`, and a page's Steps picture fires from its load (`FIRES FROM page load · /users`), draws the data it reads, the handlers it wires, the server actions it crosses to and the pages it leads to as boundaries. A response's status written as `{ status: 201 }` is read too. Re-index after upgrading.

- **TanStack Router apps land on the Screens tab too.** Routes are read both ways a TanStack app declares them: file-based, where `createFileRoute('/posts/$postId')` carries the whole path as a literal, and code-based, where each `createRoute({ path, getParentRoute })` names a fragment that is composed through its parent into `/posts/$postId`. `navigate({ to })` from `useNavigate`, a thrown `redirect({ to })` from a loader or `beforeLoad`, and `<Link to>` / `<Navigate to>` are the transitions between them. TanStack is the one router here whose destination is the route PATTERN rather than a filled address — `<Link to="/posts/$postId" params={{ postId }}>` names the route and passes the values beside it — so a destination is read as a pattern and matched against the route it names. Addresses that are not pages are left off the map: a `_auth` segment is a pathless layout and never appears in the URL, a `(group)` folder is invisible, a `__root` route wraps everything without being a page, and a file that renders an `<Outlet />` is the layout around an address while the index route beside it is the page at it. A computed `to`, a pattern no route serves, and a `navigate({ search })` that only changes the query are left out rather than guessed. Re-index after upgrading.

- **Vue Router and SvelteKit apps land on the Screens tab too.** Both drew their screens as islands with no transitions, so the tab stayed hidden; now the navigation between pages is read for each. **Vue:** the routes are read out of `createRouter({ routes: [...] })` — path, name, and the view each entry names, including a lazy `component: () => import('@/views/Login')` — and `router.push` / `router.replace` / `$router.push`, Nuxt's `navigateTo`, and `<router-link>` / `<RouterLink>` / `<NuxtLink>` are the transitions. Vue apps usually navigate by route NAME rather than by path, so `router.push({ name: 'profile' })` and `:to="{ name: 'profile' }"` resolve by name, and `router.push({ path: '/', query })` by path. **SvelteKit:** `goto('/login')`, `redirect(303, '/article/' + slug)` from a load or a form action — whose destination is its *second* argument, after the status — and the plain `<a href="/login">` that is a link in a SvelteKit app. A SvelteKit page also opens with a body now — it is joined to the page file that serves it and to the `+page.server.js` beside it — so its Steps picture draws its loader's work, its form actions, and the auth guard the loader performs (`redirect(302, '/login')` under `if (!locals.user)`) as a transition to the sign-in page, with the condition on the arrow. A computed destination, a path or name nothing declares, a relative path in a nested route, and a conditional whose two arms go to different pages are left out rather than guessed. Re-index after upgrading.

- **A React Router app lands on the Screens tab too.** `<Route path='/payment' component={PaymentScreen}>` (v5), `<Route path='/payment' element={<PaymentScreen/>}>` (v6) and `createBrowserRouter([{ path, element }])` already named a project's screens; now the navigation between them is drawn as well. `history.push('/placeorder')` and `history.replace`, `navigate('/placeorder')` from `useNavigate`, `redirect()` in a loader or an action, and `<Link to>` / `<NavLink to>` / `<Navigate to>` / react-router-bootstrap's `<LinkContainer to>` each become a transition — attributed back to the screen it starts on, with the plumbing folded and the condition on the arrow, a link written in markup drawn dashed as an inferred hop. A route with an optional parameter (`/cart/:id?`) is reached by both `/cart` and `/cart/5`. Until now a React Router project's screens were drawn as islands with no transitions at all, and a screen's Steps picture left out every page it sends you to — a checkout step showed its saved payment method but not that it goes on to place the order. A computed destination (`history.push(redirect)`), a path no route serves, and a relative path inside a nested route are left out rather than guessed, and an ordinary `paths.push('/x')` on an array is never mistaken for navigation. Re-index after upgrading.

- **Double-click a box to go there.** On the Steps tab a double-click on any step starts the picture from it — the same as the panel's *Start here* — so an endpoint the page calls, or another screen drawn as a boundary, opens as its own chapter in one gesture; on the Screens tab a double-click on a screen opens what happens from it. A boundary's panel now says it is not entered rather than that nothing leaves it.

- **The Steps tab follows a web app across its tiers.** A page's `fetch('/api/users', { method: 'POST' })` (or an `axios` / `ky` / `got` / `$fetch` call, including one through a project instance made with `axios.create({ baseURL })`) now reaches the route that serves it in the same index — drawn as a crossing to the server (`⇢ POST /api/users`) with the handler named on the box and the registration site in the panel, a boundary by default and entered with *Continue through*, so the picture reads page → handler → the endpoint → its database write → its response. A job put on a BullMQ / Bull queue lands on the `@Process` method, `Worker` or `queue.process` handler that consumes it; a NestJS `EventEmitter2` event lands on its `@OnEvent` listeners (globs included); a socket message crosses from a client's `socket.emit` to the gateway's `@SubscribeMessage` and back from the server's `emit` to the component that registered `socket.on`; and a Next.js server action called from a client file is a crossing to the server by its `'use server'` directive. Each of these is a synthesized hop — dashed, with where it was wired up — and `codegraph_explore`'s Flow section names them too. Only a literal path or event name pairs: a variable url, a path no route serves, or one two routes serve alike produce nothing. Re-index to pick the new edges up.

- **The Steps tab now draws an API as well as an app.** Anchor on an endpoint — `POST /users` in Express, NestJS, Fastify, FastAPI, Flask, Django, Spring (Java or Kotlin), ASP.NET, Vapor or Gin — and the viewer starts at the handler the route runs (or at the route itself when the handler is an inline arrow), says what fires it (`FIRES FROM POST /users · after authenticate, validate(…)` — the middleware arguments at the registration, or the guard decorators on the method and its class, or a FastAPI `dependencies=[…]`), and draws what the request sets in motion: the database calls with the model and whether they read or write (`prisma.user.create({ data })` · `database · user · write`), jobs put on a queue, emails, payments, cache reads, token checks, calls to other services, files and processes — and the **responses**, one box per status a handler can send (`201`, `404`) whose panel rows are the endpoint's contract as the code has it: `WHEN NOT user → 404 · NotFoundException('no such user')`, `always → 201 · res.status(201).json(user)`. A queue consumer or a scheduled job anchored by name says the decorator that fires it (`@Process('email')`). The legend, the panel and the chooser use the project's own words — endpoint, data call, another tier — and the bare Steps tab lists an API's endpoints by router file when there are no screens. Re-index is not needed: everything new is read from the source at request time.

- **Calls are read as written, so the database is the database.** The index keeps only the last segment of a deep member call (`create` for `prisma.user.create`), and a bare name matches by name alone — often to the wrong `create`. The Steps walk now reads each call from the source as written, classifies `prisma.user.create`, `this.usersRepository.save`, `session.commit`, `owners.save` and `_context.TodoItems.Add` by the whole chain and by the receiver's declared type (`OwnerRepository owners`, `private readonly usersService: UsersService`, `val owners: OwnerRepository`, read from the class body), and follows `this.usersService.findByEmail(…)` into the class the type names instead of the name-only guess. A hop resolved this way says so in the panel.

- **Conditions and arguments for Python, Java, Kotlin, C#, Go and C.** The `when` on a call — `if` / `elif` / `else`, `switch` / `when` / `match`, the ternary and Kotlin's `if` expression, `try` / `except` / `catch`, `and` / `or`, and the early exits before it (`if err != nil { return }` reads as `err == nil`, `if not item.title: raise` as `item.title`) — and what each call passes (`HTTPException(status_code=422, detail="bad price")`, `c.JSON(http.StatusCreated, gin.H{…})`) are now read for those languages too, in `codegraph ui`'s rails, Flow strip and Steps tab and in `codegraph_explore`'s Flow section. As before: read from the source as it stands, never stored, and a language without rules yields nothing rather than a wrong label.

- **A Steps tab in `codegraph ui` — what happens from here.** Pick a screen (or search any symbol and choose *What happens from here*) and the viewer draws everything it sets in motion as typed steps: the handlers wired to its taps and listeners, the calls that cross into native code, the native events that come back, the store actions it writes, and the calls that leave the app into the network, storage, the device or telemetry — one box per step, an arrow for every way one leads to the next, and on each arrow the condition under which it happens. The plumbing between two steps (hooks, helpers, the components in between) is folded into the arrow and listed in the side panel, exactly as the Screens tab folds a tap's chain into one transition — and every call the panel lists says what it passes, read from the source as written (`SecureStore.setItemAsync('userEmail', values.email)`, `axios.post('/auth/login', { email, password })`), so a step is not just *that* something was stored or sent but *what*. And each handler says what fires it — the JSX prop and its element (`onPress · <Button>`), the option it is written under (`onSubmit · useFormik(…)`), the listener or effect it runs from — read from the source at the call site, so `onPress={() => handleLogin(values)}` and Formik's `onSubmit` make `handleLogin` a step of its own with the event on the arrow into it. Any step is the next anchor, any link opens as a Flow strip, a cap the walk hit is announced on the step it hit it at, and the picture travels in the URL. React Native + Expo apps get the full picture today; any project gets handlers, stores and calls that leave the index.

- **React Native apps: Swift native modules and their events connect end to end.** A JS call like `captureView.finalizeCaptureSession()` — where `captureView` is bound to `NativeModules.CaptureView` and the module is a Swift class exposed through an `RCT_EXTERN_MODULE` shim — now resolves to the Swift method itself instead of stopping at the constant, so `codegraph_explore`, the Flow strip and the Steps view follow the code into native. Native → JS events now also land on listeners written inline (`addListener('onZipComplete', (data) => { … })`), attributed to the component that registers them. Re-index after upgrading to pick the new edges up.

- **A Screens tab in `codegraph ui` — the app the way its user meets it.** One box per screen, an arrow for every way of getting from one to another, and on each arrow the condition under which it happens. Click a screen and each of its transitions is labelled beside the screen at the other end of the line with the last condition checked before it happens — `→ isCollected` above `/object-detail` — laid out so that no two labels overlap and none sits under a line; hover a label, a line, or its row in the side panel for the whole condition and the chain the tap travels through (`HomeSearchResults → ItemCard → openObjectDetail`), with a link to each navigation call. A screen that returns to where it came from is drawn around the boxes rather than through them, shared chrome (a top bar rendered on ten screens) is one node a row above what it opens rather than the same arrows from every box, a screen that opens many others is wide enough to follow each line back to it and its lines take separate paths through the gap so they fan out instead of stacking, hovering picks the line nearest the pointer, and a helper that chooses the destination after login shows its fork. Projects whose graph holds screen navigation land on this tab. Expo Router apps today.

- **Expo Router apps: screens and navigation are in the graph.** Every screen file under `app/` (or `src/app/`) is now a route node named by its path — `/object-detail`, `/item/[id]`, with `(group)` folders stripped — linked to the component it renders. Calls like `router.push('/object-detail?…')`, `router.navigate({ pathname: '/item/[id]', params })`, template-literal hrefs, an href held in a local `const`, and `router.push(await pickRoute())` where the helper returns screen paths (one edge per screen it can return) resolve to the screen they open as a new `navigates` edge that remembers the href, so "where does tapping this go" and "who opens this screen" are one hop in `codegraph_explore`, `callers`, and the viewer's Flow strip instead of a dead end at a string. Re-index after upgrading to pick the new edges up.

- **Read your graph in a browser: `codegraph ui`.** Point it at a project you've already indexed and it opens a viewer for it on your own machine. Pick a symbol and you see who calls it on the left, its real source in the middle with a marker on every line that calls something, and what it calls on the right, each one drawn level with the line that calls it. Hover either end and both light up; click anything to step into it. Test callers fold into a single line so real callers stay in view, edges CodeGraph isn't confident about are folded away as "uncertain" rather than shown as fact, and a symbol no test reaches within three caller hops says so on a badge. A blast-radius strip counts what a change would reach. Search with `/` or Cmd-K across every symbol and file, start from suggested entry points (routes, hubs, files that run code when imported), and follow a trail of the path you walked that lives in the URL, so you can send someone the exact route you took. Click any file path for that file's outline in source order between everything it depends on and everything that depends on it.

  Run `codegraph ui` in an indexed project, or `codegraph ui /path/to/project` for one indexed elsewhere (`codegraph web` is an alias). It takes port 4747, or the next free one; `--port <n>` pins a specific port and `--no-open` just prints the URL for a headless box or an SSH session. Set `CODEGRAPH_BROWSER=<command>` to choose the browser, or `CODEGRAPH_BROWSER=none` to never open one.

  The viewer listens on `127.0.0.1` only, so nothing on your network can reach it, and requests claiming to come from any other host are refused. It opens an index that already exists, never creates one, and never changes your graph or a line of your code — the one thing it writes is a trail you asked it to save (see below), and `--read-only` turns even that off. It sends nothing anywhere.

- **A map of the whole project, in `codegraph ui`.** The Map tab draws your repository at module granularity — one box per directory — with dependencies pointing down, so the top of the picture is what runs first and the bottom is what everything else stands on. Nothing is placed by hand and nothing floats: a module sits one layer above whatever it depends on, line weight is how many calls, imports and type references cross the link, and the same project always draws the same picture. Hover a link for what crosses it, including the busiest symbol pairs behind the weight; click a module to isolate its links, list its dependencies and dependents with counts, and jump straight into one of its files.

  It says what it leaves out. Links carrying only a handful of references stay hidden until you select a module they touch, references CodeGraph isn't confident about are excluded from every count on the screen and the number is printed, and mutual dependencies, module loops and circular imports between files are listed rather than straightened away. The vertical order rests on the dependencies your code actually writes down — imports, qualified names, inheritance, typed receivers — because a method name shared by two unrelated folders should not be able to move a box.

  It opens on your project's source directory; a picker switches to any other top-level folder or the whole repository, a checkbox brings tests in, and `depth` splits a large folder into its sub-folders — useful on a monorepo. What you're looking at lives in the address, so the view is shareable.

- **Ask how one symbol reaches another, in `codegraph ui`.** Type "how does execute reach getFile" into the search box — or `execute -> getFile` — and the Flow strip draws the call path between them, left to right, one card per hop. Each card is opened at the exact line that makes the next call rather than at the top of the function, so reading the strip is reading the handful of lines that actually carry the work; the identifier being called is a link, and clicking a card opens it in the symbol screen with the trail already set to the path you've read so far.

  A dashed link is a hop nobody can see in the source — a callback, an interface dispatch, a React re-render, a JSX child — and it names the mechanism and, where CodeGraph knows it, the exact line the handler was wired at. When a name means several definitions, the strip says so and names the one the path runs through, offers the alternatives in a picker, and can draw them together as one branching diagram.

  The **"Read as flow"** button on the trail turns a walk you did by hand into the same strip. It is the same path finder `codegraph_explore` leads its answers with, so the picture and what your agent tells you can't disagree.

- **When a path runs out, the Flow strip says where — and why.** A flow that doesn't reach what you asked about now ends in a small block: *"Where the graph stops."* It names the kind of dispatch that ended it — a computed member call, a `getattr`, a reflective invoke, a typed message bus — and the line it's on, and the card beside it opens at that exact line so you can read the code the block is talking about. Where the key is written in the source (`handlers['save']`) it shows the key and shortlists the symbols that could be on the other side, marking any you already named; where the key is a runtime value it says so rather than guessing.

  It also lists what CodeGraph chose not to follow: name-only matches it wasn't confident enough about, with their confidence, and a count of the other calls the symbol makes that this path didn't need. Nothing is invented — no edge is guessed and none is added to your graph — and a flow that does reach what you asked for never shows the block at all. It's the same finding `codegraph_explore` announces to your agent when a flow breaks, so the screen and the answer agree.

- **Read a whole file, with its call graph in the margin, in `codegraph ui`.** The file screen gained a **Source** tab: the file itself, top to bottom, with the same gutter markers as the symbol view and the same right-hand list of what each line calls, positioned level with the line that calls it. A 6,800-line file scrolls at full speed — only the lines on screen are ever drawn, and the text pages in behind you while the markers are there from the first frame.

  In the left margin is an arc for every call that stays inside the file, drawn from the calling line to the line the callee is defined on. Nothing is laid out by an algorithm — the author already put the symbols in order, so source order does the work, and this is the one place a file's internal call structure is legible at a glance. Hover a line to light the arcs the function under your cursor takes part in, and click an arc to jump to the other end. On a file with more than forty of them the picture narrows to the symbol you're reading instead of drawing a wash of overlapping sweeps, with the total in the header. A rail on the far left lists the file's symbols and follows you as you scroll, when the window is wide enough for it.

- **The viewer keeps up with your project while it's open.** Save a file and `codegraph ui` says so within about a third of a second: a banner on the file's screen explaining that the index hasn't caught up yet, and the file's **current** source in place of a body sliced at line numbers it no longer has — the call arcs, gutter markers and call list go away with the old numbering rather than pointing at the wrong lines. It's the same answer `codegraph_node` gives your agent about a file that changed after its last sync.

  When anything re-indexes the project — your agent's background sync, `codegraph sync`, a git hook — whatever is on screen re-reads the graph and a small "Index updated · reloaded" note appears. A symbol that moved because you added a line above it is followed to its new place, with your trail intact, instead of turning into a dead link.

  Nothing polls: the viewer watches for these two things and is told about them. If it loses touch with the server it retries a few times with a growing delay, then stops and says "Not live" in the top bar rather than hammering a port that isn't answering.

- **Take a flow or a map with you: copy it as an image, or save it as an SVG.** The Flow strip and the Map both gained **Copy image** and **Download SVG**. Copy image puts a PNG on your clipboard, ready to paste into a pull-request comment or a chat — the fastest way to say "here is what your change actually touches" without asking anyone to install anything. Download SVG saves a file for a README: it is real text rather than a bitmap, so it stays sharp at any size and the symbol names in it are selectable.

  Both render the light theme whichever one you are reading in, because the image is going to be read on somebody else's screen, and both carry a caption saying what the picture is. What comes out is exactly what is on screen — the same hops, the same dashed dynamic-dispatch links with their wiring sites, the same "where the graph stops" block, the same modules dimmed or brought forward by your selection — because the image is drawn from the same measurements the screen is, not photographed off it. An eight-hop strip comes out around half a megabyte, well inside what GitHub takes inline.

- **"Where does anything start?" has a screen now.** The **Entry points** tab in `codegraph ui` (or press `e`) is the first thing worth opening on a codebase you have never seen. Every route with the symbol that serves it and the `file:line` you will find it at, grouped by the file the URL is registered in — your router, not your handlers — and headed with the framework CodeGraph detected it from. Under that, the files that actually *do* something when they load (a CLI, a worker entry, a build script), the tests ranked by how much of the project each one exercises, and the symbols the most code depends on.

  None of it is guessed from a filename: a file "runs something" because the graph recorded a call from the file itself, and a project with fewer than three routes simply has no Routes section rather than an empty one. Every list says how much of itself it is showing, and says "at least" wherever the real total can only be a floor.

  Any row that names a symbol can start a **flow**: press `Flow ›`, then type a second symbol or press `→ here` on another row, and you get the path between them — so "how does `POST /v1/payroll/cycles/{cycleID}/run` reach the database" is two clicks. Typing into the search box now finds entry points too, under their own heading below the symbol matches, so a URL comes back with its handler attached instead of on its own.

- **Syntax colouring in `codegraph ui` now comes from CodeGraph's own reading of your code.** The viewer used to run a second syntax highlighter over source CodeGraph had already parsed, with its own separate set of grammars. It doesn't any more: the colouring is taken straight from the parse that built your graph, so a file is coloured by exactly the grammar that decided what its symbols are. Three things you will notice — the name a definition declares now stands out on the line that declares it, wherever it appears; calls written inside a string (`${user.name()}`, `#{...}`, `$"{...}"`) are read as code and are now clickable links like every other call site; and built-in type words such as `string`, `int` and `void` look the same in every language instead of one way in Go and another in TypeScript. A big file paints far faster, most visibly in TypeScript, which was by a wide margin the slowest before.

  Two formats change for the worse and it is worth saying so: Liquid, Razor, YAML, Twig, XML and `.properties` files are shown without colouring now, and in `.svelte`, `.vue` and `.astro` files the `<script>` block is coloured but the surrounding markup is not. Nothing about navigation changes there — call sites in those files still link, exactly as before.

  This also takes about 3 MB of grammar files and two dependencies out of the install.

- **The viewer's screens are now a component library other tools can render.** The Symbol view, the Flow strip and the Map are packaged as `@colbymchenry/codegraph-ui` — the same components `codegraph ui` draws, not a copy of them — so another application can show you a symbol's callers, a call path or your architecture over its own copy of the graph. Everything a screen knows arrives through one small interface it is handed, so the tool doing the rendering decides where the data comes from and where a click goes; a design-token stylesheet ships with it so the screens can be themed to match whatever they are embedded in. It is versioned with the engine, so the reader and the graph it reads always match.

  Nothing changes for `codegraph ui` itself — it is the same viewer, now the library's first user. The package is prepared, not yet on npm.

- **See what a type is built on, and what is built on it, in `codegraph ui`.** Open a class, interface, struct, trait or enum and a small tree now sits above its members: what it extends and implements, going all the way up rather than stopping at the direct parent, and everything that extends or implements it, going down. Inheritance is drawn with a solid line and implementation with a dashed one, so the two never read as the same relationship, and every row opens the type it names.

  For an interface, the list below it is the answer to a question source code cannot give you: a call through that interface can land on any of them, and where there are enough of them to make a static answer meaningless the block says so in a sentence. Go's implicit interface satisfaction is included — a struct that satisfies an interface without either file mentioning the other appears in the fan, marked as matched by CodeGraph rather than written down, along with the line it was matched at. Long fans fold behind a "+N more implementations" button rather than being cut off, and if there is more below than was walked the tree says that too.

  Members that redeclare something from a type above are marked in the outline ("overrides Base", or "satisfies Clock" for an interface), so a 40-member class shows at a glance which parts are its own and which are a contract it is filling. The number of implementations shown here is the same number `codegraph_explore` reports to your agent when it announces an interface dispatch.

- **Find the code nothing reaches, in `codegraph ui`.** A new **Dead code** tab lists the symbols no import, call or reference anywhere in your project reaches — biggest first, grouped by the file they live in, with the number of lines each one would take with it. A class nobody uses brings its methods along as a single finding rather than eleven. Every row opens the code.

  The screen is built to be believed rather than to look impressive. A line above the list says, and keeps saying, that this means "no static reference in the index" and not "unused" — reflection, a framework registry and a template can all reach code a graph cannot follow. Under the list, every reason a candidate was left off is printed with its count, so you can see the list is twenty findings out of two and a half thousand candidates rather than twenty out of twenty-one.

  Those exclusions are the feature. Anything exported, or declared in a header, is off the list by default, because something outside your repository can import it — one switch adds them back, with a warning band. So are test and generated files, abstract and interface declarations, anything a decorator registers, members that override something further up, names the language calls for you (`constructor`, `__enter__`, `main`), vendored directories, and files nothing in your project reaches at all. Two more rules catch what the graph itself missed: a name CodeGraph failed to resolve somewhere is never called unreferenced, and neither is a name shared with a symbol that *is* used — the twin may simply have been picked instead. Last, before any row is shown, the files that could reach it are read and the identifier counted: written down twice, something uses it and we did not see it.

  On the **Map**, a module nothing depends on now says so in its own count line instead of counting itself — usually your entry points, sometimes something you forgot to delete. Tool-generated files and modules are dimmed wherever they appear: on the map, in its file list, in search results and on the file screen.

- **Keep a walk you want to come back to: saved trails in `codegraph ui`.** Follow a path through the code, press **Save trail** on the trail bar, give it a name, and it's kept. Saved trails are listed on the empty screen and on the Entry points tab, above the suggestions — a walk somebody named beats any ranking — and opening one puts you back at the symbol you left with the whole path in the trail bar. Explaining "how a request is served" to a new teammate is now a link and a name rather than a paragraph.

  A saved trail survives your project changing. Each step is remembered by what it is — its qualified name, its kind, the file it was in — rather than by where it sat, so editing the file above a function doesn't lose it. When something does move, the trail says so on its own row: which step moved to another file, which one was renamed or deleted, and which part of the walk still opens. It never quietly stitches over a missing step, because a trail is a path and a step that skips one would show a call that doesn't exist.

  Trails are plain JSON, one file per trail, under `.codegraph/ui/trails/` — already ignored by git, so they stay yours by default. **Export** hands you the file if you'd rather commit one for the team. This is the only thing the viewer writes: it still never indexes, never changes your graph, and never touches a line of your code. Start it with `codegraph ui --read-only` and it won't write even that — saved trails can still be opened, just not saved or deleted.

### Fixes

- Spring mappings now include every declared path combination and resolve constants declared in the same file, while unresolved paths no longer appear as false root routes. (#1461)
- `codegraph callers`, `codegraph callees` and `codegraph impact` now resolve qualified names, group results and JSON edges by definition, and accept `--file` to narrow ambiguous names; thanks @ferrine. (#1512, #1656)
- `codegraph callers`, `codegraph callees` and `codegraph impact` (CLI and MCP) now report missing names with did-you-mean suggestions instead of another symbol's results, and exact matches with no callers stay empty; thanks @uvmplus. (#1473, #1481)

#### MCP / indexing

- The prompt hook no longer injects unrelated projects when run from your home directory or a broader directory containing a stray workspace manifest. (#1454)

- Indexing now succeeds when Node.js's SQLite lacks FTS5, with search falling back to name and fuzzy matching; thanks @aniruddhaadak80. (#1532)

- `codegraph_explore` now makes clear that suggested call counts are advisory, so agents keep exploring when an answer is incomplete; thanks @rongbc. (#1504, #1570)

- C++ functions following anonymous namespaces containing raw-string templates are now indexed correctly, even when template text resembles an unfinished macro call. (#1505)

- Indexing now warns when parser errors leave a file with no symbols, including C++ raw strings with 16-character delimiters, so missing code is no longer silent. (#1522)

- `codegraph index <path>` now refuses uninitialized paths and names the nearest initialized parent instead of silently rebuilding it; thanks @danusha2345. (#1524, #1689)

- Sync now recovers the same connections as a clean index after interrupted reference resolution, including inherited calls and callbacks that previously stayed missing. (#1577)

- `codegraph_explore` now re-serves source to fresh subagents and after context compaction, with cross-call dedup available only through an explicit `CODEGRAPH_EXPLORE_DEDUP=1` opt-in; thanks @danusha2345. (#1620, #1624)

- **Watcher scope now matches `git ls-files --exclude-standard` (#1728).** `buildDefaultIgnore` / `buildScopeIgnore` read `.git/info/exclude` and `core.excludesFile` (not only the root `.gitignore`), and seed directories git reports as ignored-untracked so nested `.gitignore` effects prune the live watcher the same way the indexer skips them. Single-file auto-sync was already incremental (`pendingFiles` → scoped `sync({ paths })`); the remaining gap was watching trees git had excluded.

- **Live sync no longer lets the write-ahead log grow without a bound when a reader is holding it open (#1539).** Incremental sync now uses the same writer pause that full indexing already used, and if checkpointing still cannot finish once the log is past its documented size limit — typically because the query pool is reading at the same time — sync stops with a clear error instead of keeping writing until the disk fills. The previous behaviour could leave a multi-tens-of-gigabyte log beside a few-gigabyte index on a large project. Close concurrent readers and retry, or raise `CODEGRAPH_WAL_VALVE_MB` if the limit is too tight for the project.

- **A second `codegraph serve --mcp` on the same project no longer silently kills auto-sync (#1740).** Direct mode (`CODEGRAPH_NO_DAEMON=1` or proxy→in-process fallback) now takes an exclusive `.codegraph/writer.pid` lock; a second writer exits immediately with guidance to stop the other server or unset `CODEGRAPH_NO_DAEMON` so clients share the daemon. The shared daemon already multiplexes N clients onto one watcher — this closes the same-OS dual-direct gap the docs warned about for Windows/WSL but did not guard.

- Indexing no longer checks whether files outside your project exist. A relative import that points above the project directory (`../../something`) made CodeGraph probe that location on disk while resolving it. Nothing outside the project was ever read, and no such file was ever added to the index or linked to, but the check itself should not have happened — such an import now simply resolves to nothing. Symlinks inside your project that point at code kept elsewhere are unaffected and still index as before. Thanks @ErQrYfkrju. (#1631)

- `codegraph install` now honors `CLAUDE_CONFIG_DIR` and `CODEX_HOME` for global Claude Code and Codex setup so CodeGraph loads in your chosen profile (thanks @seanchann; #1627).

- Files opted in with `includeIgnored` now stay indexed on Git older than 2.36, and embedded repositories remain visible to the watcher (thanks @maxmilian and @newshowardz777; #1549).

- `codegraph init` and `codegraph index` now list unsupported file extensions and explain that CodeGraph is inactive when no supported source files are found (#1502).

#### Screens, links and navigation

- **A `<Card/>` is the Card that file imports, not the first one in the repo.** When two components share a name — `Section`, `Picker`, `FrameCard`, one per feature folder — the render link out of a component picked whichever happened to be indexed first, ignoring what the file actually declares or imports. That cost twice over: a component gained a link to something it never renders, and the one it does render was left with nothing pointing at it, so callers, impact and the Screens tab all dead-ended there. In one Expo app it left a folder sheet's navigation standing alone on Screens with no screen behind it, while the link pointed at an unrelated card in another sheet; in Excalidraw every `<Excalidraw/>` in the app pointed at a documentation preview file instead of the component itself. A tag now resolves to the component that file declares, then to the one it imports — through a `@/` path alias as readily as a relative path — and a JavaScript tag prefers a JavaScript component over a same-named class in a mobile app's native half. No links are added or removed, only pointed at the right place. Re-index after upgrading.

- **Where the app goes after login is a fork, not two always-es.** A navigation whose destination comes back from a helper — `router.replace(await resolvePostLoginRoute())` over `return (await hasSeenWelcome(…)) ? '/home/' : '/welcome/'` — drew both screens with no condition, reading as if the welcome screen always shows. The two arms share a line, and only a column can tell them apart; each synthesized edge now carries its literal's own position, so the guard reader says which arm it is: `WHEN await hasSeenWelcome(…)` → home, and its negation → welcome. And the scan starts at the helper's body, so a literal-union return type — `Promise<'/welcome/' | '/home/'>`, whose routes are string literals too, written first — no longer stands in for the navigation itself. Re-index after upgrading to pick the positions up.

- **A screen that talks to native code keeps its own navigations.** In a React Native or Expo app, a `router.push` written inside a listener for a native event was credited to whichever screen had *started* that round trip, not to the screen the push is written on. In one app that moved seven transitions off the capture screen and onto the review screen it opens — leaving the review screen looking as though nothing in the app could reach it, stranded in the "no transition reaches this" band at the bottom of the Screens tab, and printing Swift conditions like `Thread.isMainThread` on a JavaScript navigation. A navigation now belongs to the screen whose file it is written in; an event arriving from native code, from an HTTP call or off a queue is no longer read backwards as if it were a caller.

- **A link written under a condition says so on the Screens tab.** A checkout stepper whose tabs are each enabled by their own prop, and a navbar whose admin links only render for an admin, both read as **always** — every transition written in markup was drawn with no condition at all, while the ones written as calls carried theirs. They are read the same way now: a store's checkout tabs say `step1` … `step4`, its navbar says `userInfo && userInfo.isAdmin` for the admin links and `!userInfo` for sign-in, and 59 of that store's 74 transitions carry the condition they actually run under, up from 20. A template language with no condition rules of its own still says nothing rather than guessing.

- **A link in markup no longer reads as a helper's return value.** `<Link to='/shipping'>` was labelled `return /shipping`, which in this picture means the destination came back from somewhere else and was inferred. It is written right there, so it now reads `link /shipping` — and an internal `<a href>` reads `a`. Only a destination that genuinely arrives from elsewhere still says `return`.

- **A link that goes to one of several places now draws all of them.** A destination written as a choice — `!isAdmin ? keyword ? \`/search/${keyword}/page/${x}\` : \`/page/${x}\` : \`/admin/productlist/${x}\`, which is how a paginator shared between a storefront and an admin list is written — used to draw nothing at all, because one edge carried one destination and picking an arm would have been a guess. Every arm is now its own transition, labelled with the path THAT arm takes, so a store's paginated addresses are on the map instead of sitting there unreachable. The same goes for `redirect(307, user ? \`/profile/@${user.username}\` : '/login')` in a loader, and for `router.push(cond ? '/a' : '/b')`. Arms that name the same route still make one transition, and an arm nothing can read still contributes nothing.

- **API endpoints are no longer drawn on the Screens tab.** A store's thirty Express endpoints sat beside its nineteen pages as boxes nothing navigates to and nothing leaves — in a picture that is only about navigation — and they stretched the row of unreachable pages hundreds of boxes wide. A screen is now a route named by its path alone; a route named with the method that reaches it (`GET /api/orders`, `POST /api/users/login`, `ANY /api/users`) is a request, not somewhere a user can be. Every route still appears on Entry points, which is the list of everything a request or a user can arrive at.

- **A screen you could reach but never leave.** Three separate things left a page's own navigation off the map, and a store's home and sign-in pages showed nothing leaving them. **One component, several addresses:** a screen rendered at more than one route — a listing page that is also the search and the paginated results — handed all of its navigation to whichever route happened to be written first, and the rest were drawn as dead ends; every address it serves now gets it. **A link written as a choice:** `<Link to={redirect ? \`/register?redirect=${redirect}\` : '/register'}>` is how a link that carries state is written, and markup was read by a weaker reader than calls were, so it saw nothing; both now use the same one. **A destination whose other half is unknowable:** `const redirect = location.search ? location.search.split('=')[1] : '/'` followed by `history.push(redirect)` is how every app sends a user on after signing in — the `/` is where it lands by default, and reading neither half lost the whole transition. Where both halves ARE readable and disagree, it is still a fork and still nothing.

- **A conditional inside a conditional no longer reads the wrong arm.** A paginator written `!isAdmin ? keyword ? '/search/…' : '/page/…' : '/admin/…'` was split at the first `:` rather than the matching one, so an admin's page links pointed at the storefront's pagination. The arms are paired properly now, and a three-way choice — which is more destinations than one link can name — is left alone.

- **In a repository with several apps, a link no longer points into a different one.** Every app has a `/` and most have a `/login`, and the route table was built for the whole repository at once, so whichever app was indexed first claimed each address — a `<Link to="/posts">` in one app resolved to another app's `/posts`. Measured on a monorepo of 477 apps: **82% of navigations pointed at a route belonging to a different app**, and all of them now point within their own. Screens transitions are also no longer attributed to an unrelated page when several routes are declared in one file, as a code-based route tree or an Express router file is.

- **A server action written through a wrapper starts its transition on the right page.** `export const signIn = validatedAction(schema, async (data) => { … redirect('/dashboard') })` — the arrow inside is no symbol of its own — now belongs to `signIn` on the Screens tab, and `signIn` is attributed to the page whose component hands it to `useActionState(signIn, …)`, read from the source when the graph holds no such edge.

#### Steps — what a call does, and when

- **A handler called from under a binding says what it passes.** A press that runs `tryCatchSync(onClosePress)` drew a box for the wrapper and stopped — leaving the one thing a reader asks ("what is being wrapped?") unsaid, even though every other call-shaped site already prints its arguments. The panel and tooltip now say `tryCatchSync(onClosePress)` — the argument is the answer.

- **A step the walk stopped at keeps its whole name.** A boundary — another screen, or a cap the walk hit — ends its name with an ellipsis by design, but the box was not sized for it, so a longer name lost its last letters instead (`/scan-to-verif…` for `/scan-to-verify …`). The anchor's start mark clipped a long path the same way (`/sheets/forgot-passw…`). The box now makes room for both.

- **A server action, or any handler written inside a wrapper, draws what it really does.** `const signIn = validatedAction(schema, async (data) => { … })` showed one call out of nine, because the constant held a reference to its schema and that counted as having a body of its own. Its picture is now whole — the lookup, the early returns, the `Promise.all`, the redirect.

- **A call is no longer followed to a same-named method of its own class.** `crypto.createHash('sha256').update(…)` inside a service that happens to have an `update` method was followed into that method, so a login endpoint read as though it updated the user, extra replies and all. A method of your own class is written `this.update(…)`; a receiver that is not `this` now ends the walk instead of guessing.

- **A step is ordered by the call it was actually reached at.** An endpoint whose handler is written inline at the registration (`router.post('/users/login', async (req, res) => { … })`) had every call in the handler read as part of the registration itself, which put its reply before the work that produces it.

- **A client call with a type argument binds too.** `useSWR<Team>('/api/team', fetcher)` and `ky.get<User>('/api/users/1')` reach their routes like the untyped forms.

- **A row reads in the code's order.** The boxes one step away from a handler now sit left to right as the code runs them — the database read, the token signed while the reply is built, the reply, the error reply — instead of in an arbitrary order; a call written inside another call's arguments (`generateToken(…)` in `res.json({ token: generateToken(…) })`) comes before it, and its link says `inside res.json(…)` in the panel and the tooltip.

- **Each way an endpoint answers is its own box.** A handler that replies `200` or `401` now draws two reply boxes instead of one `200 · 401`, so each line from the handler carries its own condition on the picture — `→ user && (await user.matchPassword(password))` into the 200, `→ NOT (…)` into the 401 — the way a screen's transitions do; replies whose status the code does not spell out share one box labelled by the call.

- **A reply that sets no status is a 200.** `res.json(user)`, `res.send(…)`, `reply.send(…)`, `NextResponse.json(…)`, a `JSONResponse` or `jsonify(…)` with no status in the chain now count as `200`, so an endpoint's success reply has a box of its own beside the `401`'s and its row carries its code; a status set by the statement just before (`res.status(202); res.json(user)`) is that reply's. An Express handler written inline at the registration keeps its own replies too — they were filtered out with the framework noise.

- **The Steps walk stays out of test doubles.** A production handler that calls an interface method is no longer followed into `TestUserDataRepository` (or any implementation in a test folder); the counts in the summary no longer say "outside the indexs".

- **Conditions read as words, and each scenario gets its own row.** In the Screens and Steps tabs a condition is now written the way you would say it — `WHEN NOT (isUploadInProgress || elapsed < 5000) AND user?.organization_id`, the joining words set apart from the code — with the code inside each guard left as written, and a transition or step with several call sites (four early returns that each go home) is listed as four rows, the clauses they all share said once above them, instead of one string joined with `||`. A guard that is itself an either/or keeps its parentheses everywhere conditions are shown, including `codegraph_explore`'s Flow section.

- **Every call now says when it happens.** In `codegraph ui`, a symbol's callee and caller rails and the Flow strip's connectors carry the branch conditions the call site sits under — `when !isUploading && isCollected` — and `codegraph_explore`'s Flow section prints the same on each hop (`↓ calls (when isCollected)`). The conditions come from the `if` / `else` / ternary / `switch` / `&&` branches around the call, the early returns before it (`if (busy) return` reads as `!busy`), and Swift's `guard`; an inline callback inherits the conditions of the place it is defined. Read from the source as it is now, never stored: nothing about your index changes. TypeScript, JavaScript and Swift today.

#### Routes and framework detection

- **A SvelteKit layout is no longer a second screen at a page's address.** `+layout.svelte` and `+error.svelte` sit at the same path as the `+page.svelte` beside them and were each indexed as a route, so one address appeared in the index two and three times over. Only a page is a route now.

- **A framework whose package lives in a subfolder is detected again.** In a project that keeps its dependencies one level down — a `frontend/` and a `backend/`, or an `apps/web/` — the framework check ran once before any file had been indexed, found no folders to look in, and remembered that empty answer for the rest of the run. Every React, React Router and Next.js behaviour that depends on knowing the framework is there silently did nothing for those projects.

- **Express routes behind `app.use('/api', router)` are named by the path a request takes.** A router mounted at a prefix — through as many `router.use('/users', usersRouter)` levels as the app nests, by import or `require` — now names its routes `POST /api/users` instead of `POST /`, and the chained form `router.route('/:id').get(getProduct).put(protect, updateProduct)` (split across lines or not) registers one route per method. Entry points, the Steps tab and the client-to-route pairing all read the real paths.

- **An Express handler written through a wrapper is the endpoint's handler.** `const authUser = asyncHandler(async (req, res) => { … })` — the `express-async-handler` idiom — now names the handler in Entry points and starts the Steps walk at it, with the database reads, the token check and the `401` row read from the arrow's body.

- **A framework declared in a workspace's `package.json` is detected.** React, Next.js, Expo Router, Express and NestJS are found when their dependency lives in `frontend/`, `backend/`, `apps/web/` or `packages/api/` rather than at the repository root, so a monorepo's pages and endpoints exist in the index.

- **FastAPI routes are named by the path a request takes.** `APIRouter(prefix="/items")` and a literal `app.include_router(router, prefix="/api/v1")` — nested through an aggregate router, through a module import or an alias — now compose onto the route names (`GET /api/v1/items/{id}` instead of `GET /{id}`); a prefix that is a setting rather than a string leaves that mount alone rather than guessing.

- **ASP.NET Minimal API endpoint groups are routes.** The handler-first form — `groupBuilder.MapPost(CreateTodoItem)`, `MapPut(UpdateTodoItem, "{id}")` inside an `IEndpointGroup` / `EndpointGroupBase` class (the Clean Architecture template and its descendants) — now registers `POST /api/TodoItems` and `PUT /api/TodoItems/{id}`, with the `/api/` head read from the app's own `MapGroup($"/api/{groupName}")` and a class's `RoutePrefix` honoured, each bound to its handler so the Steps tab starts there and lists its `TypedResults` replies by status code.

- **A FastAPI service that lives in one directory of a monorepo is detected.** `backend/pyproject.toml` and `backend/app/main.py` count, not only files at the repository root — the official full-stack template's routes now appear in Entry points and the Steps tab.

#### Symbols, tests and the viewer

- Kotlin functions and methods now carry their signature — `(params): ReturnType` — in `codegraph_explore`, `node` and the viewer, instead of no signature at all. Re-index Kotlin projects after upgrading. (#1495)
- TypeScript/JavaScript value aliases — `export const alias = fn`, `export { fn as alias }`, object-literal `api = { run: fn }`, and same-file `const local = fn` — now forward calls edges to the aliased function, so callers and impact on the implementation include consumers that call through the alias instead of stopping at the binding. Genuine wrappers (`() => fn()`) are unchanged. Re-index after upgrading. Thanks @valkyriweb. (#1482, #1485)
- `codegraph affected` now finds Go, Python and JVM test files that previously went unreported, while preserving custom `--filter` behavior (thanks @danusha2345; #1507, #1688).

- Calls inside declaration initializers in Kotlin, Java, TypeScript, JavaScript, Scala, Rust and Python now appear under the declaration that owns them, making callers and impact results more accurate after re-indexing with `codegraph index -f` (thanks @danusha2345; #1510, #1511).
- Java fields initialized with anonymous classes now expose their methods and calls in the graph.
- Kotlin property accessors, initialization blocks and destructuring declarations now retain their calls with the correct owner.
- The viewer continues to count module-level initializer calls as top-level file activity in entry points and file screens.
- `codegraph_explore` again lists a dynamic-dispatch link when the same two symbols are also joined by an ordinary call.
- Rust unit structs (`struct Unit;`) and their trait implementation relationships now appear in the graph after re-indexing. (#1513, #1514)
- Imports from Node built-ins or npm packages no longer connect to unrelated type members with matching names; re-index after upgrading to clear existing false dependencies. Thanks @ctype-lab. (#1537)

- Inheritance relationships no longer attach external Rust or npm supertypes to unrelated local symbols with the same name, including in Svelte, Vue and Astro components; re-index after upgrading to clear existing false relationships. Thanks @ctype-lab. (#1536)

- PHP static calls through imported class aliases now reach the correct class when services and repositories share method names, so callers and impact analysis show the right dependencies after re-indexing. (#1545)
- TypeScript/JavaScript: a call through a field of the enclosing class — `this.mailer.send()` — now resolves on the field's declared type, so a delegating wrapper that shares the method's name no longer records itself as its own callee and `callers`, `impact` and trace stop lying on that shape. A field whose type is external or a builtin stays unresolved rather than guessed. Re-index after upgrading. (#1496)
- TypeScript and JavaScript collection calls through local variables and their nested properties no longer link to unrelated project methods; re-index after upgrading. (#1566)

- Objective-C headers now index in a project that has no `.m` file. A `.h` file is read as C from its name alone, and only later — once its contents are read — recognized as Objective-C; the grammar for that was never loaded up front, so the file failed with a parser error and nothing in it reached the index. Adding any `.m` file used to make the same header work, which is what made this look arbitrary. Thanks @Juddd. (#1628)

- TypeScript interface methods and properties are now indexed, so `node`, `callers` and impact can find platform `.d.ts` APIs while declaration-only files keep their lower ranking on flow queries; re-index TypeScript projects after upgrading. (#1638)
- Lua and Luau function expressions assigned to locals, table members, or keyed table fields are now indexed as callable nodes. Calls from `local f = function() ... end`, `M.f = function() ... end`, and callback tables such as `M.handlers = { onClick = function() ... end }` are attributed to the named function or method instead of collapsing onto the file node, so callers and impact no longer omit these handlers. Re-index after upgrading. (#1616, #1650)
- **Functions bound with `const` inside another function are symbols now.** `const handleClear = () => {…}` inside a React component — every handler that skips `useCallback` — was invisible to `callers`, `callees` and impact, answering "Symbol not found" exactly the way a function with no callers would. It is indexed like its module-level twin, contained by the enclosing function, with its own calls. Re-index after upgrading. (#1669)
- `codegraph callers`, `callees`, and `query` now clearly report when their result limit hides additional matches, including exact totals in callers/callees JSON output; the `codegraph_callers` and `codegraph_callees` MCP answers carry the same "showing N of M" note. (#1639, #1674)
- CommonJS controllers written as `exports.getItems = async (req, res) => {…}` or `module.exports.x = function () {…}` are now indexed as exported functions, so `node`, `callers` and impact find every Express handler in that style and the calls inside them belong to the handler instead of the file. Re-index JavaScript projects after upgrading. (#1675)
- Python parameters annotated with a quoted forward reference — `def f(o: "Alpha")`, or anything under `from __future__ import annotations` — now resolve the methods called on them, the same as the unquoted annotation. Re-index Python projects after upgrading. (#1684)
- **A C macro call written with designated initializers no longer swallows every function after it.** Betaflight resets each config struct with `RESET_CONFIG(type, dst, .field = value, …)`, a shape the C grammar cannot parse; past a hundred or so fields its error recovery ran the enclosing function to the end of the file, the next function vanished from the index and every later one was filed under the first, where name matching then treated it as an unreachable closure. The argument list of such a call is now blanked before parsing, offsets kept, so the file's functions come out with their real extents. On that tree 45 functions in `pid.c` alone moved back to top level and their 117 callers resolve at exact-match confidence. Re-index after upgrading. (#1729)
- A method called on the result of another call — `d.setdefault(k, []).append(v)`, `make().run()` — no longer produces a call edge to an unrelated top-level function that merely shares the name, in Python and JavaScript/TypeScript. The receiver is kept so the inner call still resolves; the outer method stays unresolved rather than guessed. Re-index after upgrading. (#1683, #1681)
- A Python call through an imported project module whose name collides with a builtin collection method — `ledger.append(row)` after `from . import ledger` — is no longer dropped as `list.append`. The builtin-method filter now lets the receiver through when it is an imported module that resolves to a file in the project, so `resolveViaImport` can attach the real edge; a stdlib/PyPI receiver (`os.remove`) still produces none. Re-index after upgrading. (#1681, via #1704)
- Python method calls on module-scope builtin collections no longer create false calls or file dependencies to unrelated project methods with the same name; re-index after upgrading. (#1652)
- **A definition its language makes file-local no longer captures calls from other files.** A C `static` in another source file (`.c`/`.cc`/… — not a header's `static inline`, which is textually included), a Kotlin/Java/C#/Swift/Scala/Dart/PHP `private` member, a Go unexported name in another package, and a Rust non-`pub` item outside its module subtree cannot be what a name in another file means, but name matching accepted them whenever the names agreed: an Android `editor.apply()` onto an unrelated class's `private fun apply`, a JavaScript `fail(...)` onto a Go `func fail`, a Rust `.count()` onto a private `fn count` in another crate, and C USB helpers onto a `static` in a `.c` they never link. Such a target is now declined after the whole name-matching pipeline settles — the reference stays unresolved rather than falling through to a fuzzy namesake. Same-file definitions, a child Rust module reaching its ancestors' private items, and Rust `impl Trait for Type` methods stay resolvable. Re-index after upgrading. (#1730, #1731)
- **A binding in a module that exports nothing is no longer a cross-file target.** On vite, every `import { defineConfig } from 'vite'` across the playground resolved onto a `const vite = await createServer(…)` sitting at module scope in `playground/ssr-html/test-stacktrace.js` — a file with an import and no export, so that binding is reachable from nowhere but itself. Name matching commits as soon as one candidate survives, and nothing asked whether an import could reach the survivor; that one binding took 157 edges. A JS/TS file holding an `import` and no export of any kind now offers its locals to no other file. Classic scripts, CommonJS (including `exports["x"] = …`), a later `export { … }`, and names contributed through `declare global` are all unaffected. Across vite this removed 320 wrong edges and added 18, each addition a reference that was previously ambiguous rather than newly invented. Re-index after upgrading. (#1719)
- **A bare call inside a JavaScript or TypeScript method no longer resolves to the method itself.** When a method and a module-scope function share a name, `serialize(this.raw)` written inside `Record.serialize` means the function, but the nearest same-named definition won the tie and the graph recorded the method calling itself. A call written without a receiver can never reach a method in JS/TS, so methods are no longer candidates for it; `this.serialize()` and `other.serialize()` resolve as before. (#1714)
- **Fuzzy matching no longer lands on a closure it cannot reach.** A function nested inside another function is only callable from inside its container, and exact-name matching already declined such candidates; the fuzzy fallback did not, so a builtin method call (`res.text()`, `items.push()`) whose only same-named project symbol was some file's closure resolved onto that closure. The fallback now checks that the one candidate it would commit to is reachable, and declines otherwise — it does not filter the candidate list first, which would turn a crowd of same-named definitions into a single "unique" survivor and hand it every call of that name. On vite that removes the 12 edges onto nested functions and adds none. Re-index after upgrading. Thanks @bompus. (#1708, #1709)
- **An import that names the emitted extension resolves to its source.** Under `moduleResolution: node16 | nodenext | bundler` TypeScript requires `import { x } from './util.js'` for `util.ts`, and no file of that name exists, so the import resolver returned nothing and every name imported that way fell through to bare-name matching: a method wrapping the same-named helper it imports (`renderDockStyles() { return renderDockStyles(); }`) resolved to itself, and cross-module edges in such projects were name guesses. `.js` / `.jsx` / `.mjs` / `.cjs` specifiers now retry with the source extensions TypeScript compiles from when the emitted file is absent; a real `.js` beside the `.ts` still wins. On a 582-file repo whose `.ts` files import this way, import-backed `calls`/`imports` edges went from 4,002 to 7,312 and the eight wrapper-method self-edges disappeared. Re-index after upgrading. Thanks @bompus. (#1705, #1706)

- **The Map groups a repository the way that repository is shaped.** It always drew top-level directories, so a project whose whole program lives under one `src/` opened as a picture of four boxes — `src`, `ios`, `.github`, `(root files)` — with two thirds of the code inside one of them and nothing to say about it. The Map now picks its own grouping: the shallowest one that is not a single box holding the program, so a mobile app opens on `src/app`, `src/components`, `src/api`, `ios/CaptureView` and the rest, and a project packaged as `frontend/src/…` opens on the screens, components and reducers instead of on the word `frontend`. A repository whose top-level directories really are its modules is left exactly where it was. A new **Grouping** control on the right says which one was chosen and lets you take it a level in or out, and a leaf directory is now named for itself rather than as `…/(root files)`. Each box now also says how much leans on it — how many files elsewhere reference straight into it — with a bar along its bottom edge scaled against the most depended-on box on screen, so the folder you have to be careful with is the one you can see at a glance rather than the one with the longest name. The Map also has a **Key** now, like the Screens and Steps tabs — including what the dashed maroon lines mean, which only appear once you select a module: that module reaching back UP into something that depends on it.

- **The Symbol tab opens the Symbol tab.** With no symbol open and no trail to return to, clicking **Symbol** in the top bar took you to the landing page — which, on any project that has screens, is the Screens tab. So the button said Symbol and gave you somebody else's view. It now has an address of its own (`#/s`) that opens the "nothing selected" screen: the search prompt and the where-to-start list of routes, entry files and the symbols the most code depends on.

- **Files under an `e2e/` directory count as tests.** Their calls no longer appear as production callers in Steps, dead-code and test badges.

- **Production code under a `samples` or `examples` package path is no longer treated as test code.** A Kotlin or Java project whose package path runs through `com/google/samples/…` (Now in Android, for one) had nearly every file counted as a fixture, so the Map opened on `build-logic`, the entry points hid the app, and dead-code and test badges were wrong. Only the project layout above a `src/` folder decides now; the package path below it never does.

- **React handlers written with `useCallback` are now symbols.** `const handleSubmit = useCallback(() => {…}, [])` — the way nearly every handler in a React or React Native component is written — is extracted as a function named by its binding (also `React.useCallback`, `useEffectEvent`), so `onPress={handleSubmit}` and `addListener('x', handleSubmit)` resolve to it, its calls are its own rather than the component's, and a tap's handler shows up in `codegraph_explore`, the Screens tab and the Steps tab. A JSX attribute value (`onPress={handleSubmit}`, `renderItem={renderRow}`) and a handler a hook hands back in an object (`return { handleSubmit, handleRetake }`) are now function-as-value references from the component or hook, so the graph knows which functions are wired as handlers.

- **Stores exported on a later line are read like any other.** `const useStore = create((set, get) => ({ … }))` followed by `export default useStore` (or `export { useStore }`) now has its actions extracted as functions, the same as an `export const` store — previously the two-statement form, common in React Native apps, left every action invisible.

- **API objects exported as a default namespace resolve through to their functions.** `import Api from './api'` + `Api.upload()` where the module ends in `const Api = { upload, createFolder }; export default Api` now links the call to `upload` itself (through the object's own imports), and a default import of any const named by an `export default NAME` statement finds that const rather than guessing the file's first exported function.

- **The Map covers a multi-root project.** A React Native app's `ios/` beside its `src/` — or any second root holding a fifth of the code — is now on the picture, one level deeper, instead of the map silently drawing only the larger root.

- **The Claude Code prompt hook's context now arrives inline.** The hook capped its injection at 16,000 characters, but Claude Code shows hook output inline only up to 10,000 and otherwise persists it to a file with a 2 KB preview, so on any repo where explore filled the cap the model saw a file path and the first 2 KB. The cap is now 9,000 characters, under the limit with room for the wrapper. (#1694)

- **`codegraph_explore` is loaded from the first prompt in Claude Code.** Claude Code defers every MCP tool behind a tool-search step, so a fresh session saw only the tool's name until the model searched for it, and the server's "call `codegraph_explore` instead of Read" had nothing loaded to act on. The tool now carries `anthropic/alwaysLoad` in its `_meta`, which exempts it on existing installs, and `codegraph install` writes `alwaysLoad: true` on the Claude Code server entry (re-run it to add the key). Copilot CLI's tool search holds MCP tools back the same way once ~30 tools are connected, so its entry now carries `deferTools: "never"`. (#1696)

- Fixed a long-running `codegraph ui` session serving a symbol that a sync had already deleted. The viewer keeps one connection to your index open, and its in-memory lookup didn't notice when another process — your agent's sync, or `codegraph sync` — rewrote the file underneath it, so a symbol screen could keep showing a body with no callers while search correctly reported it had moved. Because a symbol's identity includes the line it starts on, this happened after almost any edit above it.

- Python calls and file dependencies through `from package import module as alias` now appear in the graph, so renamed imports no longer hide live callers or imported modules. Thanks @JoeyNPP. (#1626)

## [1.6.0] - 2026-08-26

### Highlights

- **GitHub Copilot is now supported** — `codegraph install` sets it up in VS Code, the Copilot CLI, and JetBrains IDEs, next to the agents it already knew.
- **Set up in one command** — `codegraph install --yes --init` wires up your agents and indexes your project with no prompts, ideal for a fresh container or CI.
- **Better answers for your agent** — `codegraph_explore` no longer repeats code it already showed you, always brings back the files and symbols you asked for by name, and spends its space on the code that actually answers the question rather than look-alikes, generated files, and type shims.
- **Your graph stays right as you keep coding** — a long-running index no longer drifts from a fresh one, and edits to `codegraph.json` (such as `exclude`) apply immediately without a restart.
- **No more silent crashes or hangs** — deeply nested C/C++ files, Swift Vapor projects, and large sync batches that used to kill or stall indexing now finish cleanly.
- **A disk-space leak is fixed** — a force-killed session could leave the database's write-ahead log behind to grow without bound (tens of gigabytes was reported); the leftover log is now folded back into the index and trimmed automatically the next time the project opens. The index itself has no size limit.
- **Works from a workspace or monorepo root** — the MCP server finds your indexed project when launched from a folder above it, and says so clearly when it can't find one.
- **More accurate code graphs** for TypeScript, Rust, Erlang, C/C++, and Python projects — see the full list below.
- **Also new**: per-project Codex setup (`codegraph install --location=local`), a `deprioritize` setting to keep helper folders from outranking your real code, the `codegraph context` command, and usage stats that now stay entirely on CodeGraph's own servers.

After upgrading, run `codegraph index` once in each project so your existing graph picks up these fixes — `codegraph status` reminds you when it's needed.

### New Features

- GitHub Copilot is now a supported agent: `codegraph install` can configure Copilot Chat in VS Code (`copilot-vscode`), the GitHub Copilot CLI (`copilot-cli`), and the Copilot plugin in JetBrains IDEs (`copilot-jetbrains`). Installed Copilot surfaces are auto-detected like every other agent, existing MCP server entries in their config files are preserved, and `codegraph uninstall` reverses the setup cleanly. Restart VS Code or your JetBrains IDE after installing so Copilot picks up the server.

- `codegraph install --init` wires up your agents and builds the current project's index in one command, and `codegraph init --yes` runs without any prompts — so a fresh container or CI job can bootstrap CodeGraph with a single non-interactive line (`codegraph install --yes --init`). The installer still never indexes anything unless you ask for it with the flag, and the usual safety refusal for a home directory or filesystem root applies. (#1578)

- Codex CLI can now be set up per project instead of only user-wide: `codegraph install --location=local` writes `./.codex/config.toml` and the CodeGraph block in your project's `AGENTS.md`, so CodeGraph is wired into that repo only rather than every Codex session on the machine. `codegraph uninstall --location=local` reverses it, and the global install is untouched either way. Codex only applies a project's config once you've marked the project trusted, so the installer says so after a local install. Thanks @maxmilian. (#1531)

- A new `deprioritize` setting in `codegraph.json` keeps the paths you name from outranking your product code in search and `codegraph_explore` answers, without removing anything from the index. It takes gitignore-style patterns just like `exclude`, but is ranking-only: helper-script trees, generated output, or optional add-on directories whose generic symbol names (`usage`, `run`, `status`) would otherwise crowd out the code that actually answers a query stay fully indexed and findable — and a query that genuinely targets such a tree still returns it. Thanks @maxmilian. (#982)

- `codegraph_explore` no longer re-sends source it already returned earlier in the same conversation. A file it has already shown you comes back as a short pointer — the path, the symbols and the exact line range, with confirmation that the file hasn't changed since — and the space that frees is spent on code you haven't seen yet, so a follow-up call covers new ground instead of repeating the last one. If a file was edited in between, its source is always shown again in full. Set `CODEGRAPH_EXPLORE_DEDUP=0` to turn this off.

- When an agent connects over MCP, CodeGraph now states up front that it indexes 30+ languages — TypeScript/JavaScript, Python, Go, Rust, Java, C#, C/C++, PHP, Ruby, Swift, Kotlin, and more — so agents no longer assume a language isn't supported and skip the graph. (#671)

- Anonymous usage telemetry is now stored entirely on CodeGraph's own first-party infrastructure — no third-party analytics vendor receives any of it, and the endpoint that receives it makes no outbound requests at all. Individual events are deleted after 90 days, leaving only anonymous daily totals. Nothing about what is collected changed, your IP address is still never read or stored, and every off-switch works exactly as before (`codegraph telemetry off`, `CODEGRAPH_TELEMETRY=0`, `DO_NOT_TRACK=1`). `TELEMETRY.md` remains the complete field-by-field list.

### Fixes

#### Better answers from `codegraph_explore`

- Naming a file by its path in a `codegraph_explore` query now works reliably: the path is resolved against the index and that file is guaranteed a place at the top of the answer. Previously the path was broken into fragments — bracketed route segments like SvelteKit's `[id]` made this worst — and pieces like `page` or `runs` matched every sibling file, so the file you actually named could be crowded out of the answer entirely. A path that doesn't match any indexed file is now called out instead of silently ignored.

- Naming a kebab-case file **without its extension** in a `codegraph_explore` query — `background-image-table` rather than `background-image-table.tsx`, the way import paths and prose spell it — now returns that exact file too. Previously the name was split at the hyphens, and in a kebab-cased frontend those pieces (`background`, `image`, `table`) are among the most common words in the codebase, so look-alike sibling files filled the answer while the named file never appeared. Hyphenated words that don't name an indexed file, like "cross-call" or "non-blocking", are left alone.

- Plainly-worded `codegraph_explore` questions now find camelCase code: a query like "auto-scroll to bottom" can reach a function named `scrollFeedToBottom`, because query words are matched against the words inside identifiers, not just whole names.

- Variables and constants now count when `codegraph_explore` picks its starting symbols, so state held in plain variables — `$state`-style variables in Svelte, for example — no longer gets overlooked.

- `codegraph_explore` now concentrates its answer on the code that actually answers your question instead of spreading it across files that merely share a word with it, so more of the answer arrives in a single call. Thanks @LeDuyViet for the detailed measurements and reproduction. (#1500)

- Files only weakly related to your question now come back as a name, symbol and line number instead of spending the answer on their source — name one of them in a follow-up `codegraph_explore` to get it back in full. (#1500)

- A generated CRUD or protobuf layer no longer crowds out the hand-written code sitting beside it: generated files are now recognized by the `// Code generated by … DO NOT EDIT.` style banner written at the top of the file, not just by a filename that looks generated. Re-index after upgrading to pick up the new detection. (#1500)

- Test and spec files in a repository's top-level `test/` or `spec/` directory are now recognized as such, so they no longer take room from the code you asked about. (#1500)

- Generated type-declaration files that announce themselves with a "Generated by … by running …" banner — Cloudflare Wrangler's `worker-configuration.d.ts` is the common one — are now recognized as generated. Previously a file like that could take most of a `codegraph_explore` answer on nothing more than a few common words, pushing the hand-written code you asked about out of the response entirely. Re-index after upgrading to pick up the new detection.

- A hand-written type-declaration file — an ambient `.d.ts` of global shims, vendored typings, module augmentation — no longer takes over a `codegraph_explore` answer about how something works. Files like these declare common names (`Body`, `Message`, `ImageMetadata`) and nothing else, so a plainly-worded question could match one strongly enough that it ranked first and crowded the actual handler out of the answer. They are now ranked lower for questions about behaviour, and are still listed by name so one follow-up call fetches them. Asking about a type by name still returns its declaration first, and a shared types module the rest of your code imports is unaffected.

- When you name a symbol in a `codegraph_explore` query, its definition now actually comes back. Two cases previously lost it. If the symbols you named don't call one another — sibling functions inside the same factory or module are the everyday example — CodeGraph stopped treating them as symbols you had asked for, and answered with whatever sat at the top of their file instead; on one 1,400-line file that meant a same-stem `QueuedMessage` interface on line 70 came back while the `queueMessage` function on line 1087 did not. And when an answer had to be trimmed to fit, it was trimmed from the bottom of the file down, so a symbol near the end of a long file was always the first thing cut. Trimming now protects the definitions you named wherever they sit in the file.

- A file built around one very long function no longer takes the whole `codegraph_explore` answer for itself — or disappears from it. Previously such a file was shown in full however big it was, which used up the room every file after it needed, and when the function was larger than the entire response the file was dropped without a word. These files now come back as a bounded window on whole lines — the signature and the top of the body, plus the call site when the call path runs through it — with the rest one follow-up `codegraph_explore` away.

- `codegraph_explore` no longer lets the first file in an answer spend the room set aside for the files below it, so the rest of the answer still arrives. Previously a large file near the top could quietly use up everything left, and the files ranked under it — each already judged relevant enough to include — were dropped with no source at all; on one question only one of six made it into the answer. Every file now keeps what it was given, and a question that really is about one file still concentrates on that file.

- When a `codegraph_explore` answer runs right up against its size limit, it now drops the trailing notes rather than a whole file's source. Previously the last file was cut even though trimming the notes alone would have fit, so a file that had already been read, ranked and rendered was thrown away at the last moment. Across a range of real projects this returns one more file and up to 20% more source per call.

- Every file `codegraph_explore` decides to include now actually arrives. A file shown in full could still spend room set aside for files below it — the fix above covered files shown as excerpts but not files shown whole — and the answer's own size bookkeeping under-counted each file's heading, so the answer ran past its limit and a fully prepared file was discarded at the end. A file that no longer fits whole is now shown as excerpts instead of vanishing, and one that overshoots by a little is trimmed to fit rather than dropped.

- The list of files an answer could not cover — the "explore these names for their source" pointers — is no longer thrown away when the answer is full. It is now budgeted for and trimmed to fit, so a full answer still tells you what it left out and which names to ask for next, instead of ending with no pointers at all.

- When a `codegraph_explore` answer shows a file as excerpts, a large excerpt that no longer fit was dropped entirely instead of being shortened. If the file's first excerpt happened to be a trivial one — an import block, a one-line helper next to the code you asked about — the excerpt carrying the actual answer was the one thrown away, and the file came back with a quarter of the room it had been given. On real projects that meant the top-ranked file delivered a fraction of its share while a far less relevant file took the rest. Excerpts are now shortened to fit, whole method by whole method, and only dropped when what is left is too small to hold anything readable.

- The blast-radius section of `codegraph_explore` flagged "no covering tests found" whenever no test called a symbol directly — falsely branding helpers that tests exercise through their callers as untested (about 40% of flagged symbols in a measured sample). The check now follows caller chains up to 3 hops and reports indirect coverage as "tested via callers"; when nothing is found it states exactly what was checked instead of an unconditional warning. Thanks @inth3shadows for measuring the false-positive rate. (#1475)

- When a file changed on disk after its last index sync, `codegraph_node` and `codegraph_explore` could return a different symbol's code under the requested name — current file bytes cut at outdated line positions — while presenting it as verbatim, trustworthy source. This hit hardest on projects queried through `projectPath` (for example, sub-projects of a monorepo), which have no live file watcher to flag pending edits. Both tools now verify each file against the index before showing sliced code: an out-of-date file is either shown whole with its full current source, or its code is withheld with a clear "changed on disk" notice — never served as a wrong slice. A fresh re-index restores normal output automatically. Thanks @inth3shadows for the thorough report and verification passes. (#1474)

#### Finding your project, live updates, and the CLI

- The MCP server now finds your project when it's launched from a workspace folder above it: if the launch directory has no index of its own but exactly one indexed project sits below it (a repo container, an agent workspace, a monorepo root), that project becomes the session's default — live file watching and the shared daemon included — instead of every tool call failing until a `projectPath` or `--path` is supplied. Thanks @nakisen. (#1606)

- When no project can be resolved at all, the MCP server now says so instead of starting silently: a startup log line names the directory it searched, and tool calls list the indexed sub-projects it can see nearby so you can pass one as `projectPath`. Previously the server looked healthy from the outside while every tool quietly had no project to answer from. Thanks @nakisen. (#1607)

- Editing `codegraph.json`'s `exclude` or `include` (or a `.gitignore`) while the MCP server is running now takes effect immediately. Previously the running file watcher kept the scope it had when it started, so a newly excluded file was removed by `codegraph sync` and then quietly re-added by the watcher seconds later — which looked like `exclude` not working at all — until the server was restarted. A scope change now refreshes the watcher and triggers a full reconcile, and a changed file the watcher hands to sync is re-checked against the current scope first, so the CLI and the live server can no longer disagree about what belongs in the index. Thanks @K1nG11. (#1590)

- `codegraph status` now sees new files inside brand-new directories. Git reports an entirely-untracked directory as a single collapsed entry, so source files created there — a freshly scaffolded `frontend/`, for example — were missing from the pending-changes report, which could claim everything was up to date while those files had not yet been indexed. Thanks @maxmilian. (#1213)

- The `codegraph context <task>` command documented in the CLI help now actually exists — it builds a ready-to-inject context pack for a task (relevant symbols, their relationships, and code) in markdown or JSON, restoring the contract external integrations like Memorix rely on (`--path`, `--format json`, `--max-nodes`, `--no-code`). (#1611)

- On Windows, the Claude Code prompt hook written by `codegraph install` failed with "command not found" when hooks run through Git Bash, which needs the `.cmd` extension to find the launcher. The installer now writes the platform-correct command, and re-running `codegraph install` (or `codegraph upgrade`) repairs an existing install in place. (#1466)

- Looking a symbol up by name no longer reads the whole graph. Every search made one full pass over all indexed symbols for each word you typed, and a question that named several symbols made two more passes per name — including for a word that matches nothing, which is the common case. The cost therefore grew with the size of the project, and it was paid again on every message when the prompt hook is enabled. These lookups now go through the name index instead. Results are identical; only the time to get them changes, and it no longer grows with the project. Thanks @maxmilian.

#### Indexing reliability and disk usage

- Indexing no longer crashes the whole process — a segmentation fault with no message and no partial index — on a C/C++ (or any other) file with extremely deep nesting, such as the parser stress-test fixtures in the clang and gcc test suites or a fuzzer corpus. Such a file is now handed to the fallback parser and recorded with a parse warning while the rest of the repository indexes normally. Thanks @apollo600 for the exact diagnosis. (#1581)

- Indexing no longer hangs on a Swift Vapor project containing a call with a long argument list. A single `.get(...)`-style call with many labeled arguments and no `use:` handler — the shape generated request builders produce — could stall `codegraph index`, `codegraph sync`, and the MCP server indefinitely. Route detection now handles such files in milliseconds, and every previously-recognized route shape still parses exactly as before. Thanks @maxmilian. (#1544) (Swift)

- Syncing a large batch of changed files no longer crashes with "Maximum call stack size exceeded" partway through. The crash aborted reference resolution after the files' symbols were already stored, leaving the graph with far fewer connections than a fresh index would have — and it hit exactly the scenarios that re-parse many files at once, including the automatic repair above. Thanks @netbrah for pinpointing the failure. (#1558)

- A long-lived index no longer drifts away from what a fresh `codegraph index` would produce. When a file gained or lost a symbol, references to that name in files the sync never touched kept pointing at the definition that was correct before the change, and — because nothing distinguished two same-named definitions — the winner could come down to the order files happened to be written, which differs between a full index and a sync. On this project's own repository, replaying 80 commits through `sync` left 5.7% of connections wrong; it is now 1.3%, and the wrong-answers-still-being-asserted half drops by 99.7%. Since call edges are what flow questions follow and what `codegraph_explore` ranks files by, this quietly degraded answers as an index aged, with nothing to indicate it. Syncing is unchanged in speed, and an edit that only changes a function's body does no extra work at all. Set `CODEGRAPH_NO_REBIND=1` to opt out.

- Fixed a v1.5.0 regression where a perfectly valid file could be permanently recorded as having 0 symbols, with no error reported. When a file's first parse attempt was interrupted — a parsing worker crash or timeout, most likely on slow or heavily loaded machines — the automatic retry stored an empty result for any language on the native extraction path, so the file's functions and classes silently vanished from search, callers, and impact until the file was next edited. Retries now store the file's real symbols, and a file already recorded as symbol-free is detected and repaired automatically by the next sync or re-index after upgrading. Thanks @Baiae for the report. (#1541)

- A CodeGraph process that gets force-killed — by the stuck-process watchdog, a crash, or the OS — no longer leaves the database's write-ahead log behind to grow without bound. Previously each killed session stacked more data onto the same log file and nothing ever shrank it, which on machines where sessions were killed regularly could quietly eat tens of gigabytes of disk. The log's resting size is now bounded (64 MB by default; `CODEGRAPH_WAL_HEAL_MB` changes it), and any oversized leftover is folded back into the index and trimmed automatically the next time the project is opened. This bounds only the log — the index itself has no size limit, and a large repository's log is still allowed to grow in proportion to its index while it is being built. Thanks @tiendungdev for the exceptional Windows report that pinned this down. (#1431)

- The background server's watchdog no longer kills a healthy server that is just waiting on a slow disk: like indexing already does, it now checks whether the database files are still making progress before concluding the process is stuck. Fewer spurious kills also means fewer leftover write-ahead logs. (#1431)

- `codegraph status` now shows the write-ahead log's size next to the database size and warns when killed sessions have left it oversized, and every line in the background server's log now carries a timestamp so kills and restarts can be placed in time. (#1431)

- A background daemon left behind by an out-of-memory kill or force-kill can no longer block every future session when the operating system reuses its process ID: daemon management now verifies a recorded process is really a CodeGraph daemon before trusting or signaling it, and `codegraph unlock` clears stale daemon artifacts as well as the indexing lock. Thanks @hcg1023 for the report and @danusha2345 for the fix. (#1553)

- When indexing has to fall back to parsing a file with its comment lines stripped — a last-resort recovery after repeated parser crashes — the file is now flagged with a visible warning instead of being reported as cleanly indexed. The recovered result can be incomplete, and reporting success made a fresh index quietly disagree with a later re-parse of the same unchanged file. Thanks @jeremypetz for the precise init-versus-sync symbol accounting that exposed this. (#1565)

- Data-only C/C++ headers near the file-size limit no longer hold a parser worker for several minutes before timing out; the default large-file parse budget is now bounded, while an explicitly configured larger timeout is still honored. (#1555)

- Reopening an index after a crash during bulk loading now restores every dropped database index, and a successful recovery sync marks the index complete instead of leaving it permanently flagged as interrupted. (#1556)

- Files skipped because they are too large or repeatedly fail to parse are now recorded with the reason, so unchanged rejected files are no longer rediscovered and retried on every status check and sync — and a later successful parse of such a file replaces the record with its real symbols. Thanks @netbrah for the exceptional failure analysis behind this batch, and @danusha2345 for the fixes. (#1557)

- C/C++ function-pointer analysis now bounds its compiled-pattern caches, so very large repositories can no longer exhaust the JavaScript engine's regular-expression code space during indexing. (#1559)

- JSX rendering analysis now runs only on JavaScript-family files, so JSX-looking strings in C/C++ (or any other language) no longer create impossible call edges — in pure-C projects and in mixed-language monorepos alike. (#1560)

#### Language and framework accuracy

- Calls to the methods of an exported object-literal constant — `export const api = { call() { … } }` used as a module's namespace, a common way to organize a TypeScript API surface — now resolve to the method, both in the defining file and through imports. Previously such a call linked to nothing (or to the constant itself), so `codegraph callers` and impact analysis reported zero callers for methods that are called from everywhere. Re-index after upgrading to pick up the edges. Thanks @IAliceBobI for the precise report and root-cause. (#1573)

- Import aliases defined in a shared TypeScript config are now picked up. Nx-style monorepos keep every `@scope/...` alias in a `tsconfig.base.json` that the root `tsconfig.json` only inherits through `extends`, so CodeGraph found no aliases at all and every cross-package import fell back to matching on name alone — which quietly attaches results to unrelated symbols that happen to share a name, exactly where a monorepo needs `codegraph_impact` and `codegraph_callers` to be right. Chains several configs deep, a config inherited from a package in `node_modules`, and a `baseUrl` declared in an inherited config are all followed now, and a `tsconfig.base.json` is read directly when the root `tsconfig.json` is only a project-references shell or isn't there at all. Re-index after upgrading. Thanks @maxmilian. (#1534)

- Methods implemented in a generic or lifetime-parameterized `impl` block (`impl<T> Source for BufSource<T>`, `impl<'a> Iterator for Parents<'a>`) are now recorded under the implementing type instead of the trait. Previously such a method could not be found by its type — "who calls `BufSource::read`" had no answer — and it collided with the trait's own declaration, which could even invent a call-graph edge out of an impl body that contains no call at all. Impls on a reference (`impl Trait for &Foo`) and on a module-qualified type (`impl Trait for m::Foo`) are attributed to their type too. Re-index after upgrading. Thanks @Dshuishui. (#1588) (Rust)

- A method call on a struct field — `self.inner.run()` with `inner: Inner` — now resolves to the method on the field's declared type. Previously the call was reduced to the bare method name and matched whichever same-named method was nearest, which was often the calling method itself, recording recursion that isn't in the source (a few hundred such self-edges in ripgrep alone), or a method of an unrelated type. References and `Box`/`Rc`/`Arc` fields are looked through, as Rust's own method calls are; a field whose type is external (a std or third-party type), a generic parameter, or a container like `Option`/`Vec` is left unresolved rather than guessed. Re-index after upgrading. Thanks @Dshuishui. (#1585) (Rust)

- Erlang functions that share a name but differ in arity are now separate symbols with the language's own `module:fun/arity` identity, so the everyday `f/1` delegating to `f/2` shows as a real call edge instead of a self-loop, each arity keeps its own `-spec` and source span, `-export([f/1])` marks exactly that arity as public, and asking `codegraph_explore` for a symbol the way Erlang spells it — `cowboy_req:header/3` — returns that definition. Re-index Erlang projects after upgrading. Thanks @Dshuishui. (#1610) (Erlang)

- Erlang behaviour dispatch no longer miscounts a call site's arity when an argument is a binary literal like `<<1,2,3>>` — the commas inside were counted as argument separators, which silently dropped (or could mislink) the dispatch edge to the behaviour callback. (#1358) (Erlang)

- A C++ `.h` header whose only C++ construct is a plain derived type — `struct Derived : Base` with no export macro, `class` keyword, or access section — is now recognized as C++ (previously only the export-macro form was). Such a header was read as C, so the derived struct vanished from the index and a phantom function named after the base type appeared in its place. The check now also covers the whole file rather than its first few kilobytes, so a long C-compatible preamble no longer hides the signal. Re-index after upgrading to pick up affected headers. Thanks @Jaysenpeng. (#1592)

- C, C++, Objective-C and Rust unions are now indexed as first-class `union` nodes. A `union` declaration previously produced no symbol at all, so it never appeared in search or `codegraph_explore`, and anything attached to it disappeared with it — in Rust, every `impl SomeTrait for MyUnion` lost its edge, the methods from that impl were left pointing at a type the graph did not contain, and asking which types implement a trait quietly skipped the union ones. A union-shaped dispatch table in C now resolves its function pointers like a struct-shaped one. A `typedef union { … } Name;` in C keeps the typedef's name and is no longer mistaken for a plain type alias. Thanks @ctype-lab. Re-index after upgrading to pick up unions in existing projects. (#1515)

- Python classes used as values — `return SomeSerializer` from a factory method, `handler = SomeClass` aliases, registry dicts and lists, and classes passed as arguments — now produce reference edges in the graph. Previously these idioms were invisible, so on Django and Django REST Framework projects, asking for a serializer's callers or the impact of editing it missed the views that actually use it. Re-index after upgrading to pick up the new edges. (#1478)

- SAP HANA `.xsjs` / `.xsjslib` imports now resolve across files: an extensionless `import { x } from './helpers'` in a `.xsjs` file finds `helpers.xsjslib`, so the cross-file call edge is created and `codegraph_callers` / `codegraph_impact` see it. Previously the import path resolved to nothing and the call fell back to same-name matching, which could bind the edge to an unrelated file that happened to export the same symbol. Complements the `.xsjs` / `.xsjslib` extraction support. Thanks @maxmilian. (#556)

## [1.5.0] - 2026-07-21

# ⚡ The Rust engine release — with near-instant sync

**This release rebuilds CodeGraph's parsing engine as a native Rust kernel, overhauls the resolution pipeline around it, and makes the live graph effectively instant: a save now reaches the graph in well under a second, even on a 27,000-file repository. It is the largest performance upgrade in the project's history — and every graph is verified byte-for-byte identical to the previous engine.**

- **Native Rust parsing for 20 languages** — TypeScript, JavaScript (+TSX/JSX), Java, Python, Go, C, C++, Rust, C#, Ruby, PHP, Swift, Kotlin, Scala, Dart, R, Lua, and Luau now parse in a compiled Rust kernel (Metal and CUDA ride the C++ path). Platforms without a prebuilt binary, and individual files with syntax errors, fall back to the previous engine automatically — same graph either way, proven on repositories from small libraries to the Linux kernel.
- **Adaptive to your machine** — CodeGraph sizes its parse workers, resolver pool, and caches from what the system actually has: real core counts (container/cgroup-aware, not the host's), honest available memory on macOS and Linux, and measured per-project resolution cost. A big workstation gets the full parallel pipeline; a 2-core VPS gets a pipeline tuned to finish reliably instead of running out of memory — the Linux kernel (70k files) indexes to completion on a 2-core, 6GB machine in under 12 minutes — down from 26 at the start of this cycle.
- **Resolution is dramatically faster across the board** — adaptive parallel resolution, smarter method-candidate lookup, and memoized supertype/conformance walking. The Swift compiler repository (27k files, Swift + C++) went from over 3 minutes to about 100 seconds within this release cycle; Rust, Lua, and Java-family projects all see double-digit improvements.
- **Sync is now near-instant** — a save reaches the graph in well under a second, even at compiler scale. The always-on watcher fires after a 300ms quiet window for lone saves (bursts of edits still coalesce), and hands the exact changed paths to sync instead of re-scanning the whole tree — measured save-to-fresh-graph work of ~0.3s on a 4,400-file Java project and ~0.4s on the 27,000-file Swift compiler repository, byte-identical to a full reconciliation.

Full details in the entries below.

### New Features

- Indexing TypeScript, TSX, JavaScript, JSX, Java, Python, Go, C, C++, Rust, C#, Ruby, PHP, Swift, Kotlin, Scala, Dart, R, Lua, and Luau projects is faster: parsing and symbol extraction now run in a native engine when a prebuilt binary is available for your platform (release bundles include one), producing exactly the same graph — verified byte-for-byte against the previous engine on real repositories, from small libraries up to vscode-, dubbo-, django-, git-, protobuf-, tokio-, rust-analyzer-, jellyfin-, rails-, symfony-, swift-nio-, kotlinx.coroutines-, ggplot2-, Kong-, Scala-3-compiler-, and Flutter-scale codebases (Lombok-generated members, C function-pointer tables, and Unreal-Engine-style macro-heavy headers included; CUDA and Metal sources ride the C++ path). The speedup is largest on resource-constrained machines like CI runners. No setup needed: platforms without the native binary, and individual files with syntax errors, automatically use the previous engine, and `CODEGRAPH_KERNEL=0` turns the native path off entirely.
- Reference resolution now runs in parallel on large projects. When a project has enough pending references to make it worthwhile (roughly 150k+, typical for big Java/Kotlin/Spring codebases), resolution fans out across worker threads while results are applied in the exact order the single-threaded path would have used — the graph comes out byte-for-byte identical, about twice as fast end-to-end on a 4,000-file Java project in our testing. Small projects keep the single-threaded path automatically (the fan-out costs more than it saves there). Set `CODEGRAPH_NO_PARALLEL_RESOLVE=1` to disable, or `CODEGRAPH_PARALLEL_RESOLVE_MIN=<count>` to tune when it engages.
- Indexing large projects got another sizeable speedup — about a quarter less wall-clock on the same 4,000-file Java project, with the graph still byte-for-byte identical. Two changes: the database no longer interleaves expensive checkpoint housekeeping into the middle of resolution on a fresh index (it's folded once at the end instead), and while one batch's results are being written out, the worker threads are already resolving the next batch instead of sitting idle.
- The dynamic-dispatch analysis that runs at the end of indexing (callback, event, and framework wiring) now runs its passes in parallel on large projects, cutting that stage roughly in half there — and a pass that crashes now retries safely instead of failing the whole index, which also makes very large codebases that previously died in this stage more likely to index to completion. Graphs remain byte-for-byte identical.
- On large projects, indexing writes its relationship data noticeably faster: secondary database indexes are set aside during the bulk of reference resolution and rebuilt once at the end, instead of being maintained row by row. Graphs remain byte-for-byte identical, and small projects are unaffected.
- Very-large-codebase reliability: on multi-million-symbol projects, an analysis pass that fails on a worker thread is now skipped with a clear message instead of being retried in a way that could take down the whole index, and the end-of-indexing index rebuild no longer risks tripping the liveness watchdog on huge graphs. Validated end-to-end on the Linux kernel (70k files, 2M symbols, 6.4M relationships) — it now indexes to completion even on a 2-core machine.
- Indexing is significantly faster — a fresh `codegraph init` on a medium TypeScript project takes about a third less wall-clock time, with the same graph produced byte-for-byte. The gains come from batching database writes, storing files on a dedicated writer thread, memoizing repeated import-resolution lookups, skipping per-row search-index maintenance during the bulk build (rebuilt once at the end), and — on completely fresh databases only — deferring disk durability until the index completes, since an interrupted first index is simply re-run. Set `CODEGRAPH_NO_FAST_INIT=1` to keep full crash-durability during the initial build, or `CODEGRAPH_NO_STORE_WORKER=1` to store on the main thread.
- `codegraph install` and `codegraph upgrade` now offer CodeGraph Pro beta access after finishing — answer yes, type your email, and you join the same waitlist as the getcodegraph.com homepage form. Strictly opt-in and asked at most once per machine total: nothing is sent unless you say yes and enter an email, either answer is remembered so no later install or upgrade ever re-asks, and non-interactive runs (`--yes`, scripts, CI) never see the question.
- Every release is now cryptographically verifiable: npm packages publish with npm provenance (the "Provenance" badge on npmjs.com, proving each version was built by this repository's release workflow from a specific commit), and the GitHub Release bundles carry signed build attestations you can check with `gh attestation verify <file> -R colbymchenry/codegraph`.
- Indexing inside CPU- or memory-limited containers (Docker, CI runners) now sizes its worker pools from the container's actual allowance instead of the host machine's, and giant codebases no longer balloon temporary database files during indexing (previously tens of GB of transient disk on Linux-kernel-scale projects). Together these prevent out-of-memory and out-of-disk failures on constrained machines; set `CODEGRAPH_RESOLVE_WORKERS` to override the resolution worker count explicitly.
- Indexing very large projects on multi-core machines got faster again: the parallel-resolution workers now periodically refresh their read-only database connections, which lets database housekeeping advance instead of silently building up a backlog behind long-lived readers — a backlog that was taxing the indexer's own writes. Graphs remain byte-for-byte identical; the win is largest at Linux-kernel scale on many-core machines.
- Indexing on macOS now uses the machine's real memory headroom when sizing its parallel-resolution workers. macOS deliberately keeps RAM filled with reclaimable cache, so the previous free-memory reading came back tiny (~1GB on an otherwise idle machine) and silently halved the worker pool — a medium Java project's fresh index ran about 15–20% slower than the hardware allowed. Graphs remain byte-for-byte identical; the same fix also lets a memory-driven analysis cache engage fully on macOS for large C codebases.
- Fresh indexing got a sizeable across-the-board speedup: during the initial build, the database's secondary lookup indexes are set aside and rebuilt once after parsing instead of being maintained row by row — the same proven trick the later linking phase already used, now applied to the whole parse lane — and the reference-resolution loop likewise stops maintaining lookup indexes it never reads, rebuilding them at the end when almost nothing is left in the table. A medium Java project's parse phase runs about 58% faster and its full fresh index about 19% faster end-to-end; a Linux-kernel-scale index that took ~15 minutes on an 8-core machine now completes in about 11, with the resolution phase alone dropping by a third. Graphs remain byte-for-byte identical, and incremental syncs are unaffected.
- Saving a file now updates the graph almost immediately: the file watcher fires after a 300ms quiet window for one or two changed files (bursts still coalesce under the full debounce, and `CODEGRAPH_WATCH_DEBOUNCE_MS` remains the upper bound), and watcher-triggered syncs reconcile exactly the changed paths instead of stat-walking the entire repository. Directory deletions and event storms still run the full scan-diff, so nothing the events can't describe is ever missed. Measured: save-to-fresh-graph sync work drops to ~0.3s on a 4,400-file project and ~0.4s on a 27,000-file one, with resulting graphs byte-for-byte identical to a full reconciliation.
- Indexing Swift and other protocol/interface-heavy codebases got dramatically faster: the conformance walk that checks whether a method lives on a receiver's supertypes (protocols, base classes, extensions) now remembers its answers for the duration of each resolution batch instead of re-querying the graph for every call site — on the Swift compiler repository (27k files) that walk ran nearly a million times per index. A fresh index of that repo drops from about 185 seconds to under 100, with the graph byte-for-byte identical. Method-candidate lookup also gained a per-name owner index, so overload-heavy names (`init` in Swift, `execute` in Java) no longer pay a full candidate scan per receiver type.
- Resolving method calls through local variables (`recv.method()`, Lua's `recv:method()`, R's `recv$method()`) got much cheaper on repos where the same receiver is called over and over: the declaration scan that types the receiver now remembers what it has already scanned per scope instead of re-reading the same source lines for every call site, and the regex patterns it scans with are compiled once per receiver instead of per call. Kong's fresh index drops another 8% on top of the require-resolution fix (23% cumulative), with graphs byte-for-byte identical everywhere — including Java projects, where this same scan successfully types tens of thousands of receivers.
- Indexing Lua and Luau projects got a sizeable speedup: resolving each `require(...)` no longer rescans the project's entire file list four times — a per-project filename index answers the same lookup instantly, cutting per-require resolution from about a millisecond to microseconds. A fresh index of Kong (1,870 Lua files) runs about 16% faster end-to-end, with the graph byte-for-byte identical. The same housekeeping also closes a latent staleness edge where COBOL copybook lookups could keep serving a cached file list after files changed.
- Parallel reference resolution now engages adaptively instead of by a fixed project-size cutoff: the indexer measures the actual per-reference resolution rate on the first batch and spins up the worker pool mid-run whenever the remaining work justifies it. Languages whose references are expensive to resolve benefit most — Rust especially: a fresh index of tokio runs about 23% faster, with the graph byte-for-byte identical. Small projects and low-core machines (2-core CI runners) keep the single-threaded path exactly as before.
- The dynamic-dispatch analysis at the end of indexing now skips passes that provably can't produce anything for the project at hand: React re-render bridging when no class has a `render` method, React Native and Expo cross-platform pairing when the required languages aren't present, and MyBatis mapper linking when there's no mapper XML. Previously each of these scanned the whole graph before coming up empty — on a 4,000-file Java project that was about 0.9 seconds of wasted analysis per fresh index. The interface-implementation bridging pass also got cheaper on real work: it no longer re-fetches a hub interface's method list once per implementer, and classes that extend or implement nothing are skipped before any per-class lookups. Graphs remain byte-for-byte identical.
- Indexing large C and C++ codebases spends much less time in the function-pointer dispatch analysis (the pass that connects handler tables like a command table or an ops struct to their call sites): each source file is now read and prepared once instead of four times, files that can't contribute any dispatch wiring are skipped outright in the later linking steps, and on platforms with the native engine the per-file scanning itself now runs natively too. On a Linux-kernel-scale tree the pass runs about a third faster end-to-end, with graphs byte-for-byte identical; platforms without a native binary keep the same results on the previous path.

### Fixes

- Codex installs, upgrades, and uninstalls now preserve TOML array-of-table sections that appear after CodeGraph's MCP configuration instead of accidentally removing them. (#1351)
- TypeScript, TSX, and JavaScript files now parse with up-to-date grammars — modern syntax such as `using` declarations and import attributes no longer trips parse errors that could drop surrounding symbols. (The previously bundled grammars dated from 2023.)
- Rust files also parse with an up-to-date grammar now (the previously bundled build dated from 2023), which additionally sharpens method-call attribution: calls through struct fields resolve with receiver context instead of falling back to ambiguous bare-name matching, removing a class of wrong call edges on common names like `len` and `start`.
- Ruby files also parse with an up-to-date grammar now (the previously bundled build dated from early 2024), which fixes a misparse of safe-navigation operator-method calls (`recv&.!= x`) that had recorded the wrong callee name.
- PHP files also parse with an up-to-date grammar now (the previously bundled build dated from 2023): files using modern PHP features — enum constants, PHP 8.4 property hooks, parenthesis-free `new X()->method()` chaining — no longer hit parse errors or misparses that dropped or garbled their symbols, so codebases like Symfony and Laravel index substantially more accurately.
- Swift files also parse with an up-to-date grammar now (the previously bundled build dated from 2023): swift-testing `#expect`, `#Preview`-style macros, the `package` access modifier, and typed `throws` no longer produce parse errors that dropped surrounding symbols — server-side Swift projects like Vapor see the biggest recovery.
- Dart's grammar is now bundled with CodeGraph itself instead of being resolved from a third-party package whose Dart build floated on an unpinned upstream reference — a routine dependency update can no longer silently change how Dart files parse.
- Searching or exploring by field names now finds the code that defines them. A query made of object keys or API field names (`profileInfo isTrialEligible quotaInfo billingMethod`) used to return unrelated results while the defining files never appeared, because three retrieval steps each dropped multi-word camelCase terms: an internal case-comparison bug, a match step that only considered classes (never functions or methods), and exploration seeding that required exact symbol-name matches. All three are fixed — `codegraph_explore` with a bag of field names now surfaces the controllers and services that assemble those fields. (#1196)
- `codegraph.json`'s `includeIgnored` works again for the "folder of repos" layout: when one `.gitignore` rule covers a parent directory (`/repos/`) holding several embedded git repositories, opting in the individual repos (`"includeIgnored": ["repos/a/"]` — the exact spelling `codegraph init`'s own hint suggests) previously matched nothing and indexed zero files, looping the same suggestion back at you. Both spellings now work — name the parent directory to opt in everything under it, or name individual repos to opt in just those — and the hint no longer re-suggests repos that are already configured. (#1295)
- Method calls on literals (`", ".join(...)` in Python, `"x".split(...)` in JavaScript, and the like) no longer produce call edges to unrelated project functions that happen to share the builtin's name — a codebase with a function called `join`, `get`, or `update` could show phantom callers from every string-builtin use. Additionally, a function nested inside another function is now only matched as a call target from inside its container, since it isn't reachable from anywhere else. Blast-radius and affected-test results get cleaner on Python and JavaScript codebases especially. (#1230)
- Go method calls through a struct field (`target.conn.Exec(...)`) no longer bind to unrelated same-named local methods when the field's type is external — `conn *sql.DB` calls were being attributed to a local interface that happened to declare `Exec`, fabricating internal dependencies. Chained field calls now resolve by inferring the field's declared type from the struct definition: in-project types (including unexported ones like chi's `tree *node`) gain correct, validated call edges that never existed before, and external types (standard library, third-party modules) are left unlinked instead of guessed. (#1276)
- TypeScript/JavaScript method calls through an imported singleton (`import { store } from './store'; store.notify()`) now resolve to the class method instead of the exported constant, so `codegraph callers` sees cross-file callers of the method — previously only same-file calls were attributed and a method used everywhere could look unused. The same declaration-based type inference applies across the languages that share it (Python, Java, Kotlin, Go, and more), and a failed inference keeps the old edge rather than guessing. (#1292)
- `codegraph node <symbol> -f <file>` now prints the symbol's source body. Pinning an ambiguous name to a specific file (the whole point of `-f` when many files define the same function) returned only the location and caller trail with no code. (#1284)
- Deleting a whole directory is now picked up by watch mode: the files inside it are removed from the index on the next auto-sync instead of lingering as stale records until an unrelated edit happened to trigger one. Operating systems often report a directory deletion as a single event on the directory itself (with no per-file events for its contents), which the watcher previously discarded. (#1285)
- `codegraph sync` now gets the same slow-disk fix that made full indexing fast in 1.4.0: database checkpointing is deferred for the whole incremental run instead of firing every few megabytes of writes. On mechanical drives and other high-latency storage, a small sync on a large index no longer stalls for minutes at near-zero CPU — the cost of a sync scales with what changed, not with the size of the existing index. The same `CODEGRAPH_NO_WAL_DEFER=1` switch turns it off. (#1248)
- C functions declared with a project-specific attribute macro in front of a typedef'd return type (`SEC_ATTR UINT32 MyFunc(VOID)` — common in embedded and kernel code) are now indexed under their real names. Previously the parser tripped over the unknown macro and stored the parameter list as the function name, leaving entries like `"(VOID)"` in the graph and making the real function unfindable. (#1211)
- Macro-heavy C and C++ code indexes much more completely. Fifteen ubiquitous idioms that previously tripped the parser into error recovery — dropping symbols, garbling names, minting phantom entries, or losing the relationships between real ones — now parse cleanly: the `#ifdef __cplusplus` / `extern "C" {` compatibility guard in C headers, iterator macros in statement position (`list_for_each_entry(pos, head, member) { … }` and the whole Linux-kernel/git/jemalloc family — including calls that wrap across lines, like `hlist_for_each_entry_rcu(…)` with a `lockdep_is_held` argument), the Linux/sparse declaration annotations (`static int __init foo(void)`, `void __user *buf`, `container_of(p, struct T, m)`), parameterized annotations (`__free(kfree)`, `__printf(4, 0)`, `__counted_by(count)`), per-CPU and work-queue declaration macros (`static DEFINE_PER_CPU(struct T, name);`, initialized forms included), bare type arguments to allocator and list macros (`kzalloc_obj(struct T)`, `list_entry(p, struct T, member)`), `va_arg` with a qualified or pointer type, GNU named-variadic macro definitions (`#define dbg(fmt, args...)`), `static notrace`-style compiler markers, C23 `auto` declarations, trailing parameter annotations (`int argc UNUSED`, git's house style), namespace-management macros alone on a line (`FMT_BEGIN_NAMESPACE`, Qt's `Q_OBJECT`), and function attribute macros in front of C++ return types. On the Linux kernel this cuts files lost to parse errors on the core directories by nearly half and recovers thousands of call and reference relationships that error recovery silently dropped, with git and redis seeing smaller cuts of the same kind; graphs stay byte-for-byte identical between the native and fallback engines. A related fix stops the macro handling itself from corrupting `#define` lines that mention the same macro names, which removed a class of phantom parse errors in fmt-style headers.
- C++ methods defined out-of-line inside a namespace (`namespace sim { Output MyClass::Apply(...) { ... } }`) now carry the namespace in their qualified name, matching their class. Fully-qualified call sites from other files (`sim::MyClass::Apply(...)`) resolve to the definition again, so `codegraph callers` and file impact no longer come up empty for this pattern. (#1291)
- C++ methods defined out-of-line on a template class (`template <typename T> T Box<T>::get() { ... }`) no longer keep the template parameter list in their qualified name. They now index as `Box::get` — identical to an inline definition of the same method — so they link to their class and resolve from call sites again, and pathological multi-line template parameter lists can no longer blow the qualified name past filesystem name limits. (#1286)
- Go route detection no longer misidentifies ordinary method calls that share HTTP verb names — `cache.Put("key", value)`, `store.Get("config", out)`, `bus.Handle("user.created", handler)` and the like were being indexed as HTTP routes, polluting route listings in cache-heavy codebases. A registration now has to look like one: its first argument must be a `/`-prefixed path (all routers) or a Go 1.22 `"METHOD /path"` pattern on `Handle`/`HandleFunc`, which now also extracts the method instead of listing the route as `ANY`. (#1259)
- Progress output on Windows no longer mixes ASCII `|` rails with the Unicode `│ ◆ ●` frame around them. In terminals that render Unicode (Windows Terminal, VS Code, ConEmu/Cmder, JetBrains, Alacritty), the whole `codegraph init` / `index` / `sync` block now draws with matching box-drawing characters; unrecognized legacy consoles keep the safe all-ASCII output that avoids garbled characters. `CODEGRAPH_ASCII=1` / `CODEGRAPH_UNICODE=1` still override in either direction. (#398)
- CLI output now honors the `NO_COLOR` convention and new `--color` / `--no-color` flags, and goes plain automatically when piped: commands like `codegraph status`, `query`, `callers`, and `files` no longer embed ANSI color codes when stdout isn't a terminal, and a piped `codegraph init` / `index` / `sync` prints simple per-phase lines instead of progress-animation control characters. `FORCE_COLOR` or `--color` forces color back on for pipes that render it. (#1281)
- Callers and impact analysis no longer silently under-count a function that calls the same callee many times. When one caller contained several call sites to the same callee and an internal resolution batch boundary happened to split them, cleanup after the first batch removed the later sites' pending rows before they were ever attempted — their edges were never created, deterministically, and which edges went missing shifted with unrelated changes to the project's total reference count. Post-pass cleanup now targets the exact database row each processed reference came from. Found while validating the operator-call fix on nlohmann/json, where `write_cbor`'s 11 calls to `to_char_type` indexed as 10. (#1269)
- C++ explicit operator calls — `a.operator+(b)`, `p->operator+(b)`, `a.operator[](3)`, and the other symbolic forms — now produce a `calls` edge to the operator method, so an operator invoked only through the explicit syntax no longer looks uncalled in callers and impact analysis. tree-sitter parses these call sites with the operator name stranded in an error node (never as a normal member access), so the call's target was silently read as just the receiver variable; the operator name is now recovered from the error node and resolved through receiver-type inference like any other member call — a same-named operator on an unrelated class can never capture the edge. Infix uses (`a + b`, `a[i]`) need real type inference and are tracked separately. (#1247)
- `codegraph init` and `codegraph index` no longer look hung after "Resolving refs" reaches 100%. The dynamic-dispatch linking that runs after resolution (callbacks, React re-renders, C function pointers, and the rest) had no progress display, so on repos where it takes a while — large C codebases especially — the bar just sat frozen at 100% until it finished. That work now shows as its own "Linking dynamic dispatch" progress phase, and the heaviest pass — C function-pointer linking — additionally reports progress within the pass, so a large C codebase advances the bar smoothly instead of parking it on one number for the bulk of the phase.
- Indexing no longer prints repeated "SQLite is an experimental feature" warnings that garbled the progress display. The warning comes from Node's built-in SQLite and fired once per parsing worker; it's now suppressed on every launch path.

## [1.4.1] - 2026-07-10

### New Features

- The MCP server now notices when a newer CodeGraph release exists and tells you — a long-running server used to drift behind releases silently until something broke. On startup it checks the latest release in the background (never blocking, at most once a day, cached across all servers on the machine) and surfaces a one-line "update available — run `codegraph upgrade`" notice in the server log, in the instructions your agent sees on connect, and in `codegraph_status`. Nothing updates by itself, and being offline just means no notice. Opt out with `CODEGRAPH_NO_UPDATE_CHECK=1`; `DO_NOT_TRACK=1` disables it too. (#1243)

### Fixes

- `codegraph upgrade` on a Windows npm install actually runs npm again — modern Node refuses to launch `npm.cmd` directly, so the upgrade failed with a spawn error before doing anything. npm is now invoked the way a terminal would run it. (#1238)
- `codegraph uninstall` now actually uninstalls CodeGraph. It used to remove only the agent configurations and leave every installed binary behind, so `codegraph` still ran afterward — especially confusing when both an npm global install and a standalone install were present and removing one still left the other answering on PATH. Uninstall now finds every install on the machine (the standalone bundle, the npm global package, the launcher link) and removes them all, after showing you exactly what it found and asking first (`--yes` skips the prompt). Machine-level settings like your telemetry choice are preserved, a source checkout is never touched, and the new `--keep-cli` flag restores the old configs-only behavior. (#1071)
- `codegraph_explore` no longer lets ordinary English words in a natural-language question hijack the ranking when they happen to match a code symbol's name. A question like "how does the upgrade flow check the latest version" used to treat "check" as a symbol the agent asked for by name, rank that unrelated definition's file first, and crowd the files the question is actually about out of the answer entirely. Precisely written symbol names (camelCase, PascalCase, snake_case, qualified names) still get top billing exactly as before, as do plain-word symbol bags whose words belong together in the same file.
- PHP method calls made through a class property — `$this->dep->method()`, the dominant call shape in constructor-injection codebases (Symfony, Laravel) — now resolve to the method on the property's declared type, so callers and impact analysis see production call sites instead of reporting a DI-heavy method as uncalled or test-only. Promoted constructor parameters, typed properties, classic constructor assignment (including multi-line signatures), and typed setter injection all count; interface-typed properties resolve to the interface method, and inherited methods resolve through the type hierarchy. Only property-shaped declarations are consulted — a same-named local variable or parameter elsewhere can never mistype the property — and a property whose type can't be recovered statically stays unlinked rather than guessed. Thanks @w0lan. (#1220)
- `codegraph upgrade` now also refreshes what previous versions installed into your agents — the CodeGraph section in CLAUDE.md / AGENTS.md / GEMINI.md and the MCP entry — so upgrading no longer leaves agents following instructions written for tools that have since been renamed or removed. Refresh-only: agents you never configured are not touched, and your permission and hook choices are preserved. Also available manually as `codegraph install --refresh`, and skippable with `CODEGRAPH_NO_INSTALL_REFRESH=1`. (#1238)
- `codegraph upgrade` on an npm install now upgrades through npm again instead of quietly creating a second copy that never wins the PATH race — previously `codegraph --version` kept reporting the old version forever, no matter how many times you upgraded. (#1238)
- After every upgrade, CodeGraph now checks that the `codegraph` command your terminal resolves actually serves the freshly installed version — confirming you don't need a new terminal, or telling you exactly which stale install is shadowing the new one. (#1071)
- The safety watchdog no longer kills a healthy index on severely degraded storage. It used to judge liveness purely by the event loop, so one long database write on a struggling disk looked identical to a hung process and could get a valid, in-progress index terminated. During `codegraph index`/`codegraph init` the watchdog now also checks whether the index files on disk are advancing before it acts: slow-but-progressing work is left alone (bounded by a hard cap), while a genuinely hung process is still killed exactly as fast as before. (#1231)
- Incremental sync now picks up cross-file relationships that only become resolvable after an edit — for example, when a file gains an export that another, unchanged file was already importing or calling. Previously the reference in the unchanged file was never revisited, so callers, impact, and flow results silently omitted the new edge (while status reported a clean index) until a full re-index. References that can't be resolved yet are now remembered and automatically retried whenever a change introduces a symbol that could satisfy them — this also covers a class gaining a new method that other files already call. Thanks @loadcosmos for the report with a minimal reproduction. (#1240)
- The reverse case is fixed too: when an edit removes or moves a symbol (or deletes its file), callers in unchanged files now re-resolve during the same sync — rebinding to the symbol's new home when it moved, or waiting to reconnect automatically when it comes back — instead of silently losing their relationship until a full re-index. (#1240)

## [1.4.0] - 2026-07-10

### New Features

- Indexing is dramatically faster on slow storage — mechanical HDDs, network folders, and virtualized disks. The database no longer folds its write journal back into the main file thousands of times during a bulk index (that folding was ~95% of all disk activity); it now streams writes sequentially and folds them back in a few large, coalesced passes that run off the main thread. In a disk-throttled benchmark matching the reported hardware, a mid-size Java project went from over 25 minutes to under a minute, and there is no change on fast disks. Opt out with `CODEGRAPH_NO_WAL_DEFER=1`; tune the fold-back threshold with `CODEGRAPH_WAL_VALVE_MB`. (#1231)
- New `CODEGRAPH_PARSE_TIMEOUT_MS` environment variable to raise the per-file parse budget on unusually slow storage, the same way `CODEGRAPH_PARSE_WORKERS` already tunes the worker count. (#1231)

### Fixes

- Indexing on slow storage (mechanical HDDs, network folders) no longer collapses into false "parse timeout" failures. When disk writes stalled the coordinating thread, parses that had already finished — including empty files — were being misjudged as hung, their workers killed, and the files silently dropped from the index. A parse result is now judged by the worker's own clock, so a stalled coordinator accepts the finished result instead of killing the worker; only a genuinely hung parse is terminated (after a wider grace window). Files that do hit the timeout are retried at the end of indexing instead of being silently lost. Thanks @KnifeOfLife for the exceptional report. (#1231)
- Parse workers now receive their grammar files from memory instead of each re-reading them from disk on spawn, eliminating a feedback loop on slow disks where every worker restart added more disk contention — and making worker restarts cheaper everywhere. (#1231)

## [1.3.1] - 2026-07-09

### Fixes

- Fixed `codegraph init` grinding for 20+ minutes (or getting killed by the safety watchdog) near the end of the "Resolving refs" step on large PHP and JavaScript codebases — a regression since 0.9.x reported on a 12,000-file project that used to index in about a minute. One of the dynamic-dispatch analysis passes only ever applies to Swift/Kotlin closure collections, but it was scanning every function in every language, and on `.push(`-heavy JavaScript (or any codebase with big generated functions, especially with non-ASCII text) its per-match bookkeeping went quadratic. The pass now skips languages it can't apply to, does its line accounting in constant time, and stays responsive even inside a single pathological function — the graph produced is identical. Thanks @sniperrenren for the report. (#1235) Two failure modes are fixed: on multi-million-symbol projects (e.g. the Linux kernel, ~95,000 files) the final analysis phase ran out of memory and crashed the process outright, and on large projects on slower machines (reported on a 24,000-file Java project on Windows) the same phase could stall long enough that the safety watchdog killed a healthy, still-progressing index at ~98% (#1212). The whole phase now streams its work instead of holding whole-graph snapshots in memory, keeps the process responsive throughout, and skips analysis passes for languages a project doesn't contain — which also makes the tail of indexing noticeably faster on single-language repos. The resulting graph is identical, and a genuinely wedged process is still detected and killed.
- Indexing and `codegraph sync` stay responsive through their heaviest internal steps on huge projects: the post-index database maintenance (which on a multi-gigabyte index could stall the process for minutes and get a fully successful index killed by the safety watchdog at the finish line) now runs on a background thread, storing a giant generated file no longer freezes the process mid-extraction, and the reference-resolution bookkeeping between progress updates is broken into small responsive steps. The resulting graph is byte-for-byte identical.
- Fixed a race that could leave a freshly-attached MCP session permanently silent: when a client's first messages arrived glued together during the daemon's connection handshake (roughly one attach in five on a busy machine), the daemon could drop them and stop reading that connection entirely — every tool call from that session then hung with no reply. The handshake now hands the connection over losslessly, and the fix is validated by hammering the previously-flaky attach test 25× under load.
- The first tool call after the shared daemon starts no longer waits behind the query workers' cold start (which can take many seconds on a busy machine) — it's served directly until the first worker is warm, so a fresh session answers immediately.

## [1.3.0] - 2026-07-07

### New Features

- CodeGraph now indexes **Nix** (`.nix`) — flakes, NixOS and home-manager modules, overlays, and package sets join the graph: `let` and attrset bindings, functions (simple, destructured `{ pkgs, ... }`, and curried), and `inherit` bindings all become searchable symbols, with call edges between bindings. File-level wiring follows the ways Nix actually connects files: `import ./relative/path.nix` (with `import ./dir` reaching the directory's `default.nix`), NixOS module `imports = [ ./hardware.nix ../common ]` lists, flake-style `modules = [ ./configuration.nix ]` lists, and the nixpkgs `callPackage ./pkgs/foo { }` idiom — so "what does this configuration actually pull in" and "what uses this module" are answerable on real setups. Dynamic references (`import <nixpkgs>`, variable paths, flake-input module references) are deliberately left unlinked rather than guessed. Thanks @TyceHerrman. (#324, #332, #648)
- The NixOS module system's option wiring is bridged, so flow questions cross the module boundary instead of going dark at it: a config write like `launchd.user.agents.myapp = { ... }` or `home.file.".gitconfig" = { ... }` links to the module that declares that option (`options.launchd.user.agents = mkOption { ... }` — flat and nested declaration spellings both count, quoted keys like `system.defaults.NSGlobalDomain."com.apple.dock"` match their exact declaration), which makes "how does enabling this service produce the launchd daemon / generated config file" traceable end-to-end and "what sets this option" answerable across modules, tests included. Precision is deliberately conservative: interpolated `${...}` paths, options declared in more than one module, and submodule-internal option namespaces stay unlinked rather than guessed, and every bridged hop is labeled as heuristic module-system wiring rather than shown as a plain reference.
- CodeGraph now indexes **ArkTS** (`.ets`) — the language of HarmonyOS / OpenHarmony apps. Everything TypeScript gets extracted (classes, interfaces, enums, type aliases, imports/exports, call edges), plus ArkTS's own constructs: `@Component` / `@ComponentV2` structs with their decorators (`@Entry`, `@State`, `@Prop`, `@Link`, `@Local`, `@Param`, …) captured and searchable, `build()` view trees linked parent→child so "which pages render this component" is answerable, chained attributes connected to the `@Extend`/`@Styles` functions they invoke, `@Builder` methods and functions wired into the call graph, and `.onClick(this.handler)`-style event bindings linked to their handler methods. Modular HarmonyOS projects resolve across module boundaries too: a bare `import { CartRepository } from "data"` follows the `oh-package.json5` `file:` dependency to the right module — honoring each module's declared `main` entry, from `.ets` and `.ts` consumers alike — while ambiguous names in multi-app monorepos deliberately stay unlinked rather than guessed, and mixed `.ets`/`.ts` codebases cross-link freely. Validated on real HarmonyOS apps including the official OpenHarmony samples monorepo. (#396, #512, #648, #890)
- ArkUI's dynamic hops are bridged so flow questions cross them instead of going dark, each labeled as dynamic dispatch rather than shown as a plain call: methods that assign a reactive property (`@State`, `@Local`, …) link to the component's `build()` (the re-render hop — assignment-gated, so a method that merely reads state gets no edge); `emitter.emit(eventId)` links to the matching `emitter.on/once` subscriber when both sides share a statically-recoverable event key (numeric ids pair within one file only, named constants within one module, so unrelated samples in a monorepo never cross-link); and `router.pushUrl({ url: 'pages/Detail' })` links to the target page's `@Entry` struct, with ambiguous urls left unlinked rather than guessed.
- Interrupted or incomplete indexing is now visible instead of silent: a run killed mid-index (crash, out-of-memory, watchdog) leaves a marker that `codegraph status` reports as a truncated index, a completed run that dropped files reports itself as partial — both in the human output and in `status --json` — and `codegraph index` prints a warning with the exact counts when its result doesn't add up to what the scan discovered.
- CodeGraph now indexes **Terraform and OpenTofu** (`.tf`, `.tfvars`, `.tofu`) — resources, data sources, modules, variables, outputs, providers, and every `locals` attribute become symbols (e.g. `aws_s3_bucket.my_bucket`, `var.region`, `module.vpc`, `local.prefix`), and uses like `var.region`, `module.vpc.id`, `data.aws_caller_identity.current`, or `aws_s3_bucket.my.arn` are wired up cross-file, so search, callers, and impact queries return real results on infrastructure repos instead of nothing. Module calls are bridged across the module boundary: a `module` block's inputs link to the child module's variables, `module.vpc.vpc_id` reaches the child's `output "vpc_id"` definition, and the block's local `source` path links to the module's files — so "what breaks if I change this module's variable" reaches every caller instead of dead-ending at the declaration (registry and git sources are deliberately left as visible boundaries rather than guessed). Cross-component wiring through the cloudposse/atmos `remote-state` module connects too: `module.vpc.outputs.vpc_cidr` in one component reaches the `vpc` component's own output when the component name is statically declared (a literal, or a variable with a literal default) and exactly one directory matches — anything dynamic or ambiguous stays unlinked. Aliased providers are first-class: `provider "aws" { alias = "east" }` gets its own symbol, and `provider = aws.east` on a resource (or a module's `providers` map) links to that configuration, found up the module tree the way Terraform actually inherits it. `moved`/`import`/`removed` state-migration blocks and `check` assertions reference the resources they name, so a refactor's paper trail is part of the graph. `.tfvars` assignments link to the variables they set, including var-files kept in a subdirectory. Resolution follows Terraform's real per-directory scoping, so same-named variables across modules never cross-link and "what depends on `var.project_id`" in a multi-module repo never mixes in unrelated modules. Thanks @Javviviii2. (#83, #310, #648)
- CodeGraph now indexes **CUDA** (`.cu`, `.cuh`) — kernels, device/host functions, structs, and classes become symbols, and the host→kernel call edge survives the `<<<grid, block>>>` launch syntax, so questions like "how does this call reach the GPU kernel?" trace across the CPU/GPU boundary instead of going dark at the launch site. Real-world launch styles all connect: templated launches (`my_kernel<Traits, 256><<<grid, block>>>(args)`), launches through a local function pointer (`auto kernel = &my_kernel<...>; ... kernel<<<grid, block>>>(args)` — each branch-assigned target linked), brace-initialized launch configs (`<<<dim3{1,1,1}, dim3{256,1,1}>>>`), and kernels defined through a name-in-first-argument macro (flash-attention's `DEFINE_FLASH_FORWARD_KERNEL(kernel_name, ...) { ... }` style), which now index under their real kernel names. CUDA that lives in plain `.h`/`.hpp` headers — where much real-world device code sits, launch-template headers included — is recognized by content and indexed the same way. Validated on llm.c, flash-attention, and NVIDIA CUTLASS. (#387, #648)
- C++ symbols defined inside `namespace` blocks now carry the namespace in their qualified name (`flash::compute_attn`, C++17 `namespace a::b {` included), and namespace-qualified calls (`ns::fn(...)`) resolve to their definitions — previously such calls never linked at all, which hid much of the call graph in namespace-heavy C++ codebases from callers and impact analysis.
- C++ calls that spell out template arguments (`fn<T, 256>(args)`) now link to the function they instantiate, the same normalization templated base classes already had.
- CodeGraph now indexes **Solidity** (`.sol`) — contracts, libraries, interfaces, structs, enums, modifiers, events, errors, and state variables become first-class symbols, with call edges that follow `emit`, `revert`, modifier guards (`onlyOwner`-style, including base-constructor chains like `constructor() ERC20(...)`), and library/method calls (including `using` directives). `import` directives resolve to the imported file, so cross-contract questions like "trace `transferFrom` through allowance and balance updates" or "how does `onlyRole` reach the role storage?" work out of the box on real Solidity codebases (validated on solmate, solady, and OpenZeppelin Contracts). Thanks @naiba. (#374, #648)
- Erlang behaviour dispatch is now followed through the graph: a framework call through a variable module — cowboy's `Handler:init`/`Middleware:execute` folds, a plugin manager's `Mod:callback(...)` — links to the repo's implementations of the behaviour that declares that callback, so flow traces and impact cross the OTP callback boundary instead of stopping at it. The links are precision-gated: the callback arity must match, exactly one behaviour may own that callback shape (a collision stays unlinked rather than guessed), the implementer must actually export the callback, and the fan-out is bounded — a behaviour with hundreds of implementers stays a visibly dynamic boundary. Every bridged hop is labeled as dynamic dispatch with its wiring site, never shown as a plain static call.
- CodeGraph now indexes **Erlang** (`.erl`, `.hrl`) — functions, with clauses and arities of the same name grouped as one symbol spanning all of them, plus records with their fields, `-type`/`-opaque` aliases, `-define` macros, and `-spec` signatures attached to every function. Cross-module `mod:fn(...)` calls resolve to the target module's function, `fun name/arity` values are captured as references (so callback registrations like `lists:foreach(fun submit/1, ...)` link up), `-include`/`-include_lib` connect to the header files they pull in, `-behaviour` declarations link a callback module to its behaviour (and only ever to a module — a same-named macro or function elsewhere in the repo is never mistaken for one), and `-export` lists (plus `-compile(export_all)`) drive each function's public/private flag. OTP's indirection idioms are followed where the target is static: `spawn`/`apply`/`proc_lib`/`timer`/`rpc` calls that name their target as `(Module, Function, Args)` arguments produce call edges, and `gen_server:call`/`cast` connects to the target module's `handle_call`/`handle_cast` — its own when targeting `?MODULE` (including the `-define(SERVER, ?MODULE)` idiom), and the named module when a registered name follows OTP's name-the-server-after-its-module convention (`gen_server:call(other_mod, ...)`, directly or through a `-define(STORE, other_mod)` macro); a registered name that matches no module stays unlinked. Macros participate in the graph too: a `-define` body's calls belong to the macro, each `?MACRO(...)` use site links into the call chain (and bare `?CONSTANT` reads are tracked as references), so a call path hidden behind a macro — `set_password → ?SQL_UPSERT_T → sql_query_t` — traces end-to-end and "where is this macro used" is answerable. escripts index like any module (the shebang line is understood), and OTP application resource files (`.app.src`, `.app`) join the graph: `{mod, ...}` links an app to its callback module and `{applications, [...]}` connects umbrella sibling apps — resolving only ever to modules, so an OTP app name like `ssl` is never mistaken for a same-named function. Truly dynamic dispatch (`Mod:handle(...)`, message sends, var-module spawns) is deliberately left unlinked rather than guessed. `codegraph_explore` also understands Erlang-native symbol spelling in queries — `mod:fn/3` and `init/2` find the symbols they name. (#635, #648)
- CodeGraph now indexes **Visual Basic .NET** (`.vb`) — classes, Modules, interfaces, structures, enums, properties, events, `MustOverride` abstract members, and `Declare` P/Invoke signatures, with `Inherits`/`Implements` hierarchy edges, call edges (resolved through VB's ambiguous call-vs-index parentheses), and `New`/`As New` instantiation links. Real-world VB styles parse cleanly: WinForms designer files, interpolated and multi-line strings, XML literals (embedded `<%= %>` expressions included), single-line and multi-line LINQ queries, multi-line lambdas, `Handles`/`WithEvents` event wiring, Custom Events, date literals, classic type-character identifiers (`i%`, `name$`), and non-English (Unicode) identifiers. (#648, #639, #170)
- CodeGraph now indexes **COBOL** (`.cbl`, `.cob`, `.cpy`) — programs, sections and paragraphs with `PERFORM`/`GO TO` call edges, `CALL` cross-program calls, `COPY` copybook imports (standalone copybooks included), and DATA DIVISION records with 88-level condition names, in both fixed and free source format. Impact queries work on data items: every `MOVE`/`ADD`/`COMPUTE`/`SUBTRACT` write-site links back to the field it changes, so "what touches this copybook field" answers across programs. CICS flows connect too: `EXEC CICS LINK`/`XCTL` program targets, `EXEC SQL INCLUDE` copybooks, and pseudo-conversational `RETURN TRANSID(...)` hops resolve to the program owning the transaction id. (#590, #648)
- CodeGraph now indexes **CFML** (`.cfc`, `.cfm`, `.cfs`) — both the classic tag-based style (`<cfcomponent>`/`<cffunction>`) and modern bare-script `component { ... }` syntax, including `extends`/`implements`, embedded `<cfscript>` blocks (at any nesting depth, including inside `<cfif>`/`<cfloop>`/`<cftry>`), call edges, and calls embedded in `#hash#` expressions inside `<cfquery>` SQL bodies. Files saved with a UTF-8 byte-order mark and tags with unquoted attribute values — both common in long-lived CFML codebases — are handled too. Thanks @ghedwards. (#1118)
- CFML inheritance written as a component path now links to the right component. `extends="coldbox.system.web.Controller"` names its supertype by dotted path and `extends="../base"` by relative path (the FW/1 style) — both previously produced no inheritance edge at all, which on framework-style CFML apps hid most of the type hierarchy from impact and blast-radius analysis (on ColdBox's own core, over 90% of inheritance was invisible). Resolution is deliberately conservative: the target's directory layout must corroborate the declared path — so a supertype that lives in an out-of-repo library (testbox, mxunit, an installed framework) correctly stays unlinked rather than being guessed at, and an ambiguous path produces no edge rather than a wrong one. (#1152)
- CFML method calls made through a local variable, typed argument, or component property now resolve to the right method — the same receiver-type inference the other object-oriented languages already had. `var svc = new UserService(); svc.save()`, `createObject("component", "path.UserService")`, a typed `<cfargument>` or cfscript parameter, and `variables.`/`this.`-scoped fields — including the pseudoconstructor pattern (`variables.svc = new UserService()` in `init()`) and WireBox-injected properties (`property name="svc" inject="UserService"`) — all now link the call to the declared component's method, with methods inherited from a supertype resolved through the inheritance links above. This makes callers, impact/blast-radius, and `codegraph_explore` flow traces follow CFML service calls instead of dropping them or guessing among same-named methods.
- The Claude Code context hook now recognizes prompts that describe code in plain words — in any language — by checking the prompt's words against the symbol names actually in your project's index. Asking about "the state machine des commandes" finds `OrderStateMachine` with no keyword involved. Confidence decides how much gets injected: structural questions and prompts naming a real symbol still get full context up front; a plain-words match gets a short pointer to the matching symbols so the agent queries them itself; everything else stays silent, exactly as before.
- Anonymous usage telemetry now counts how often the context hook injected context, offered a hint, or stayed silent — fixed counter names only; the prompt's content is never stored or sent. This makes the hook's accuracy measurable instead of guessed. The counters record what actually happened, not what was attempted: a lookup that errors or comes back empty counts as a distinct silent outcome, never as delivered context (#1143, thanks @inth3shadows).
- Metal shader files (`.metal`) are now indexed. Metal Shading Language is close enough to C++ that vertex/fragment/kernel functions, structs, type aliases, and the calls between them all land in the graph — so shader pipelines in Apple-platform projects show up in impact analysis and flow traces instead of being silently skipped. Metal's `[[buffer(0)]]`-style attribute annotations are handled so they can't corrupt what gets extracted. Thanks @FluxKo for the report. (#1121)
- CodeGraph now indexes legacy **iBatis 2** SQL maps (`<sqlMap>`), not just MyBatis 3 `<mapper>` files. `<select>`/`<insert>`/`<update>`/`<delete>`, iBatis's `<statement>`/`<procedure>`, and `<sql>` fragments inside a `<sqlMap>` become searchable statement symbols — for both namespaced maps and the namespace-less `Map.statement` id style — and `<include>` references resolve to the fragment they pull in, so search, callers, and impact queries return results on iBatis codebases that previously produced no statement symbols at all. Thanks @ESPINS for the report and the reproduction. (#1182)
- You can now force gitignored first-party source **into** the index with an `include` list in `codegraph.json`. The case this solves: a project tracked by a second VCS (SVN, Perforce, …) alongside Git, where some real source is committed to that VCS and deliberately listed in `.gitignore` so it never lands in Git — git never lists those files, so CodeGraph never indexed them, and neither `includeIgnored` (which only revives *embedded git repositories* inside a gitignored directory) nor `exclude` (its opposite) could help. Add a root `codegraph.json` with, e.g., `{ "include": ["Tools/", "Local/typescript/"] }` and CodeGraph discovers those files directly off disk — overriding `.gitignore` — and indexes them on the full index, incremental `sync`, and file-watching, on both git and non-git projects. Patterns are gitignore-style and matched against project-root-relative paths (a directory, a recursive `**` glob, or a single file). An explicit `exclude` still wins, and built-in skips like `node_modules`, `dist`, and `.git` are never re-included. This complements the existing `exclude` (its opposite — keep tracked files *out*) and `includeIgnored` (opt *in* to gitignored embedded repos).

### Fixes

- Indexing a large Java or Kotlin **Spring** monorepo is dramatically faster. The reference-resolution phase — the bulk of a first-time `codegraph index` — could run for the better part of an hour on a big multi-module project and now finishes in a few minutes, producing the same graph. The cause: every `receiver.method()` call in the codebase was repeatedly scanning the project's entire set of configuration keys looking for a match — work that only ever applies to Spring `@Value` / `@ConfigurationProperties` bindings, never to ordinary method calls. As part of the same fix, a method call whose name coincides with a configuration key (say a `service.process()` call alongside a `service.process` entry in `application.yml`) is no longer mislinked to that config key — it now resolves to the actual method. Thanks @bayernjava for the report. (#1180)
- `codegraph init` at a parent repository whose `.gitignore` excludes its child repositories no longer silently indexes nothing and reports success. The "super-repo of gitignored child repos" layout — a top-level Git repo that `.gitignore`s each `service-*/` or `packages/*` child so `git status` stays quiet — used to index only the parent's few top-level files and print "Done" with 0 nodes, even though running `codegraph init` inside any child worked fine (CodeGraph respects `.gitignore` by default, so the excluded children were skipped). Now, when an index comes up empty, CodeGraph detects the gitignored child repositories that were skipped, names them, and — in an interactive terminal — offers to index them (writing an `includeIgnored` entry to `codegraph.json` and re-indexing on the spot); non-interactive runs print the exact `codegraph.json` snippet to add. Projects that legitimately keep gitignored reference clones out of a working index are never nagged: the offer only appears when the index would otherwise be empty. Thanks @small-thanks for the report. (#1156)
- The MyBatis mapper reader is sturdier on real-world XML. Single-quoted attribute values (`id='getById'`, legal XML and common in older mappers) are no longer skipped, so those statements make it into the graph. Statements and `<include>`s that were commented out with `<!-- ... -->` no longer produce phantom symbols. And two vendor-split statements — the same `id` with `databaseId="oracle"` / `databaseId="mysql"` — written on a single line no longer silently drop one of the pair. Thanks @ESPINS for the report, the reproductions, and the fixes. (#1182)
- `codegraph init` and `codegraph index` no longer get killed by the safety watchdog at the "Resolving refs" step on large method-name-heavy codebases (big Java/enterprise monorepos were the main victims, especially on slower machines). Resolution used to come up for air only every 500 references, so a dense stretch of expensive ones could starve the watchdog long enough for it to assume the process was stuck and kill a perfectly healthy index. Resolution now checkpoints after every reference, and two of the expensive steps got much cheaper: repeated method lookups on the same type are now cached, and source files are no longer re-split line-by-line for every call being resolved — indexing such repos is several times faster as a result. Generated or minified single-line files are also skipped during receiver-type inference instead of being scanned per call. Thanks @UchihaYong and @wangmeng-95 for the reports. (#1122)
- An index left incomplete by an interrupted run now heals itself on the next sync instead of silently staying wrong forever. If indexing died partway through resolving references (a crash, Ctrl-C, or the watchdog kill fixed above), the affected files still looked indexed but their caller/impact edges were missing — a too-small blast radius clustering by package or module, e.g. a Spring `@Resource`-injected method reporting 3 of its 10 real caller files — and because incremental syncs only re-resolve files that changed, the damage was permanent until a full re-index. Any sync (a watched file change, or a bare `codegraph sync`) now detects the leftover references and finishes resolving them, `codegraph status` warns when an index is in that state instead of passing it off as healthy, and a rare early-stop that could abandon resolution on repos whose first files reference only external libraries is fixed too. Thanks @KnifeOfLife for the report and the package-correlation observation that pinned it down. (#1187)
- The shared background server no longer shuts down out from under a live editor/agent session that simply hasn't queried CodeGraph in a while. A safety timer meant to reap an *abandoned* server — one whose client vanished without the connection ever closing — was reaping **any** server that saw no requests for 30 minutes, including a perfectly live session that just wasn't asking CodeGraph anything; that silently dropped the session (and every other session sharing the same background server) to a slower in-process mode for the rest of its life. On one machine over a day it fired 20 times on live sessions and caught zero real phantoms. The timer now checks whether the connected clients are actually still alive and only reaps when none of them are, so a quiet-but-live session keeps its shared server while a genuinely abandoned one is still cleaned up. (#1200)
- CodeGraph's background server no longer leaves a lingering Node process behind when your editor or agent kills its launcher during startup. If the app that started CodeGraph (Codex, Claude Code, or another MCP client) was killed within the server's first fraction of a second — a config probe, a cancelled request, a startup timeout — while keeping the connection's pipes open, the server could be handed off to the system before its orphan-detection watchdog had captured a reference point, leaving it running (idle, ~30 MB) until the launching app itself exited; over a long day of repeated launches these accumulated. The server now records its parent at the earliest possible moment, and the npm and standalone launchers pass the real host's process id down to it so it can watch the host directly instead of only the launcher that may already be gone. As a final backstop, a server that never receives a single request after starting now shuts itself down instead of waiting for the host to exit — tunable with `CODEGRAPH_STARTUP_HANDSHAKE_TIMEOUT_MS` (0 disables). Thanks @ruslan33321 for the report. (#1185)
- The automatic context hook for Claude Code now fires for structural questions asked in nearly thirty languages — French, Spanish, Portuguese, German, Italian, Dutch, Polish, Czech, Romanian, Hungarian, Greek, Swedish, Danish, Norwegian, Finnish, Russian, Ukrainian, Turkish, Indonesian, Vietnamese, Thai, Hindi, Arabic, Farsi, Hebrew, Japanese, Korean, and both simplified and traditional Chinese — instead of just English and simplified Chinese. Previously a natural question like "comment marche la state machine des commandes ?" injected nothing unless it happened to contain a code-shaped symbol name, making the hook look broken for non-English teams. English questions phrased with derived word forms ("explain the architecture…", "what are the dependencies…") now fire too, and prompts in any other language still fire when they name a symbol from the index. Thanks @anthonyle-roy-lgtm for the report. (#1126)
- Lua and Luau method calls with capitalized names (`obj:Method()` — the standard Roblox convention) now link to the right method. Because Lua's method-call syntax looks identical to a Luau type annotation, a capitalized call like `lg:Log()` was misread as declaring the variable's type, so whenever two or more classes shared a method name (`Init`, `Update`, `Destroy`, …) the call was silently dropped from callers, impact/blast-radius, and flow traces. Lowercase method names were unaffected. Thanks @inth3shadows for the precise root-cause analysis and repro. (#1124)
- Removed dead code left behind by the discontinued managed-reasoning feature. Its `codegraph login` flow was unplugged before ever shipping in a release, but the unused module still shipped inside the platform bundles, and a security review flagged its Windows browser-open step (it routed the login URL through `cmd`, which would have been unsafe had the flow ever been wired back up). The leftover module and its tests are now fully deleted. Thanks @inth3shadows for the report. (#1114)
- The Claude Code context hook no longer treats ordinary English words that merely start with "call", "trace", "affect", or "connect" — callus, calligraphy, Connecticut, connective, affectionate, Tracey — as structural questions, which used to inject full CodeGraph context into prompts that had nothing to do with code structure. Genuinely structural forms (calls, callers, callbacks, call site, traced, tracing, affected, connections, connectivity, …) still fire exactly as before. Thanks @inth3shadows for the report. (#1138)
- A stuck git command can no longer hang CodeGraph indefinitely. The git checks behind worktree detection and git-hook setup, and the installer's optional `npm install -g` step, now time out and fail gracefully instead of blocking forever — this matters most for the background MCP server, where an unbounded git hang (network filesystems, a wedged fsmonitor) could previously freeze it long enough for the safety watchdog to kill it. Thanks @inth3shadows for the report. (#1139)
- The context hook's new plain-words matching works immediately on projects indexed by an older CodeGraph version. The word lookup it relies on is built at index time, so a project indexed before the upgrade had an empty one, and the hook would silently find nothing until something else happened to refresh the index; the hook now fills it in on first use (a one-time step — normally the background MCP server's startup catch-up gets there first). Thanks @inth3shadows for the report. (#1142)
- Several accuracy fixes to the plain-words matching: a renamed symbol (for example a NestJS route after its module prefix is applied) stays findable under its new name (#1141); a word that only appears in your code as an import statement's package name is no longer presented as a matched symbol (#1144); plural words no longer generate garbled lookup keys ("services" no longer also looks up "servic") (#1145); and a name matching both the singular and plural of one word can no longer squeeze out a genuine two-word match (#1146). Thanks @inth3shadows for the reports.
- Heavily-reflected Unreal Engine C++ classes are no longer dropped from the index. Reflection markup that decorates members — `UPROPERTY(...)`, `UFUNCTION(...)`, `UCLASS(...)`, `GENERATED_BODY()`, `UE_DEPRECATED_*(...)`, `DECLARE_DELEGATE_*(...)` — are no-semicolon macro calls that tree-sitter doesn't recognize, so each drops into error recovery; in a big class the errors pile up until the whole `class_specifier` collapses and the class, its base clause, and its members vanish (`UCharacterMovementComponent`, with ~240 such macros, disappeared entirely, breaking every subclass/type-hierarchy and blast-radius query that went through it). These line-leading annotation macros are now blanked (offset-preserving) before parsing so the class survives. Thanks @luoyxy for the report and root-cause analysis. (#1093 follow-up)
- Unreal Engine members and methods prefixed by an export/visibility macro are no longer lost. The `*_API` macro doesn't only sit on the class header — it prefixes almost every exported member of a large UE class (`ENGINE_API virtual void Tick(...)`, `static ENGINE_API void AddReferencedObjects(...)`); the parser read the macro as an extra type token and each such declaration fell into error recovery, so on headers like `Actor.h` and `World.h` hundreds of return types piled up as orphan errors and could still tip the class into collapse. Member/method-level `*_API` / `*_EXPORT` / `*_ABI` macros (Unreal, Qt/Boost, LLVM) are now blanked before parsing, mirroring the existing class-header recovery. (#1093 follow-up)
- Unreal Engine annotation macros that appear mid-line — an enum value's `UMETA(DisplayName=...)`, a parameter's `UPARAM(ref)`, or a deprecation tag wedged into a `using` alias (`using FOnNetTick UE_DEPRECATED(5.5, "...") = ...;`, which alone collapsed `UWorld` in `World.h`) — are now stripped too. These sit in positions the line-leading recovery structurally can't reach, and a single one could take down the surrounding enum or class. They are matched by an Unreal-only name list (`UMETA`, `UPARAM`, `UE_DEPRECATED*`) so no standard-C++ or other-library code is affected. Together these three fixes recover the main class of every large Unreal Engine header tested (`Actor`, `ActorComponent`, `SkeletalMeshComponent`, `World`, `LightComponent`, `CharacterMovementComponent`). (#1093 follow-up)
- A C++ header whose only C++ signal is an export-macro-annotated class is no longer misdetected as C — which had silently dropped the class. A `.h` file defaults to C and is only reclassified as C++ when it shows a C++-specific construct, but that check couldn't see through an export/visibility macro: `class ENGINE_API UFoo : public UObject` didn't match the macro-blind `class Name :` pattern, so a lean Unreal-Engine-style header carrying just `GENERATED_BODY()` and no explicit `public:` / `virtual` / `namespace` / `template` was parsed as C. The C extractor emits no class nodes, so the class — and its inheritance link — vanished from the graph, quietly undoing the macro-class recovery added in #1061. The C++ detection heuristic now recognizes an export-macro-annotated `class`/`struct` declaration (the same shape #1061 blanks before parsing), matching how `class Foo : Bar` was already detected; the two-token `<keyword> <MACRO> <Name>` before a `[:{]` never occurs in valid C, so genuine C headers are unaffected. Thanks @luoyxy for the report and root-cause analysis. (#1133)

## [1.2.0] - 2026-07-02

### New Features

- Method calls made through a local variable now resolve to the method in many more languages. When code does `const logger = new Logger(); logger.log();` (or the equivalent), CodeGraph infers the local variable's type from its declaration or initializer and links the call to the right method — so these calls now show up in callers, impact/blast-radius, and `codegraph_explore` flow traces instead of being dropped. Previously only C++ handled this; it now also covers TypeScript, JavaScript, Python, Java, C#, Kotlin, Swift, Go, Rust, Dart, Scala, and PHP. (#1108)
- Ruby method calls made on a receiver (`logger.log`) now record an edge to the method. Previously the Ruby indexer kept only the receiver and discarded the method name, so a method called through a variable or object had no recorded callers and was missing from impact/blast-radius and flow traces; combined with the local-variable type inference above, `logger = Logger.new; logger.log` now links to `Logger#log`. Calls to a class method (`Foo.bar`) and object construction (`Foo.new`) are still recorded too. (#1110)
- The same local-variable method-call resolution now extends to Lua, Luau, R, and Pascal/Delphi. A method invoked through a local — Lua/Luau `local lg = Logger.new(); lg:log()`, R `lg <- Logger$new(); lg$log()`, or Pascal `var lg: TLogger; ... lg.Log` — now links to the right method instead of being dropped. (#1112)

### Fixes

- Indexing a large project no longer gets killed partway through with a "Main thread unresponsive — killing the wedged process" message. The safety watchdog that stops a genuinely stuck index was mistaking slow-but-normal work for a hang: on a big repo, linking up references and cross-file relationships can legitimately run for a while, and that work now regularly yields so the watchdog can tell real progress from a true stall. Projects that previously failed to finish `codegraph init` / `codegraph index` (and had to fall back to `CODEGRAPH_NO_WATCHDOG=1`) now complete normally, while a genuinely hung process is still caught. Thanks @zmcrazy, @YoungLiao, and @GeeLab-Mob for the reports. (#1091)
- On Windows, a console window no longer briefly flashes when CodeGraph runs as a background MCP server. When the npm launcher started the bundled runtime — which happens every time an editor starts the server or reconnects after the daemon idles out — and during its self-heal step that extracts a missing platform bundle, Windows would pop up a black console (conhost) window for a moment. Both now launch hidden, matching how the daemon already behaved; the browser-open step of `codegraph login` was hardened the same way. Thanks @luoyerr for the report and root-cause analysis. (#1092)
- C++ forward declarations no longer crowd out the real class definition. A `class Foo;` forward declaration — common in large C++ and Unreal Engine codebases, where a heavily used class is forward-declared across dozens of headers — was indexed as its own class node every time it appeared. So exploring that class returned mostly forward-declaration sites, and could even pick one of them as the representative for blast-radius, burying the actual definition and its members and callers. Bodiless forward declarations are now skipped for C and C++, exactly as forward-declared structs and enums already were, so only the real definition is indexed. Languages where a class with no body is a complete definition — such as Kotlin's `class Empty` and Scala — are unaffected. Thanks @luoyxy for the report and root-cause analysis. (#1093)
- C++ methods that return a reference, and user-defined conversion operators, are now indexed under their correct names. An inline getter like `const FGameplayTagContainer& GetActiveTags() const` — everywhere in Unreal Engine headers — was indexed as `& GetActiveTags() const` instead of `GetActiveTags`, and a conversion operator like `operator EALSMovementState() const` kept its trailing `() const` instead of reading `operator EALSMovementState`. In both cases the garbled name meant you couldn't find the symbol by name and its callers weren't linked. Both now read cleanly, matching how pointer-returning and value-returning methods already worked. (#1096)
- C++ functions written with an inline-specifier macro before the return type are now indexed correctly. In Unreal Engine, inline helpers are commonly written `FORCEINLINE FString GetEnumerationToString(...)`; the `FORCEINLINE` macro made the parser read the return type as part of the function's name (`FString GetEnumerationToString` instead of `GetEnumerationToString`) and lose the real return type, so the function couldn't be found by name and its callers weren't linked. CodeGraph now recognizes the standard Unreal inline macros (`FORCEINLINE`, `FORCENOINLINE`, `FORCEINLINE_DEBUGGABLE`), so both the name and the return type are captured. (#1100)
- The same function-name recovery now covers inline macros from common third-party C++ libraries, not just Unreal Engine — including pugixml (`PUGI__FN`, `PUGIXML_FUNCTION`), Godot (`_FORCE_INLINE_`), Boost (`BOOST_FORCEINLINE`), and generic `ALWAYS_INLINE` / `FORCE_INLINE`. Functions decorated with these are now indexed under their real names. On a large Unreal project vendoring these libraries this cleaned up the large majority of remaining function-name garbling. (#1101)
- C++ function names are now recovered even when decorated with a macro CodeGraph doesn't specifically know about. A function written `SOME_LIBRARY_MACRO ReturnType doWork(...)` previously had the macro or return type absorbed into its name whenever the macro wasn't one CodeGraph recognized; now the real name (`doWork`) is recovered regardless of the macro, so it's findable and its callers link — no per-library configuration needed. The recognized-macro list was also broadened (Qt, Folly, Abseil, LLVM, V8, Eigen, rapidjson) so those additionally capture the return type. This only ever cleans up an already-garbled name and is limited to C and C++, so ordinary names — and languages like Kotlin and Scala where identifiers can legitimately contain spaces — are unaffected. (#1102)
- The set of C++ libraries whose macros are recognized for full return-type recovery was expanded well beyond Unreal Engine — now spanning Mozilla, Protobuf, {fmt}, nlohmann/json, GLM, Bullet, Skia, OpenCV, EASTL, Cocos2d-x, GLib, SQLite, and the common Windows calling conventions (so `HRESULT WINAPI CreateThing(...)` indexes as `CreateThing` returning `HRESULT`). Functions from libraries not on the list still get their name recovered automatically; being listed additionally recovers the return type. (#1103)
- Graph traversal and blast-radius results no longer drop or miscount relationships in a handful of edge cases. When a symbol could be reached by more than one path, an impact/blast-radius query could leave out a direct dependency between two symbols that were already linked another way; separately, the lower-level graph traversal used by the library API could keep only one of several relationships between the same pair of symbols (for example a symbol that both calls and references another), count a caller reached through two different call sites twice, or return slightly more results than the requested size limit on a very highly-connected symbol. These were long-standing and mostly masked by later de-duplication, so day-to-day query results were largely unaffected, but the traversal now returns the complete, correctly-bounded set. Thanks @inth3shadows for the precise, individually-traced reports. (#1086, #1087, #1088, #1089, #1090)
- Method calls to same-named classes in different files now resolve to the right definition. If two files each declared, say, a `Logger` class with its own `log()` method, a call could be linked to whichever definition happened to be indexed first — so a call in one file wrongly pointed at the class in another, mixing up that method's callers and blast radius. This affected calls written as `obj.log()`, `Logger.log()`, and `Logger::log()` across many languages, including C++, Python, TypeScript, Java, C#, and Rust. When a method name is ambiguous, CodeGraph now prefers the definition in the calling file itself — the correct target in the common case — while Java/Kotlin calls that an `import` already pins to another file are unaffected. Thanks @inth3shadows for the minimal repro and root-cause analysis. (#1079)
- Live auto-sync now gives up cleanly if indexing keeps failing, instead of retrying forever in the background. If a repeatable indexing error kept every sync from completing — for example a file that reliably crashes one language's parser, a full disk, or a corrupt database — a long-running server or daemon would retry every couple of seconds indefinitely, quietly wasting work and filling the logs while the graph silently stopped updating. Auto-sync now backs off after repeated failures and, if they persist, disables itself with an actionable message (naming the underlying error and pointing you to run `codegraph sync`), so the stalled index is surfaced rather than hidden — matching how prolonged file-lock contention and watch-limit exhaustion already degrade. A single successful sync resets everything, so an occasional transient hiccup is unaffected. Thanks @inth3shadows for the precise, code-traced report. (#1127)
- Method calls made through a typed function parameter now resolve to the right method in more languages. A call like `function use(lg: Logger) { lg.log(); }` wasn't linking to `Logger`'s `log()` — CodeGraph inferred a receiver's type from a local declaration but not from a bare typed parameter, so when another class defined a method of the same name the call was dropped from callers and impact/blast-radius. Typed parameters (and method receivers) are now recognized in TypeScript, JavaScript, Rust, Go, Dart, and PHP — including TypeScript generic types (`repo: Repository<User>`) — matching how Java, C#, Kotlin, Swift, Scala, Python, and Pascal already handled it, so the parameter case is now covered across every language that has typed parameters. Thanks @inth3shadows for the isolated repro and root-cause analysis. (#1125)

## [1.1.6] - 2026-06-30

### Fixes

- The standalone installer (`install.sh`) no longer leaves old versions piling up on disk. Each upgrade installed the new release into its own directory and re-pointed the launcher at it, but never removed the previous ones — so on macOS and Linux a full vendored Node runtime (tens of MB per version) accumulated with every update. The installer now keeps only the version it just installed and removes the older ones automatically (the npm installer's download-fallback cache prunes the same way). Windows installs already replaced a single directory in place, so they were never affected. Anything still left behind under `~/.codegraph/versions` from earlier upgrades is safe to delete. Thanks @lalanbv for the report. (#1074)
- `codegraph index` can now rebuild an existing oversized index from an older version, instead of hanging until the watchdog kills it. The previous fix (#1065) stopped *new* indexes from sweeping in a gitignored corpus of nested repos, but a project that had already built the multi-gigabyte graph before upgrading couldn't recover: `codegraph index` is meant to rebuild from scratch, yet it cleared the old graph by deleting every row one at a time, and on a graph of well over a million symbols that took longer than the 60-second responsiveness watchdog allows — so the command was killed before indexing even started, leaving the bad index in place. A full re-index now discards the old database outright and starts fresh, which is near-instant regardless of the old size and also frees the disk the bloated database was holding. Thanks @AriaShishegaran for the detailed follow-up report. (#1067)

## [1.1.5] - 2026-06-30

### Fixes

- C++ classes annotated with an export or visibility macro are now indexed as real classes. This is the `class MYMODULE_API UMyComponent : public UActorComponent` style used throughout Unreal Engine — where an `XXX_API` macro sits between `class`/`struct` and the type name — as well as the equivalent `*_EXPORT` / `*_ABI` macros common in Qt, Boost, LLVM, and many other libraries. Previously that macro made the parser misread the whole declaration as a function, so the class was dropped entirely: it never appeared in the graph and its base class went unrecorded, which made "find subclasses", type-hierarchy, and impact-through-inheritance queries come back empty for effectively every gameplay class in an Unreal Engine project. The class, its members, and its inheritance link are now all captured. Thanks @luoyxy for the detailed report and proposed fix. (#1061)
- `codegraph_explore` now surfaces the options/config type behind a function when you ask, in plain language, what to change to add a parameter to it. A question like "what do I need to change to add a new parameter to X" shares no words with the file that actually defines X's options — for example a functional-options struct and its `With…` builders living in a separate `options.go`, reachable only through X's signature — so that file scored near-zero on every text and connectivity signal and got dropped: explore returned X itself but not the file you'd edit, and the agent fell back to grep. Explore now follows a named function's parameter and return types and pulls in the file that defines them when ranking would otherwise bury it, so the options/config file shows up with its fields. Well-connected types that already rank are left untouched, so ordinary "how does X work" flow questions are unchanged. (The separate tools `codegraph_search`/`codegraph_impact`/`codegraph_node` remain available via `CODEGRAPH_MCP_TOOLS` for anyone who prefers driving each step explicitly.) Thanks @wauxhall for the detailed investigation. (#1064)
- The standalone installers (`install.sh` / `install.ps1`) now warn when a different `codegraph` earlier on your PATH will run instead of the one they just installed. The usual cause is a leftover `npm i -g @colbymchenry/codegraph` from an earlier version, whose launcher keeps running its own pinned version — so the installer reports installing the latest while `codegraph --version` keeps printing the old one, with no hint as to why. The installer now points at the shadowing copy and tells you how to resolve it (remove the other install, or put the new one first on PATH). Thanks @SpringYear for the report. (#1071)

## [1.1.4] - 2026-06-29

### Fixes

- CodeGraph again respects `.gitignore` for nested repositories that git tracks as gitlinks. The recent change that taught CodeGraph to descend into nested repos recorded as `160000` "commit" pointers (#1031, #1033) did so even when your `.gitignore` excludes the directory those repos live in — so a gitignored reference or benchmark corpus full of cloned repositories got pulled into the index anyway. One project with a gitignored `benchmark/repos/` of 19 cloned repos saw over 138,000 files swept in and a 4.8 GiB graph, and a full index then stalled in the "Resolving refs" phase until the watchdog killed it. CodeGraph now treats a gitignored gitlink the same as any other gitignored embedded repo: excluded by default, and re-included only when you opt the directory in with `codegraph.json` `includeIgnored`. Nested repos in non-ignored locations — the case #1031/#1033 fixed — are unchanged. Thanks @AriaShishegaran for the detailed report. (#1065)

## [1.1.3] - 2026-06-29

### Fixes

- CodeGraph now indexes nested repositories that git records as gitlinks, so a workspace built by stacking several repos inside one another indexes completely from a single `codegraph init` at the top. When a repo contains another git repo that was `git add`ed into it — so git tracks it as a `160000` "commit" pointer rather than a folder of files — or a submodule that isn't an active, initialized submodule in your checkout, that nested repo's source used to be skipped entirely: indexing the top level stopped at the nested repo's boundary and pulled in only the outer repo's own files, so a stacked-repo project came up nearly empty (one report saw ~10 files indexed at the root). CodeGraph now descends into each such nested repo that has a real working tree on disk and indexes it as its own embedded repository, recursively, so every layer of a stacked workspace is covered. Active submodules (already handled) and plain untracked nested clones are unchanged; a nested repo under a dependency directory such as `vendor/` or `node_modules/` stays excluded; and a submodule with nothing checked out on disk is correctly left alone rather than reported as empty. Thanks @ofergr and @kun-yx for the reports. (#1031, #1033)
- CodeGraph no longer shows a misleading "different git working tree" warning when you work inside a submodule (or other nested repo) of a workspace you indexed at its root. Because indexing a workspace now pulls in its submodules and embedded clones, a query run from inside one correctly resolves up to the workspace's single index — but it was still warning that the results came from "a different working tree" and suggesting you run `codegraph init -i`, which would have split the submodule back out into its own separate index and undone the unified view. CodeGraph now recognizes that the nested repo's code is already part of the workspace index and stays quiet. The warning still appears for a genuine git worktree — a second checkout of the *same* repository on another branch, which really does have its own uncommitted symbols — since that's the case it exists for. (#1031, #1033)
- On Windows, CodeGraph's background server now shuts down cleanly instead of occasionally aborting with a crash error. When the indexed project contained a nested repository (a submodule or embedded clone), stopping the server could race the file watcher's teardown and exit with a Windows crash code rather than a clean exit. Shutdown now lets that teardown finish first, so the server stops cleanly and promptly. (Windows only; other platforms were unaffected.) (#1033)
- C++ classes that inherit from a templated base — `class Widget : public Base<int>`, a CRTP base like `class App : public CRTPBase<App>`, or a struct inheriting a template — are now linked to that base class in the graph. Previously the template arguments (`<int>`) made the inheritance go unrecognized, so these classes looked like they inherited from nothing and impact/callers analysis stopped at the boundary; the connection is now followed like any other base class. Thanks @ryancu7 for the report. (#1043)
- C++ objects constructed on the stack — `Calculator calc(0)` or `Widget w{1, 2}` — now record that the enclosing function instantiates that class, the same as heap construction (`new Calculator(0)`) already did. Previously only the `new` form was tracked, so a function that built objects with the ordinary stack syntax looked like it didn't construct them and the dependency was missing from impact/callers. Thanks @Dshuishui for the report. (#1035)
- The graph no longer stores duplicate copies of the same relationship. The same dependency between the same two symbols at the same spot could be recorded more than once, which inflated edge counts and let callers/impact results list a relationship twice. Each relationship is now stored exactly once, and existing projects are de-duplicated automatically the next time CodeGraph opens them. Thanks @inth3shadows for the detailed report. (#1034)
- `codegraph node` can now read a file from the command line. File-read mode — pass `-f`/`--file` to get a file's source with line numbers plus the files that depend on it, the same output as the `codegraph_node` MCP tool — was rejected with "missing required argument 'name'", because the command always demanded a symbol name even though file mode has none, leaving the feature unreachable from the CLI. The symbol name is now optional: `codegraph node -f src/auth.ts` (or `codegraph node src/auth.ts`) reads the file, `codegraph node parseToken` looks up a symbol, and running it with neither prints a short usage hint instead of a cryptic error. Thanks @jcrabapple for the report. (#1044)
- `codegraph query` no longer prints meaningless relevance percentages like "12042%" next to each result. The number was a raw full-text search score — useful only for ordering the results, not as a real 0–100% figure — so multiplying it by 100 produced wild values that made the output look broken. Results are already listed best-match first, so the CLI now just shows them in that order with no score, matching what the search tool reports to AI agents. If you script against `codegraph query --json`, the raw `score` is still included for sorting or thresholding. Thanks @jcrabapple for the report. (#1045)
- `codegraph explore` no longer reports an alarming, inflated result count on broad natural-language queries. The "Found N symbols across M files" summary used to count every symbol the search swept in while ranking, so a broad query (for example "publish status to the API") on a large project could announce hundreds of symbols across a big fraction of the codebase — reading as if you had to wade through all of them — even though only the most relevant handful are actually shown with their source. The summary now counts just the files explore returns source for, so the number matches what you see. Ranking and results are unchanged: the right symbols still come first, and any further relevant files are still listed by name under "Not shown above" so nothing is hidden. Thanks @jcrabapple for the report. (#1046)
- Android resource files no longer bloat the index. A `res/` tree — layouts, drawables, value bags (strings, colors, styles), menus, navigation graphs — contains no code symbols, but on an Android app it can be the overwhelming majority of files (one project: 26,000+ XML files, ~97% of everything, contributing zero symbols), which inflated the database, slowed indexing, and padded file counts and `codegraph explore`/search results with entries that have nothing to find. CodeGraph now skips Android resource directories by default — `res/layout/`, `res/values/`, `res/drawable/`, `res/menu/`, and the rest, including their locale/density/version variants like `res/values-es/` or `res/drawable-hdpi/`. Your actual code is untouched, and so is the one kind of XML that does carry symbols — MyBatis mapper files, which live under `src/main/resources/`, not `res/`. `res/raw/` is deliberately kept (it can hold real assets), and you can re-include any excluded directory with a `.gitignore` negation such as `!res/values/`. Thanks @jcrabapple for the report. (#1047)

## [1.1.2] - 2026-06-28

### New Features

- You can now exclude committed directories from the index with an `exclude` list in `codegraph.json` — even when they're git-tracked. `.gitignore` can't drop a directory git already tracks, so a vendored theme or SDK that's checked into your repo (a committed Metronic theme under `static/`, a bundled vendor library) had no supported way to be kept out — it just bloated the graph and slowed indexing. Add a root `codegraph.json` with, e.g., `{ "exclude": ["static/", "**/vendor/**"] }` and those paths are skipped on indexing, sync, and file-watching, on both git and non-git projects. Patterns are gitignore-style and matched against repo-root-relative paths. This complements the existing `includeIgnored` (its opposite — opt *in* to gitignored embedded repos). (#999)
- CodeGraph now follows C/C++ commands that are dispatched through macro-built function-pointer tables, so the handler functions they reach are no longer dead-ends in the graph. Many C projects register a handler into a struct's function-pointer field through a macro and a generated table — redis is the classic case: every command (`getCommand`, `decrbyCommand`, …) is wired into the command struct's `proc` field by a `MAKE_CMD(…)` table that lives in a generated, `#include`-d file, then invoked as `c->cmd->proc(c)`. CodeGraph now reads those macro-built tables — including ones whose struct type is itself a macro alias, whose table sits in an `#include`-d file that is never indexed on its own, or that are wrapped in conditional compilation (`#ifdef`) and defined inline with the struct. It recognizes function-pointer fields declared through a function typedef, and follows the receiver — a chained access (`c->cmd->proc`) or an array subscript through a file-scope table (`(cmdnames[i].cmd_func)(…)`) — across field types. It also follows dispatch through a bare array of function pointers with no struct wrapper at all — the opcode/handler-table pattern common in interpreters and emulators, where a table like `opcodes[op](…)` invokes one of many registered handler functions by index — linking the dispatcher to every handler in the array. The upshot: asking for the callers or blast radius of a command handler now finds the dispatcher that reaches it. For redis, `call` shows up as a caller of every command; for SQLite, the builtin SQL functions registered through `FUNCTION(...)` link to where they're invoked; for Vim, every `:ex` and normal-mode command links from the dispatcher. (#991, extending #932)
- CodeGraph no longer times out when many agents query it at once. The shared background server that serves all your editor and agent sessions used to run every query on a single thread, so a burst of concurrent requests — for example a swarm of subagents exploring a large monorepo together — queued up behind one another and, while the heavy ones ran, froze the connection so finished answers couldn't even be sent back until the whole batch drained. Past a handful of simultaneous callers that routinely surfaced as MCP request timeouts. The shared server now answers queries across a pool of worker threads, so concurrent requests run in parallel and the connection stays responsive the whole time; when it's genuinely saturated a call returns a brief "busy, retry shortly" note (not an error) instead of hanging past your client's timeout. The pool sizes itself to your machine — roughly one worker per core, leaving one for coordination — and a single editor session is unaffected (no pool, no overhead). Set `CODEGRAPH_QUERY_POOL_SIZE` to choose a specific number of workers, or `0` to revert to single-threaded in-process queries.
- Indexing now parses files across multiple CPU cores instead of one, so building a project's graph — `codegraph index`, the first index of a project, and the background re-index after changes — is faster on multi-core machines, most noticeably on large or parse-heavy codebases. The graph it produces is identical to before and re-indexing stays deterministic: parsing runs in parallel, but results are still committed in a fixed order, so the same project always yields the same graph. CodeGraph sizes the pool to your machine automatically (leaving a core free for everything else); set `CODEGRAPH_PARSE_WORKERS` to choose a specific number of parse workers, or `CODEGRAPH_PARSE_WORKERS=1` to restore the previous single-core behavior. Peak memory is unchanged — workers reclaim parser memory independently, so it doesn't grow with the number of cores. (#1015, #320)
- When CodeGraph's MCP server runs with no default project of its own — started outside any repository (for example behind an MCP gateway), or at a monorepo root whose indexes live in sub-projects — it now marks `projectPath` as a required argument on every tool call. Before, `projectPath` was always optional, so an agent talking to such a server would often omit it, get back guidance to pass it, and not reliably retry — you had to nudge it by hand every time. Now the requirement is part of the tool definition the agent sees, so it supplies the path to the project it's working on the first time. When the server does have a default project — the normal case, launched inside your repo — `projectPath` stays optional and a call without it falls back to that project exactly as before. Thanks @wauxhall for the report. (#993)
- CodeGraph's MCP tools now work in Cursor's **Ask mode** (and any other client that only permits read-only tools). Every CodeGraph tool just reads your indexed code — it never changes your workspace — but it didn't advertise that, so Cursor's Ask mode blocked every call with "you are in ask mode and cannot run non read-only tools," and you had to switch to Agent mode to use CodeGraph at all. CodeGraph now declares all its tools read-only using the standard MCP tool annotations, so Cursor (and similar clients) allow them in read-only contexts. Nothing about how the tools behave changes. Thanks @CDsouza315 for the report. (#1018)

### Fixes

- A `codegraph index` or `codegraph init` that gets orphaned or wedged now stops itself instead of pinning a CPU core forever. If you killed the command (or the terminal/agent that launched it), the underlying indexer process used to keep running in the background — the parent couldn't pass the signal along — and a genuinely stuck index had nothing watching it either, since the self-recovery watchdogs were wired only into the background MCP server. Both gaps are closed: indexing now self-terminates when its parent goes away, and a main thread that stops making progress is killed so it can't hang indefinitely. Opt out with `CODEGRAPH_NO_WATCHDOG=1` (liveness) or `CODEGRAPH_PPID_POLL_MS=0` (orphan detection), matching the server. (#999)
- Indexing no longer hangs at "Resolving refs" on a repo that commits a large JavaScript/TypeScript theme or SDK. A vendored admin theme (Metronic is the classic case — ~1,300 committed `.js` files) re-declares the same method names (`init`, `update`, `render`, `destroy`, …) on hundreds of widgets, and resolution used to score *every* same-named definition against *every* call — work that grows with the square of how many times a name repeats. On such a repo it pinned a CPU core for 15–30 minutes and effectively never finished. Resolution now declines to guess when a name is defined more times than any real codebase ever repeats one (the cutoff is generous — normal projects top out far below it and are completely unaffected), since no proximity heuristic can pick the one true target among thousands anyway. Indexing that previously wedged now completes in seconds, and precise resolution (imports, qualified names, class-name matches) is unchanged. This is the same class of slowdown as the 1.1.0 import-name fix, now closed for repeated method/symbol names. Tune the cutoff with `CODEGRAPH_AMBIGUOUS_NAME_CEILING` if you ever need to. Thanks @DANOX2 for the detailed report and repro. (#999)
- Claude Code's front-load prompt hook now fires for non-English prompts. The optional hook that injects CodeGraph context for structural questions only recognized English keywords, so a structural question written in Chinese — or any non-Latin-script language — silently injected nothing: the hook looked like it wasn't wired up despite a correct setup, with no error to explain why. The gate is now language-aware. It recognizes Chinese structural keywords (如何/流程/调用/依赖/实现/架构…), and — in any language — a prompt that names a real code symbol from your project, such as `getUserId`, `article_publish`, `user.login`, or `parseToken()` (the name is checked against the index, so an ordinary word that merely looks like code doesn't trigger it). Non-structural prompts ("fix this typo", in any language) stay a no-op as before, so nothing fires where there's no structural answer to give. Thanks @whinc for the detailed report and repro. (#994)
- The background auto-sync server now starts for projects kept on an ExFAT or FAT external drive (and some network mounts). Those filesystems don't support the operations the server relies on to coordinate and to listen locally, so it failed immediately and re-logged the same error on every retry — background indexing was broken, so you had to run `codegraph sync` by hand after changes. (The MCP tools, the prompt hook, and manual `codegraph index`/`sync` were unaffected, since none of them need the server.) The server now works around those limitations automatically — falling back to a different coordination method and relocating its local socket to your system temp directory — so background indexing works there exactly like anywhere else, with no configuration needed. Verified end-to-end on real removable-drive filesystems on macOS, Linux, and Windows. Thanks @zengwenliang416 for the detailed report. (#997)
- If you use CodeGraph as a library, the `QueryBuilder.deleteResolvedReferences()` helper no longer throws "too many SQL variables" when handed a very large list of ids — it issued one unbounded query, so a list longer than SQLite's parameter limit aborted the call. It now splits the work into batches like every other bulk query in the API. CodeGraph's own indexing and reference resolution never called this method (they use a different, already-batched path), so the CLI and MCP server were unaffected. Thanks @inth3shadows for the static analysis. (#1001)
- Swift computed properties are now indexed, so you can search for them. A computed property — a `var isCloudProxy: Bool { … }` read all over a codebase, a SwiftUI view's `var body: some View { … }`, a protocol's `var title: String { get }` requirement — produced no symbol at all, so `codegraph query` and `codegraph_explore` answered "No results found" for it, and an agent that trusts an empty result would wrongly conclude the property doesn't exist. Computed properties (and protocol property requirements) are now graph symbols you can find and explore like anything else. A computed property's getter is also read as its body, so a SwiftUI view's `body` links to the subviews and helpers it builds — making a view's render flow traceable through the property. Stored properties were already indexed; this closes the computed-property gap. Thanks @monochrome3694 for the precise report and repro. (#1020)

## [1.1.1] - 2026-06-24

### Fixes

- CodeGraph respects your `.gitignore` again when looking for nested git repositories. A directory you've gitignored — a `resource/` or `.repos/` folder of cloned reference projects, a large vendored data dir — is no longer walked into and indexed: version 1.0.0 started searching inside *every* gitignored directory for embedded git repos and pulling them all in, which could multiply a graph many times over and slow indexing to a crawl on a multi-gigabyte folder of clones, even though you'd explicitly excluded it. Indexing the repos inside a gitignored directory is now opt-in — add an `includeIgnored` list to a `codegraph.json` at your repo root, e.g. `{ "includeIgnored": ["packages/", "services/"] }`, to index the embedded repos under the directories you name. The "super-repo of independent clones" layout from 1.0.0 still works, just declared explicitly. Nested repos you *haven't* gitignored (untracked clones) are indexed as before, and a project without this layout is unaffected. (#970, #976, #622)
- The MCP server no longer drops out with `Transport closed` when its connection to the shared background server hits a transient socket error. To serve all your editor/agent sessions from a single index, CodeGraph runs one shared background server they reach over a local socket; a stray error on that socket while a session was connecting used to take the whole server process down, so your agent saw a bare `Transport closed` even though `codegraph status` and `codegraph sync` were perfectly healthy. Now such an error is handled gracefully — the session falls back to serving itself in-process instead of crashing. This is most common on WSL2 when your project lives on a Windows drive (a `/mnt/c` or `/mnt/d` path), where that socket is flaky; if you still hit problems there, set `CODEGRAPH_NO_DAEMON=1` to skip the shared server entirely and run each session in its own process. Affects Codex and any other MCP client. (#974)
- A shared background server that can't bind its socket now cleans up its own lock file before exiting, instead of leaving a stale lock that the next launch trips over — which previously could let duplicate `serve --mcp` processes pile up for the same project. (#974)

## [1.1.0] - 2026-06-23

### New Features

- **Claude Code:** an optional front-load hook makes your agent reach for CodeGraph automatically. When you ask a structural question — "how does X work", "what calls Y", "trace the flow from A to B" — CodeGraph injects the relevant source and call paths into the prompt up front, so the agent answers from the graph instead of grepping around to rebuild it. You're asked during `codegraph install` (default yes; Claude Code only, since it's the agent with prompt hooks), it's removed by `codegraph uninstall`, and `codegraph upgrade` turns it on for existing Claude setups. It's strictly additive and degradable — non-structural prompts and un-indexed projects are left alone — and you can switch it off any time without uninstalling by setting `CODEGRAPH_NO_PROMPT_HOOK=1`.
- Vue store actions, mutations, and getters are now indexed as symbols you can find and read. Whether your store is **Vuex** (`mutations` / `actions` objects in a module) or **Pinia** — both the options form (`defineStore({ actions: { … } })`) and the setup form (`defineStore('id', () => { … })`, where actions are local functions) — each action, mutation, and getter is now a real node. So `codegraph search` finds `login` or `getSessionList`, and `codegraph_explore` / `codegraph_node` show its body and what it calls, instead of "not found" because the function only existed as an object-literal property.
- `codegraph_explore` now connects a Vue component to the **Pinia** store action it calls. When code does `const store = useUserStore()` and then `store.fetchUser()`, that call now links through to the `fetchUser` action in the store module — so "what happens when this view loads its data?" traces from the component into the action's body instead of stopping at the `store.fetchUser()` line. Works for both Pinia store styles (options and setup), and stays precise (a built-in like `store.$patch()` or an unrelated same-named method isn't mislinked).
- `codegraph_explore` now follows **Vuex** string dispatch. A `dispatch('user/login')` or `commit('SET_TOKEN')` call — namespaced `'module/action'` keys included — now links to the action or mutation it names, resolved to the correct store module even when several modules share an action name (and without being fooled by a same-named `api/` helper). So "what runs when this dispatches?" traces from the call into the store handler and on to the mutations it commits. Vuex's canonical `export default { namespaced, actions, mutations }` module shape is now indexed too, so those handlers are findable symbols.
- `codegraph_explore` now connects React data-fetching flows built on **RTK Query** (Redux Toolkit's `createApi`). An endpoint defined inside `createApi({ endpoints })` and the `useGetXQuery` / `useUpdateYMutation` hook it generates were both invisible to analysis — so "what does this component fetch?" or "where does `useGetThingQuery` get its data?" dead-ended, because the hook, the endpoint, and the component had nothing linking them. CodeGraph now indexes each endpoint and each generated hook as real symbols and wires the path `component → useGetXQuery → getX → queryFn`, so the flow resolves in one explore call instead of reading the API slice by hand. Both the arrow (`endpoints: build => ({ … })`) and method (`endpoints(builder) { return { … } }`) styles are recognized, along with the `useLazyGetXQuery` variant; hand-written hooks of a similar name are left untouched.
- `codegraph_explore` now follows **Celery** task dispatch in Python. A `send_email.delay(...)` or `send_email.apply_async(...)` call now links to the `@shared_task` / `@app.task` function it runs — typically defined in a different module (`tasks.py`) from where it's triggered (a view or service) — so "what actually happens when this is dispatched?" traces from the call site straight into the task body instead of stopping at the `.delay()` line. Both decorator dialects are recognized (bare `@shared_task` and the arg'd `@app.task(bind=True, …)` form), including the module-qualified `tasks.invalidate_cache.apply_async()` call style. It stays precise: a `.delay()` on something that isn't a Celery task is never mislinked, so a project that doesn't use Celery is unaffected.
- `codegraph_explore` now follows **Spring application events** in Java. A `publishEvent(new OrderShippedEvent(...))` call now links to every `@EventListener` that handles that event — usually in a different class — so "what reacts when this is published?" traces from the publisher straight into each listener method instead of dead-ending at `publishEvent(...)`. The link is by event type, and all the common listener styles are recognized: a `@EventListener` typed on its parameter, the `@EventListener(SomeEvent.class)` form, `@TransactionalEventListener`, and the older `implements ApplicationListener<SomeEvent>`. One event fans out to all its listeners, and a plain Spring app with no event bus is unaffected.
- `codegraph_explore` now follows **MediatR** request and notification dispatch in C#/.NET. A `_mediator.Send(command)` or `_mediator.Publish(notification)` call now links to the `Handle` method of the matching `IRequestHandler<>` / `INotificationHandler<>` — usually in a different file in a Clean Architecture layout — so "what handles this command?" traces from the controller straight into the handler instead of stopping at the mediator call. The sent type is recognized whether it's constructed inline (`Send(new GetFooQuery())`), built into a local first (`var cmd = new …; Send(cmd)`), or passed in as a parameter, and it's matched by type — so a `MessagingCenter.Send(...)` or a same-named DTO that isn't a request is never mislinked, and a project without MediatR is unaffected.
- `codegraph_explore` now follows **Sidekiq** background-job dispatch in Ruby. A `DestroyUserWorker.perform_async(id)` (or `.perform_in` / `.perform_at`) call now links to that worker's `perform` method — usually in `app/workers/` away from the controller or model that enqueues it — so "what runs in the background here?" traces from the enqueue straight into the job body. Both the modern `include Sidekiq::Job` and the older `Sidekiq::Worker` are recognized, namespaced workers resolve to the right class even when several share a name (e.g. `Comments::NotifyWorker` vs `Articles::NotifyWorker`), and Rails ActiveJob's `perform_later` — a different mechanism — is intentionally left alone.
- `codegraph_explore` now follows **Laravel events** in PHP. An `event(new OrderShipped($order))` call now links to every listener that handles it — each listener's `handle()` method, usually a separate `app/Listeners/` class — so "what reacts to this event?" traces from the dispatch straight into the listener bodies. Listeners are found both ways Laravel registers them: by a typed `handle(OrderShipped $event)` (auto-discovery, including a `handle(A|B $event)` union that listens for two events) and by the `protected $listen` map in your `EventServiceProvider` (which also catches a listener whose `handle()` has no type-hint). One event fans out to all its listeners, and queued jobs — dispatched via `::dispatch()` rather than `event()` — are correctly left out.
- CodeGraph now understands **Lombok**-generated methods in Java. `@Getter`, `@Setter`, `@Data`, `@Value`, and `@Builder` generate getters, setters, `builder()`, `equals`/`hashCode`/`toString`, and the `@Slf4j` `log` field at compile time, so those methods never appear in the source — and a `user.getName()`, `User.builder()`, or `log.info(...)` call used to resolve to nothing, silently breaking call-chain analysis (the agent would conclude the method didn't exist and reconstruct it by hand). Those members are now indexed from the annotations and fields, so they appear in `codegraph search` and `codegraph_explore`/`codegraph_node`, and callers trace through them like any hand-written method. They're marked as Lombok-generated so they read as generated, not hand-written; a method you write yourself is never overridden, static fields get no accessor, and a class without Lombok is unaffected. Thanks @git87663849. (#912)
- `codegraph_explore` now follows **C and C++ function-pointer dispatch**. C does polymorphism with function pointers: a struct carries a function-pointer field, concrete functions are registered into it through a table (`static struct cmd commands[] = {{"add", cmd_add}, …}`), a designated initializer (`.handler = on_open`), or an assignment, and the code dispatches indirectly (`p->fn(argv)`). None of that was visible to analysis — the indirect call resolved to nothing, so `git`'s command runner looked like it called nothing and a vtable's implementations had no callers. CodeGraph now links the dispatch site to the registered handlers, keyed by the struct field, so "what runs when this dispatches?" traces from `p->fn(...)` into every function registered for that field. This covers the command-table idiom (git, redis) and the ops-struct/vtable idiom (curl's content-encoders, protocol handlers), including the case where a generic hook slot is reassigned from a registry (`h->func = found->fn`). It stays precise — distinct function-pointer fields don't cross-link, a plain data field is never treated as a dispatch, and a project without function-pointer dispatch is unaffected. (#932)
- `codegraph_explore` now follows **GoFrame** route bindings in Go. GoFrame's standard router wires routes reflectively: the path and method live in a `g.Meta` struct tag on a request type (`` g.Meta `path:"/user/sign-in" method:"post"` ``), the controller method that serves it is matched by that request type, and the two are joined at runtime by `group.Bind(...)` — so there was no path string and no edge from a route to its handler, and "where is `/user/sign-in` handled?" or "where are the routes bound to controllers?" could only be answered by reading. CodeGraph now indexes each `g.Meta` route as a real route node and links it to the controller method whose signature takes that request type, so a route resolves to its handler structurally in one `codegraph_explore` call. The link is by request type, not method name — so it's correct even when the two differ (a `DeptSearchReq` served by a `List` method); it tells apart the many identical request types a large app defines one-per-module (`cash.ListReq` vs `order.ListReq`) by package, including cloned addon modules; and a route whose handler isn't present is left unlinked rather than guessed. (#747)
- The MCP server now works in monorepos and multi-project setups. Before, if you started CodeGraph somewhere with no `.codegraph/` of its own — most often a monorepo root where you only indexed individual services — the server exposed **no tools at all**, so your agent couldn't query CodeGraph even for the sub-projects that *were* indexed. Now the tools are always available: point a query at any indexed project with the `projectPath` argument (its path, or anywhere inside it) and CodeGraph answers from that project's index — for as many projects as you like in one session. It also means a project you index *after* the server started is picked up without restarting, where before the tools stayed hidden because the server only checked for an index once at launch. A project that genuinely has no index still cleanly tells your agent to use its built-in tools there (and that you can run `codegraph init` to enable it), so single-project use is unchanged. Thanks @MaiLunJiye. (#964)
- The Claude front-load hook now finds your indexed sub-projects in a monorepo. The optional `UserPromptSubmit` hook that injects CodeGraph context for structural questions previously only looked for an index at or above your working directory — so if you opened the monorepo root but indexed individual packages (`packages/api`, `services/auth`), it found nothing and stayed silent exactly where it was most useful. It now also looks *into* sub-projects: a single indexed sub-project gets its context front-loaded automatically, and with several the hook front-loads the one your question names (and lists the rest so the agent can target them by `projectPath`). Single-project repos are unaffected, and the scan is bounded and skipped entirely outside a recognizable workspace root. (#964)

- `codegraph_explore` now surfaces the right code in large multi-layer projects. When you ask a backend-flow question in a repo that pairs an API server with a big frontend that mirrors the same domain words — say an `app/` admin UI sitting over an `api/` server — the server-side file that genuinely matches several of your query's terms is no longer pushed out of the results by the larger, more interconnected frontend layer. A file corroborated by two or more distinct query terms is now kept in the answer even when a denser unrelated layer would otherwise crowd it out, so "how does X read items / handle the request" returns the service or handler that does the work instead of a wall of frontend views. Single-layer projects are unaffected; set `CODEGRAPH_RANK_NO_MULTITERM=1` to revert to the previous ranking.
- Impact and blast-radius analysis for TypeScript, JavaScript, Go, Python, Rust, Ruby, C, Java, C#, PHP, Scala, Kotlin, Swift, Dart, and Pascal/Delphi now understands the readers of a constant. When you change a file-scope, package-level, module-level, or class-level constant — a config object, a lookup table, a shared constant — the other symbols in that file that read it now show up as affected, where before they were invisible (impact only followed calls, imports, and inheritance, so a constant's consumers looked like "nothing depends on this"). This makes `codegraph impact`, and the impact trail in `codegraph_explore`/`codegraph_node`, catch the "change this table, break its readers" class of change. It's on by default and adds no nodes to your graph; bundled/minified files and ambiguously-shadowed names are skipped to keep results precise. Set `CODEGRAPH_VALUE_REFS=0` to turn it off.
- C file-scope constants and globals — `static const` scalars, pointer/array lookup tables, and shared mutable globals — are now recognized as symbols in their own right. They previously weren't extracted at all, so they never appeared in search or carried any dependents; now they show up in `codegraph search` and participate in impact analysis (see above), so changing a C lookup table surfaces the same-file functions that read it.
- Java `static final` constants, C# `const` / `static readonly` constants, Scala `object` vals, and Kotlin top-level / `object` / `companion object` `val`s are now classified as constants rather than generic fields, so they participate in the constant-reader impact analysis above — change a `public static final` table, a `const string`, a Scala `object Config { val Timeout = … }`, or a Kotlin `companion object { const val … }` and the methods that read it now show up as affected. (Per-object Java `final` / C# `readonly` / Scala & Kotlin `class` instance properties are unchanged.) Kotlin constants were previously not indexed as their own symbols at all, so they now also appear in `codegraph search`.
- Swift top-level `let`s and `static let` constants (including those namespaced in an `enum`/`struct`, the common Swift pattern) are now indexed as constants and participate in the constant-reader impact analysis above — change a `static let defaultRetryLimit` or an `enum Constants { static let … }` and the same-file code that reads it shows up as affected. Computed properties and per-instance `let`s are not treated as constants.
- Dart top-level `const`/`final` and class `static const`/`static final` constants are now indexed as constants and participate in the constant-reader impact analysis above. Instance fields, `var`s, and locals are not treated as constants. (Generated Dart code with the standard `.g.dart`/`.freezed.dart`/`.pb.dart` suffixes is already skipped.)
- You can now teach CodeGraph about custom file extensions. Drop a `codegraph.json` at your repo root with an `extensions` map — `{ "extensions": { ".dota_lua": "lua", ".tpl": "php" } }` — and files with those extensions get indexed under the language you name, instead of being silently skipped because the extension wasn't one of the built-in defaults. It's opt-in and committed alongside your code so the whole team shares it, your mappings layer on top of the built-ins and win on conflict (you can even re-point a built-in, e.g. `.h` → `cpp`), and a typo'd language or a malformed config is warned about and skipped rather than breaking indexing. Projects without a `codegraph.json` behave exactly as before. (#906)

### Fixes

- `codegraph index` and `codegraph init` no longer crawl during the "Resolving refs" phase on large projects — most painfully ones that mix a big front-end and back-end, where the phase could stretch to many minutes. A package or module imported across hundreds or thousands of files (`react`, a shared UI package, Python `logging` / `typing`) was being treated as if every one of those import statements might be its definition, so the resolver compared each import against all the others — work that grows with the *square* of how widely a package is imported, which is why it blew up only on big, import-heavy repos. Imports now resolve straight to the definitions they actually point at, so those redundant comparisons are gone (reference resolution is dramatically faster on large repos), and the graph no longer accumulates the meaningless import-to-import links the old fallback created. (#915)
- MCP tool results no longer show up as oversized headings in Markdown-rendering clients (such as the Claude Code VSCode extension). Results used Markdown headings (`##`/`###`) for things like the status summary, each search hit, and every file section in an exploration, so a normal query filled the transcript with large-font lines — worst with `codegraph_search` and `codegraph_explore`, where the noise grew with the number of results. Section headers are now bold labels, which render at normal text size while keeping the same structure. Terminal/CLI output is unchanged. (#778)
- An MCP server pointed at a very large repository (tens of thousands of files) no longer hangs on the first tool call after a fresh start. On startup CodeGraph reconciles its index against the current files on disk, and on a huge repo that reconcile could run for minutes while blocking the very first request — long enough that the background server was sometimes force-restarted mid-scan, so the first query never came back at all. The reconcile now yields as it runs (keeping the server responsive instead of pinning it), and the first tool call waits only briefly for it before answering and letting the rest finish in the background — so you get a fast first response and the index still catches up. Set `CODEGRAPH_CATCHUP_GATE_TIMEOUT_MS` to tune how long that first call waits (default 3000ms), or `=0` to always wait for the full reconcile. (#905)
- `codegraph install` now wires up your agents and stops there — it no longer indexes the current directory. Building a project's graph is always the explicit `codegraph init` (or `codegraph index`), so you decide what gets indexed and when, and the steps are the same whether you installed globally or just for one project. This clears up the confusion where a project-local install silently indexed but a global one didn't, and where the docs and the tool disagreed about whether you still had to run `init`. (#826)
- React components declared with `forwardRef`, `memo`, or styled-components / emotion (`const Button = forwardRef(...)`, `const Card = memo(...)`, `const Box = styled.button\`…\``) are now recognized as components, so finding where they're used works. Before, they were indexed as plain constants, so `codegraph callers` and impact analysis reported "no callers found" even when the component was rendered across dozens of files — a dangerous false "safe to change" right before refactoring a shared component. Now every `<Button/>` usage links back to the component, so callers and blast radius are complete. This is the standard shadcn/ui declaration style, so for typical React design systems the whole UI layer is no longer invisible to impact analysis. Thanks @Arlandaren for the report and @maxmilian for the root-cause. (#841)
- React Router and Next.js routes defined in `.tsx` / `.jsx` files are now indexed. Routes written as JSX — `<Route path="/users" element={<UsersPage/>}/>`, `createBrowserRouter([...])`, and Next.js `app/`/`pages/` page files — were being skipped entirely (only routes that happened to live in plain `.ts`/`.js` were picked up), so "what renders at this path?" and the route → page-component link were missing for most React apps. Now those routes show up in `codegraph search`/`codegraph_explore` and connect to the component they render, just like the backend route → handler links on other frameworks.
- `codegraph sync` (and the file-watcher auto-sync behind always-on / MCP use) no longer drops a function's callers when you edit the file that function lives in. Re-indexing a file after an edit — even a docstring-only change — was severing the `calls` / `references` edges coming from *other, unchanged* files (e.g. `from pkg import mod; mod.fn()` callers, or any cross-file reference), so `codegraph callers` / `impact` would abruptly report "no callers" for a function that's used throughout the codebase, until the next full re-index. Sync now preserves those incoming cross-file edges across the re-index, re-matching them to the symbol even when the edit shifted its line number. Most visible in large Python package trees, but it applies to every language. Thanks @JosefAschauer. (#899)
- `codegraph index` now rebuilds the full graph from scratch, so it produces the same result as a fresh `codegraph init` instead of reporting "0 nodes, 0 edges" and looking like it wiped your index. Previously, re-running `index` on an unchanged project skipped every file (their contents hadn't changed) and showed an empty-looking summary; it now clears and re-indexes for an honest, complete rebuild every time. Use `codegraph sync` for fast incremental updates between full rebuilds. Thanks @Arc-univer. (#874)
- The file watcher that auto-syncs the graph now fails cleanly when live watching can no longer be trusted, instead of looking healthy while the index quietly goes stale. If the operating system runs out of file-watch resources, or another process holds the write lock far longer than a normal save, CodeGraph now disables auto-sync once — with a single clear message telling you to run `codegraph sync` (or rely on the git sync hooks) to refresh — rather than retrying forever or repeating the same error on a loop. And while auto-sync is disabled, CodeGraph's tool responses (and `codegraph status`) now say so plainly, so your AI agent knows to read files directly instead of trusting a frozen index. This mostly matters for long-running MCP/daemon sessions, which could otherwise keep serving stale results while appearing to work. Thanks @thismilktea. (#876)
- On Linux, hitting the kernel's inotify watch limit on a large project no longer silently leaves half the tree unwatched. CodeGraph now tells you once — naming the exact setting to raise (`fs.inotify.max_user_watches`, e.g. `sudo sysctl fs.inotify.max_user_watches=1048576`) — and keeps live-watching the directories it could register while `codegraph sync` (or the git sync hooks) covers the rest. (#876)
- A long-running MCP server now notices when a git worktree gains its own index. Before, if the server (or shared daemon) first saw a worktree before you ran `codegraph init` in it — so the lookup walked up to the main checkout's index — it pinned that decision for its whole life: even after the worktree had its own `.codegraph/`, every query kept hitting the main checkout's index and every result carried a false "this index belongs to a different git working tree" warning, until you restarted the server. The CLI got it right but the MCP server didn't, and re-indexing didn't help. The server now re-checks which index a path belongs to on each call, so the worktree's own index is picked up — and the stale warning drops — without a restart. (#926)
- A long-running MCP server now recovers when your index is deleted and rebuilt at the same path. If `.codegraph/` was removed and recreated while the server held it open — most easily by recreating a git worktree at the same path, or `rm`-ing `.codegraph/` and running `codegraph init` again — the server kept reading the old, deleted database file and served a frozen snapshot: renamed or removed symbols still showed as live, new ones were missing, and `codegraph sync` couldn't refresh it — only restarting the server fixed it. The server now detects that the database file was swapped out from under it and reopens the live one in place, so results stay correct without a restart. (On Linux and macOS; Windows doesn't allow deleting an open file, so it isn't affected.) (#925)
- The MCP server now opens and auto-syncs a project that lives inside a folder an enclosing git repository ignores. Before, if the directory you indexed sat within a larger repo that gitignored it, the shared MCP daemon failed to open the project — its log repeated `Failed to open project … path should be a` `path.relative()` `d string, but got "./"` — so the file watcher never started and the index silently went stale until you ran `codegraph sync` by hand (setting `CODEGRAPH_NO_DAEMON=1` was the only workaround). The daemon now opens the project and starts watching as expected. Most visible with Codex on Windows, but the cause wasn't platform-specific. (#936)
- A git worktree of a submodule is no longer indexed as a duplicate copy of that submodule's code. CodeGraph already skips ordinary worktrees (a second working view of a repo it indexes), but a worktree created *from a submodule* — common in monorepos that check submodules out into worktrees for parallel feature work — was mistaken for a genuine embedded repo and swept in, duplicating every symbol it shared with the real submodule checkout (one report had ~28% of its index as duplicates, inflating both query results and the database). These submodule worktrees are now recognized and skipped, while the submodule's own checkout stays indexed as distinct code. Thanks @charlesxu2026-ship-it. (#945)
- A C++ class or struct annotated with an export/visibility macro — `class MYLIB_EXPORT Foo : public Bar { … }`, the common DLL-export / visibility pattern in headers — is no longer mis-indexed as a function spanning the whole class body. Not knowing the macro is a macro, the parser read it as a type and the whole declaration as a function, so the class surfaced as a phantom `function` that showed up as a false caller in `codegraph callers`, `codegraph impact`, and blast-radius analysis, and skewed symbol counts. CodeGraph now recognizes this misparse and drops the bogus node. Thanks @spwlyzx. (#946)
- `codegraph status` no longer reports phantom pending changes for files CodeGraph deliberately keeps out of the index — a tracked file inside a committed dependency dir (a checked-in `vendor/` or `node_modules/`), or a tracked file under a `.gitignore`d directory. A full index correctly excludes these, and `codegraph sync` never indexes them, but the fast change-detector behind `codegraph status` flagged every edit to such a file as "modified" (and a new one as "added") — so the pending-changes count never cleared no matter how many times you synced. Change detection now applies the same ignore rules the full index does, so `status` agrees with `sync`, and tools built on CodeGraph's change-detection API get the same corrected list. (#766)
- Files reached through a symlinked directory that points outside your project now index instead of being silently skipped. When a folder in your repo is a symlink to a location outside the repo root — the standard layout for Dota 2 custom games and similar SDK-linked projects, where `game/` and `content/` link into the engine tree — CodeGraph followed the symlink to discover those files but then blocked every one of them while reading, logging `Path traversal blocked in batch reader` and indexing nothing under them. The reader now agrees with the directory scan and indexes those files. The guard against `../` path escapes is unchanged, and the protection that keeps your agent from being served file contents from outside the repo is also unchanged — only the indexer's own read path follows these in-repo symlinks. (#935)

## [1.0.1] - 2026-06-13

### New Features

- New `codegraph daemon` command (alias `daemons`) — an interactive manager for the background daemons. It shows what's running (your current project's daemon first, pre-selected), and you arrow-key to one and press enter to stop it, or pick "Stop all". Previously the only way to shut a daemon down was to hunt for its pid and `kill` it by hand. (#845)
- Checking your installed version is now easy to reach however you guess at it: `codegraph version`, `codegraph -v`, and `codegraph -version` all print it, alongside the existing `codegraph --version`. (#864)
- The CodeGraph MCP server now self-heals if its main thread ever locks up. A lightweight watchdog notices when the process has stopped responding and stops it so a fresh one starts on your next request — it can no longer sit pinned at 100% CPU with no way to recover. Tune the detection window with `CODEGRAPH_WATCHDOG_TIMEOUT_MS`, or turn it off entirely with `CODEGRAPH_NO_WATCHDOG=1`. (#850)

### Fixes

- Git worktrees nested inside your project — like the `.claude/worktrees/` that Claude Code creates — are no longer indexed as duplicate copies of your whole codebase. CodeGraph deliberately indexes genuine embedded repos (a real second project checked out inside yours), but a worktree is just another working view of a repo it already indexed, so each one was multiplying every symbol — one report went from ~1,850 files to over 24,000, with search and `explore` flooded by stale duplicates. CodeGraph now recognizes worktrees and skips them, while still indexing real embedded repos and submodules. Thanks @tphakala. (#848)
- Running `codegraph serve --mcp` by hand no longer just hangs in silence. That command is the MCP server your AI agent starts for itself — not a step you run directly — and in a terminal it used to sit there waiting for input that never comes, looking broken. It now recognizes when a person runs it and explains what to do instead (`codegraph status`, `codegraph daemon`), and it's been dropped from the command listing so it stops looking like something you need to launch.
- Cross-file static method calls like `ClassName.staticMethod()` now resolve correctly. CodeGraph was linking the call to the *class* instead of the method (and recording it as a construction), so `callers` and `impact` for a static method came back empty — a real blind spot in TypeScript and JavaScript codebases that lean on static utility classes (Python and other languages with the same call shape benefit too). The call now links to the method itself. Thanks @contextFlow-lab. (#825)
- `codegraph affected` now accepts `./`-prefixed and absolute file paths, not just bare project-relative ones. Passing `./src/x.ts` or an absolute path — common when the file list comes from another tool — used to silently match nothing and report no affected tests. Thanks @contextFlow-lab. (#825)
- The CodeGraph MCP server no longer risks getting stuck at 100% CPU after an unexpected internal error. Previously such an error was logged but the process was left running in a broken state, where it could spin a CPU core indefinitely and had to be killed by hand. The server now logs the error and exits cleanly, so a fresh one starts on the next request. Thanks @songhlc. (#850)
- CodeGraph no longer indexes your entire home directory by accident. Running the installer — or `codegraph init` / `codegraph index` — from your home folder or a filesystem root would index everything underneath it (caches, `Library`, every other project), producing a multi-gigabyte index and constant file-watching churn. CodeGraph now refuses these roots and points you at a specific project instead; pass `--force` if you genuinely mean to. (Combined with the macOS file-descriptor fix already in 1.0.0, this closes the report of a runaway watcher exhausting the system file limit.) Thanks @ligson. (#845)

## [1.0.0] - 2026-06-12

### Security

- Closed a path-traversal hole where a symbolic link inside an indexed project that pointed *outside* the project root could make CodeGraph serve that out-of-root file's contents (for example a file under your home directory) to the AI agent. CodeGraph now resolves symlinks when validating file access and refuses to read anything whose real location is outside the project, while still allowing symlinks that stay within it. Thanks @sulthonzh. (#527)
- CodeGraph now indexes Spring configuration files (`application.properties` / `application.yml`) by key only, and never includes their values in `codegraph_explore` or `codegraph_node` output. Previously a secret committed to one of these files — a database password, API key, or connection string with embedded credentials — could be surfaced to an AI agent that asked about nearby code, even though the agent never opened the file. The configuration keys are still indexed, so reference and impact analysis are unaffected; an agent that genuinely needs a value reads the file itself. Shopify Liquid `{% schema %}` blocks are likewise indexed by name only. (#383)

### New Features

- **CodeGraph now indexes R** (`.R` / `.r`) — functions in every assignment form (`name <- function(...)`, `name = function(...)`, nested definitions), S4 / Reference / R6 classes with their methods, `setGeneric`/`setMethod` generics, top-level variables and constants, `library()` / `require()` imports, `source()` file references, and call edges — including calls inside tidyverse pipe chains. Statistical and research codebases get the full explore / impact / callers surface. (#828) (R)
- **Workspaces holding multiple git repositories now index as a whole.** Running `codegraph init` at the root of a directory that contains several independent git repos — including the common "super-repo" layout where the parent repo's `.gitignore` hides the child repos to keep `git status` quiet — now indexes every nested project into one graph, with each child repo's own `.gitignore` respected. `codegraph sync` and live file watching pick up changes inside the nested repos too (previously change detection only consulted the parent repo, so edits in child repos were invisible until a full re-index). Git repositories inside `node_modules` (npm git-dependencies) remain excluded. (#514)

- **`codegraph_explore` now explains where a flow ends instead of going silent.** When the symbols you ask about don't connect statically — because the code dispatches through a runtime mechanism like a computed call (`handlers[action.type](...)`), Python's `getattr`, a command/mediator bus (`sender.Send(new DeleteCommand(...))`), reflection, or `new Proxy` — explore now announces the exact dispatch site (file and line) where the static path stops, and when the dispatch key is visible in the source it shortlists the likely runtime targets (for example pointing a MediatR command straight at its `Handler.Handle` method). Detection is deterministic and runs only when a flow fails to connect; fully connected flows are unchanged, and nothing about indexing or the graph itself changes. Relatedly, a custom event bus whose emit and handler connect through a single synthesized hop now shows that hop explicitly (with the registration site) — it previously rendered nothing because the connection was "too short" for the flow section. (#687)

- **Anonymous usage telemetry, documented field-by-field and easy to turn off.** CodeGraph now collects a small set of anonymous usage statistics — which commands and MCP tools get used, which languages get indexed, which agents connect — so language and agent support work goes where real usage is. Never any code, file paths, file or symbol names, search queries, or IP addresses; usage aggregates locally into daily totals before anything is sent, and the ingest endpoint is public, auditable code in the repository that enforces the documented field list. The installer asks up front with a visible default-on toggle (and never re-asks); everywhere else a one-line notice prints before the first send. Disable any time with `codegraph telemetry off`, `CODEGRAPH_TELEMETRY=0`, or the cross-tool `DO_NOT_TRACK=1` standard — off means off: nothing is recorded, nothing is sent, and buffered data is deleted. `TELEMETRY.md` documents every field.
- **Subagents and non-MCP agents can now reach CodeGraph.** Two new CLI commands — `codegraph explore "<symbols or question>"` and `codegraph node <symbol-or-file>` — print exactly what the matching MCP tools return (relevant symbols' source + call paths; one symbol's source + callers; file reads with line numbers), so any agent with a shell can use the graph. And `codegraph install` now writes a small marker-fenced CodeGraph section into each agent's instructions file (`CLAUDE.md` / `AGENTS.md` / `GEMINI.md`) pointing at both surfaces — that file is what Task-tool subagents actually see, where the MCP server's own guidance only reaches the main agent. Measured on a delegated code-exploration task: subagents went from almost never using CodeGraph (~1 in 9 runs) to using it in every run, including runs with zero grep/file-reading fallback. The section is small, survives your own content, upgrades cleanly from the old long block, and `codegraph uninstall` removes it. Thanks @liuyao37511. (#704)
- **The MCP tool list is now a focused default of four** — `codegraph_explore`, `codegraph_node`, `codegraph_search`, and `codegraph_callers`. The other four (`codegraph_callees`, `codegraph_impact`, `codegraph_files`, `codegraph_status`) remain fully functional — the CLI and library API are unchanged, and `CODEGRAPH_MCP_TOOLS` re-enables any of them — but they're no longer listed to agents by default: measured agent behavior shows they're never or rarely picked, and the information they carry already arrives inline on the tools agents do use (explore's blast-radius section, node's dependents note, a symbol's own body as its callee list). A leaner list saves context tokens every session and steers agents to the right tool by presence alone.
- **CodeGraph now goes quiet instead of failing loudly in unindexed projects.** When an AI agent's session starts in a workspace that has no CodeGraph index, the MCP server now announces itself as inactive with a short note and lists no tools at all — instead of presenting the full toolset and erroring on every call, which taught agents to distrust CodeGraph even where it works. Querying another project that isn't indexed likewise returns clear guidance (use your regular tools for that codebase; the user can run `codegraph init` there to enable CodeGraph) instead of an error, and genuine internal errors now tell the agent to retry once rather than give up on CodeGraph entirely. Indexing stays your decision — agents are told not to run it themselves. (#769)
- **Astro projects are now indexed.** `.astro` files previously weren't parsed at all — on a typical Astro site that left most of the codebase invisible to search, impact, and `codegraph_explore`. CodeGraph now extracts the TypeScript frontmatter (functions, imports, `getStaticPaths`, …) and client-side `<script>` blocks, captures function calls and `<Component>` usages in template markup so cross-component dependencies trace end-to-end, resolves the `Astro` global and `astro:*` module imports as framework-provided, and maps `src/pages/` file-based routing to route nodes (`.astro` pages and `.ts` endpoints, including `[param]` and `[...rest]` dynamic segments, with underscore-prefixed files correctly excluded). Validated on two real-world Astro sites with 93% measured cross-file coverage and every page mapping to its route. Thanks @xingwangzhe. (#768) (Astro)
- Same-named symbols across a monorepo's apps are no longer conflated. In a NestJS-style workspace with one `UserService` per app, `codegraph_callers`, `codegraph_callees`, and `codegraph_impact` now report **one section per distinct definition** — each app's callers and blast radius under its own file-labeled heading — instead of a single merged list, and accept a `file` argument to focus exactly the definition you mean (like `codegraph_node` already did). Impact in particular no longer overstates a change's blast radius by merging unrelated same-named classes. Thanks @Igorgro. (#764)
- Fixed a related source of cross-package wrong edges: PascalCase **type references from plain `.ts` files were being resolved as React components**, which could link a file's own type alias to an arbitrary same-named class in another package (on one large monorepo this produced over a thousand wrong cross-package reference edges; 96% are now gone, and the remainder are genuine shared-model imports). Component resolution now applies only to references from JSX-capable files and never guesses between multiple candidates without a positional signal. The **Svelte and Vue component resolvers had the same arbitrary-pick flaw** (Vue resolved the first same-named `.vue` file found anywhere in the tree) and now follow the same rule: same-directory first, otherwise only an unambiguous name resolves. Re-index a project to benefit. (#764) (TypeScript, React, Svelte, Vue)
- TypeScript and JavaScript **class fields are now reported as properties instead of methods**. A plain field like `public fonts: Fonts;` previously extracted as a method, misrepresenting class shape and letting calls to same-named functions resolve to data fields (a boolean field named `isArray` was soaking up `Array.isArray(...)` call edges). Fields holding arrow functions or function expressions (`onClick = () => {…}`, including wrapped ones like `onScroll = throttle(() => {…})`) correctly remain methods and their bodies are still analyzed. Field initializers are analyzed too, so `history = createHistory()` records its call — and JavaScript class fields, which previously produced no symbol at all, now appear in the graph. Re-index a project to benefit. (#808) (TypeScript, JavaScript)
- Callback registration through `this` now resolves precisely in TypeScript and JavaScript: `window.addEventListener("online", this.onOfflineStatusToggle)` or an API object like `{ mutateElement: this.mutateElement }` produces a reference edge to the **enclosing class's own method** — never a same-named method on an unrelated class, and never a data field. Builds on the callback-registration support below. (#808) (TypeScript, JavaScript)
- Callback-registration coverage deepened across four more shapes: a `this.<member>` registration whose method lives on a **base class** now resolves through the inheritance chain (`bus.on("submit", this.handleSubmit)` in a subclass links to the parent's `handleSubmit`); Java and Kotlin **method references to other classes** (`Handlers::onMessage`, `OtherClass::handle`) resolve across files, with `this::` and `super::` scoped to the defining class and references through a variable deliberately left out; and Swift bare callback names now match only the **enclosing type's** methods (implicit `self`), eliminating a class of wrong edges where a parameter like `request` linked to a same-named method on an unrelated type. (Java, Kotlin, Swift, TypeScript, JavaScript)
- PHP **string and array callables** now register: a string passed to a callable-taking core function (`usort($items, 'cmp_items')`, `array_map('absint', …)`, `call_user_func`, `spl_autoload_register`, …) links to that function — including across files — and the array forms `[$this, 'method']` and `[Foo::class, 'method']` link to the named method (the `$this` form resolves through the class and its parents). Strings passed to arbitrary functions are deliberately ignored: only known callable positions are trusted. Validated on WordPress core (+556 edges, every sampled edge a genuine registration). (PHP)
- Ruby **lifecycle-hook symbols** now register: `before_action :authenticate`, `after_save :reindex`, `around_create`, `validate :check`, `rescue_from(…, with: :handler)` and friends link the symbol to the method it names — on the class itself or **inherited from a parent** (`before_action :authenticate` in a controller resolves to `ApplicationController`'s method). `validates` (plural) is excluded since its symbols name attributes, not methods. Validated on rails/rails (+385 edges, every sampled edge genuine). (Ruby)
- Method references to a type that needed **no import** now resolve: Java/Kotlin same-package references (`.concatMapMaybe(Maybe::just, …)`), **Kotlin companion-object members** (`KtHandlers::handle`), and cross-file C++ member pointers (`&TestSuite::RunSetUpTestSuite`). Resolution stays anchored to the named type, so a same-named member on a different class never matches. (Java, Kotlin, C++)
- CodeGraph now sees where a function is **registered as a callback**, not just where it's called. A function name passed as an argument (`signal(SIGINT, handler)`, `qsort(…, compare)`, `addEventListener(…, onBlur)`), assigned to a function pointer or field (`ops->recv_cb = my_cb`, `OnClick := Handler`), or placed in a struct initializer or handler table (`{ .recv_cb = my_cb }`, `{ "get", getCommand }`) now produces a reference edge from the registration site to the function — so `codegraph_callers` and `codegraph_impact` surface callback wiring that previously looked like dead code. Works across all supported languages, including the language-specific forms: C/C++ `&fn`, Java `Class::method`, Kotlin `::fn`, Swift `#selector`, Objective-C `@selector`, Ruby `method(:fn)`, Scala eta-expansion, and Delphi/Pascal `@Handler` and `OnClick := Handler` event wiring. Callers output labels these "via callback registration". Resolution is deliberately conservative: an ambiguous name produces no edge rather than a wrong one. Re-index a project to benefit. Thanks @zmcrazy. (#756)
- The `codegraph_node` MCP tool can now **read a whole source file like the built-in Read tool — only faster, served from the index**. Pass a file path with no symbol and it returns that file's current source with line numbers (the same `<n>⇥<line>` shape Read produces, so an assistant can edit straight from it), narrowable with `offset`/`limit` exactly like Read, plus a one-line note of which files depend on it (the file's blast radius). Use it anywhere you'd reach for Read on an indexed source file. Pass `symbolsOnly: true` for just the file's structure. Configuration/data files (`.yml` / `.properties`) are summarized by key only, never dumped, so secrets in them are never surfaced. The agent-facing guidance was also retuned so assistants reach for codegraph while *implementing* a change (not only when answering questions), since one codegraph call returns the same bytes plus the blast radius, faster than re-reading the file.
- New `codegraph upgrade` command updates CodeGraph to the latest release in place — it detects how you installed (the standalone `install.sh` / `install.ps1` bundle, npm, or npx) and does the right thing for each, on macOS, Linux, and Windows. Use `codegraph upgrade --check` to see whether an update is available without installing, or `codegraph upgrade <version>` to move to a specific version. After upgrading it reminds you to re-index your projects so they pick up the newer engine's improvements. (#679)
- `codegraph status` now flags when a project's index was built by an older engine than the one you're running and recommends re-indexing (also surfaced in `codegraph status --json`), so you know when a `codegraph index -f` or `codegraph sync` will add coverage a newer release introduced.
- Cross-file impact and blast-radius coverage now spans **all 22 supported languages and 14 web frameworks**, each validated on a real-world repo — see the new coverage table in the README. This release ships the cross-file resolution behind it, including Lua and Luau `require`, Shopify OS 2.0 Liquid section templates, Delphi form code-behind, Rust cross-module calls and Rocket route macros, Swift Fluent relationships, and the SvelteKit / Nuxt / Vapor / Axum route conventions. The residual everywhere is genuine static-analysis frontiers (runtime dispatch, reflection / DI, framework-convention entry points), never hidden.
- C# `record struct` and `readonly record struct` declarations now index with the correct `struct` kind, and positional one-liner records (`public record struct Money(decimal Amount);`) index reliably in every form — previously a bodiless value-type record could be skipped entirely and could halt extraction of the declarations following it in the same file. (#831) (C#)
- C# types are now tracked by their namespace-qualified name. Same-named types in different namespaces — a domain entity and a DTO both called `CatalogBrand`, say — are told apart instead of collapsing into one arbitrary match, so a reference resolves to the right one and impact no longer conflates them. (C#)
- ASP.NET Razor (`.cshtml`) and Blazor (`.razor`) markup are now parsed for code relationships. A `@model` / `@inherits` / `@inject` directive links the view to the C# view-model, base type, or service it names; a Blazor `<MyComponent/>` tag (plus `@typeof(...)` and generic `TItem="..."` arguments) links to the component class; and the C# inside `@code { }` / `@functions { }` / `@{ }` blocks is analyzed too, so services and types used in component logic are linked. A view-model, component, or service referenced only from markup is no longer reported as having no dependents, and editing it surfaces the views that use it. (ASP.NET, Blazor)
- A Razor/Blazor type reference now resolves through the component's `@using` namespaces — including the folder's cascading `_Imports.razor` — so a simple name that exists in several namespaces lands on the right one. A `@model` / `<MyComponent>` / `@code` reference to `CatalogBrand` resolves to the `@using`'d DTO (`BlazorShared.Models.CatalogBrand`) rather than a same-named domain entity. (ASP.NET, Blazor)
- `codegraph status --json` now also reports the running CLI `version`, the index directory (`indexPath`), and a `lastIndexed` timestamp (ISO-8601, or null when nothing's indexed yet), so CI and scripts can pin the CLI version and check index freshness from a single command. A matching `CodeGraph.getLastIndexedAt()` library method exposes the same freshness check without shelling out. Thanks @12122J and @eddieran. (#329)
- TypeScript service/RPC contracts defined as a tuple of generic types — `type MyServiceList = [Service<'query_apply_record', …>, Service<'apply_confirm', …>]` — now index each entry's string-literal name as a searchable symbol. Previously these names existed only as type arguments, so `codegraph query query_apply_record` found nothing even though the names are the app's primary API surface. The pattern is common in typed RPC / BFF clients and mock servers where the types are the source of truth for a runtime proxy object. Utility types (`Pick`, `Omit`, `Record`) and route paths are deliberately left out to avoid noise. Thanks @jiezhiyong. (#634) (TypeScript)
- New `CODEGRAPH_DIR` environment variable sets the per-project index directory name (default `.codegraph`). This lets one working tree hold two independent indexes — most usefully when you open the same checkout from both **Windows** and **WSL**, which can't safely share a single `.codegraph/`: the background-server lock and the SQLite database are tied to the OS that wrote them, and SQLite locking across the WSL2/Windows filesystem boundary is unreliable. Set `CODEGRAPH_DIR=.codegraph-win` on the Windows side, leave WSL on the default, and each keeps its own index in the same folder without clobbering the other. CodeGraph also skips any sibling `.codegraph-*` directory when indexing and watching, so neither environment trips over the other's data. Thanks @rrtt2323. (#636)

### Fixes

- **opencode on Windows now finds CodeGraph.** The installer wrote opencode's global MCP entry to `%APPDATA%\opencode\`, but opencode reads its config from `~/.config/opencode/` on every platform (honoring `XDG_CONFIG_HOME`), so the entry was invisible to it. Installs now write where opencode actually looks, and `codegraph install` / `codegraph uninstall` both clean a stale CodeGraph entry out of the old `%APPDATA%` location — other servers and comments in that file are left untouched. Thanks @fucknoobhanzo for the report and @WodenJay for the first patch. (#535)
- The `codegraph_search` tool's `kind: "type"` filter — a value its own schema advertises — silently matched nothing; it now correctly finds type aliases. The `codegraph_explore` tool's parameter guidance also no longer suggests running `codegraph_search` first, which contradicted explore's call-it-first design and cost agents an extra round-trip.
- Symbols defined in Svelte and Vue `<script>` blocks were reported one line below where they actually are — a function on line 3 was reported at line 4 — which offset every script-block symbol's location in search, `codegraph_node`, and explore output. Line numbers now match the file exactly. Re-index a project to benefit. (Svelte, Vue)
- Doc comments are now captured for exported, `const`-assigned, and decorated declarations, and the documentation a symbol carries is now clean across every supported language. Previously a comment above `export class X`, `export const fn = () => …`, a plain `const fn = () => …`, or a decorated Python `def`/`class` (`@app.route(...)`, `@dataclass`) was dropped entirely — only comments directly above a plain declaration were kept. CodeGraph now finds the comment through the `export` / `const` / decorator wrapper. Comment-marker cleanup was also rounded out for every language CodeGraph supports: Rust/Swift/Kotlin doc lines (`///`, `//!`), Python/Ruby/shell `#`, Lua/Luau (`--` and `--[[ ]]`), and Pascal (`{ }` and `(* *)`) no longer leave stray markers in the stored text — validated end-to-end across all 19 code languages plus Svelte/Vue `<script>` blocks. (#780). Thanks @caleb-kaiser.
- Go method calls made through a chained factory function now resolve to the correct type. A call like `New().Method()` used to drop the receiver, so the chained method attached to a same-named method on an unrelated type — or didn't resolve. CodeGraph now captures Go return types (a pointer `*Foo` resolves to `Foo`, and a multi-return `(*Foo, error)` to its first result), infers the chained receiver's type from what the factory function returns, and resolves the method on it — including methods promoted from an embedded struct — creating the edge only when the type or an embedded type genuinely has the method. Existing Go indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Go)
- Scala method calls made through a companion-object factory, a fluent chain, or a case-class `apply` now resolve to the correct type. A call like `Foo.create().bar()` or `Builder(cfg).bar()` used to drop the receiver, so the chained method silently attached to a same-named method on an unrelated type — most often mis-attributing a standard-library `Option` / `Iterator` `.map` / `.flatMap` / `.foreach` onto your own same-named class. CodeGraph now captures Scala return types (a generic `List[Foo]` resolves to its container `List`, a qualified `pkg.Foo` to `Foo`), infers the chained receiver's type from what the inner call returns or constructs, and resolves the method on it — including methods inherited from a trait the type extends — creating the edge only when that type or one of its traits genuinely has the method (so a wrong inference produces no edge instead of a misleading one). Existing Scala indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Scala)
- Rust method calls made through a chained associated function now resolve to the correct type. A call like `Foo::new().bar()` or `Foo::with(cfg).build()` used to drop the receiver, so the chained method silently attached to a same-named method on an unrelated type — or didn't resolve. CodeGraph now captures Rust return types (`-> Self` resolves to the implementing type), infers the chained receiver's type from what the associated function returns, and resolves the method on it — including methods provided by a trait the type implements (via the new `impl Trait for Type` relationships) — creating the edge only when the type or one of its traits genuinely has the method. Existing Rust indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Rust)
- Dart method calls made through a static factory, a factory or named constructor, or a fluent chain now resolve to the correct type. A call like `Foo.create().bar()` used to drop the receiver, so the chained method silently attached to a same-named method on an unrelated type — most often mis-attributing a standard-library `Option` / `Iterator` `.map` / `.where` onto your own same-named class. CodeGraph now indexes Dart **factory and named constructors** (`factory Foo.create()`, `Foo.named()`) as first-class members so calls to them resolve, captures Dart return types (a generic `List<Foo>` resolves to its container `List`), infers the chained receiver's type from what the inner call returns or constructs, and resolves the method on it — including methods inherited from a superclass or mixin — creating the edge only when that type genuinely has the method. Plain construction (`Foo(...)`) is still recorded as instantiation. Existing Dart indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Dart)
- Objective-C methods called through a chained message send now resolve to the correct class. A call like `[[Foo create] doIt]` used to drop the receiver, so `doIt` silently attached to a same-named method on an unrelated class — most often a test helper or stdlib class. CodeGraph now captures Objective-C method return types and infers the chained receiver's type from what the inner message returns. For the ubiquitous `[[X alloc] init]` and singleton (`[[X sharedInstance] …]`) patterns — where the factory returns `instancetype` — the receiver is the class `X` itself, so the chained method resolves on `X` (including methods inherited from a superclass), creating the edge only when the class genuinely has the method. Existing Objective-C indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Objective-C)
- Pascal/Delphi methods called through a chained factory call now resolve to the correct class. A call like `TFoo.GetInstance().DoIt()` used to drop the receiver, so `DoIt` silently attached to a same-named method on an unrelated class. CodeGraph now captures Pascal return types and infers the chained receiver's type from what the factory function returns — resolving to the declared type (including an interface return like `IFoo`), and for a constructor (`TFoo.Create().…`) or a typecast (`TFoo(x).…`) to the class `TFoo` itself, since both yield a `TFoo`. The edge is created only when that type genuinely has the method (so a wrong inference produces no edge). Existing Pascal/Delphi indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Pascal/Delphi)
- Pascal/Delphi **paren-less method calls are now tracked**. Pascal lets a no-argument method or procedure drop its parentheses (`Obj.Free;`, `List.Clear;`, `TFoo.GetInstance.DoIt;`), which previously weren't recorded as calls at all — so callers, impact, and trace missed them. CodeGraph now extracts these, scoped to statement position so a field or property access (which looks identical) is never mistaken for a call. On a real Delphi codebase this added ~1,100 previously-missing call edges with no false positives. Existing Pascal/Delphi indexes should be re-indexed (`codegraph index -f`) to benefit. (Pascal/Delphi)
- Pascal/Delphi calls inside a **standalone procedure or function** (one with no `interface` declaration, defined only in the `implementation` section) are now attributed to that routine instead of the whole file. Previously such a routine had no symbol of its own, so everything it called was lumped under the unit — `codegraph_callers` returned the file, and impact couldn't tell which routine was responsible. These routines are now indexed and their calls attributed correctly. Existing Pascal/Delphi indexes should be re-indexed (`codegraph index -f`) to benefit. (Pascal/Delphi)
- Chained method calls now resolve when the chained method is **inherited from a superclass or declared on an interface/protocol** the receiver's type conforms to — for example a call on a sealed-subclass instance (`Either.Right(x).combine(...)`) that invokes a method defined on its parent type. Previously these chains found no caller edge even though the factory's type was known, so the call was invisible to callers, impact, and trace. CodeGraph now walks the type's supertypes (its `extends` / `implements` relationships) to find the method, creating the edge only when a supertype genuinely declares it (so a wrong inference still produces no edge). This makes Java, Kotlin, and C# factory and fluent chains more complete. Existing indexes should be re-indexed (`codegraph index -f`) to benefit. (#750)
- Swift method calls made through a static factory, fluent chain, or constructor now resolve to the correct class. A call like `Foo.make().draw()` or `Foo().draw()` used to drop the receiver, so the chained method silently attached to a same-named method on an unrelated class — or didn't resolve at all. CodeGraph now captures Swift return types and infers the chained receiver's type from what the inner call returns (or the constructed type), creating the edge only when that class genuinely has the method (so a wrong inference produces no edge instead of a misleading one). Existing Swift indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Swift)
- C# method calls made through a static factory or fluent chain now resolve to the correct class. A call like `Foo.Create().Bar()` or `JObject.Parse(s).Property(...)` used to lose the receiver's type, so the chained method didn't resolve and the call was invisible to callers/impact/trace. CodeGraph now captures C# return types and infers the chained receiver's type from what the inner call returns, creating the edge only when that class genuinely has the method (so a wrong inference produces no edge). Existing C# indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (C#)
- Kotlin method calls made through a companion-object factory or fluent chain now resolve to the correct class. A call like `Foo.getInstance().bar()` or `Config.create(opts).build()` used to drop the receiver entirely, so the chained method silently attached to a same-named method on an unrelated class — or didn't resolve at all — corrupting callers, impact, and trace. CodeGraph now captures Kotlin return types and infers the chained receiver's type from what the inner call returns, creating the edge only when that class genuinely has the method (so a wrong inference produces no edge instead of a misleading one). Existing Kotlin indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Kotlin)
- Java method calls made through a static factory or fluent chain now resolve to the correct class. A call like `Foo.getInstance().bar()` or `Config.create(opts).build()` used to lose the receiver's type, so when two classes had a same-named method the call silently attached to whichever was indexed first — or didn't resolve at all — corrupting callers, impact, and trace. CodeGraph now captures Java return types and infers the chained receiver's type from what the inner call returns, creating the edge only when that class genuinely has the method (so a wrong inference produces no edge instead of a misleading one). Covers factories and fluent builders that take arguments (`hashKeys().arrayListValues()`), including builders that return a nested type. Existing Java indexes should be re-indexed (`codegraph index -f`) to benefit. (#750) (Java)
- PHP: a method called through a chained static factory — `Cls::for($x)->method(...)`, the canonical Laravel per-credential / per-tenant client idiom — now records a caller edge. Previously the receiver type (what `for()` returns) was never recovered, so `codegraph_callers` returned nothing for the method and the call was invisible to `codegraph_impact`. CodeGraph now captures PHP return types — `: self` / `: static` resolve to the declaring class, `: SomeClass` to that class — and resolves the chained method on the factory's result, creating the edge only when that class actually has the method (so a wrong inference produces no edge). Existing PHP indexes should be re-indexed (`codegraph index -f`) to benefit. Thanks @cvanderlinden. (#608) (PHP)
- Search relevance: including the project name in a query (a user naturally writes `MyApp backend routes`) no longer buries the part of the codebase the query is actually about. The project name lexically matches whatever stack embeds it — a `MyAppFrontend/` directory, a `MyAppApp` class — and it was over-weighted two ways: a single PascalCase word was scored once per sub-token (`my` / `app` / `myapp`), so one concept boosted that path several times over; and the name carried full path / disambiguation weight even though it names the whole repo, not any symbol. Now path relevance counts each query word once, and a word matching the project name (derived from `go.mod`, `package.json`, or the repo directory) is dropped from path scoring and from `codegraph_explore`'s type-disambiguation bias — unless it's the only term, so a bare project-name search still works. In a mixed-stack repo, a backend question now surfaces the backend even with the project name in the query. Thanks @MiNuo1. (#720)
- Go: a function called only from inside an anonymous closure — a cobra `RunE: func(…) {…}` handler, a goroutine literal, or a callback closure stored in a package-level `var` — now shows its real caller. Previously the call leaked to the file node, so `codegraph_callers` and `codegraph_impact` reported such a function as having no meaningful caller; the call is now attributed to the enclosing declaration, so editing the function surfaces the closures that use it. Existing Go indexes should be re-indexed (`codegraph index -f`) to benefit. Thanks @Cyclone1070. (#693) (Go)
- Indexing no longer aborts when a `.gitignore` contains non-UTF-8 bytes or an unparseable pattern. A `.gitignore` transparently encrypted in place by corporate DLP / endpoint-security software (a common enterprise scenario) — or one with a stray pattern the matcher can't compile (`\[`, producing "Unterminated character class") — used to crash the entire `sync` / `index` with a screen of garbled bytes and never name the offending file, leaving `Files: 0 / Nodes: 0`. CodeGraph now skips a `.gitignore` that isn't valid UTF-8 text whole, drops only the individual unparseable patterns from a text one, and logs a warning naming the file — indexing continues either way. Thanks @zhanghang-9527. (#682)
- C++ method calls made through a singleton, factory, or chained getter now resolve to the correct class. A call like `Foo::instance().bar()`, `WidgetFactory::create().draw()`, `openSession()->run()`, or the same stored in an `auto` local first, used to lose the receiver's type — so when two classes had a same-named method the call silently attached to whichever was indexed first (or didn't resolve at all), corrupting callers, impact, and trace. CodeGraph now infers the receiver's type from what the inner call returns (capturing C++ return types for the first time) and creates the edge only when that class genuinely has the method, so a wrong guess produces no edge instead of a misleading one. Covers singletons and self-returning accessors, factories that return a different type, free-function factories, `make_unique` / `make_shared` / `new` / direct construction, and single-level member chains. Existing C/C++ indexes should be re-indexed (`codegraph index -f`) to benefit. Thanks @stabey. (#645) (C/C++)
- The shared background server no longer logs a scary-looking `[error] … undefined` line on every session start. Attaching to the shared daemon is normal, healthy behavior, but the informational message was being surfaced by MCP hosts (Claude Code and others) as an error; it's now silent by default — set `CODEGRAPH_MCP_LOG_ATTACH=1` to surface it when debugging daemon attach. Thanks @mturac. (#618)
- On Windows, CodeGraph's background processes no longer pile up without bound and saturate CPU over a long session. When the editor or agent that launched CodeGraph exited, its helper process couldn't tell its parent had gone — Windows reports process lineage differently than macOS and Linux — so the helper kept running, the shared background server never saw the client disconnect, and its idle timer never fired to shut it down. CodeGraph now detects parent-process exit directly on Windows, so helpers and the idle background server wind down promptly, the same as they already did on macOS and Linux. (#692, #576, #680)
- The MCP server now shuts down cleanly when its editor/agent connection drops abruptly, instead of risking an orphaned process that pins a CPU core. Editors talk to a stdio MCP server over a socket; if that socket failed with an error rather than closing cleanly — which can happen when the editor window is reloaded or the launching process is killed — the server didn't treat it as a disconnect and could be left running. CodeGraph now treats any failure of its input stream as a shutdown signal and tears the stream down, so an orphaned server exits promptly. (#799)
- The shared background server has two further safeguards against ever lingering: it now drops a client the moment it detects that client's process is gone (even if the disconnect arrived uncleanly — a force-quit or a dropped connection that never closed the socket), and it won't stay running indefinitely with clients attached but no activity. Together these guarantee it always winds down, on every platform. (#692)
- A session no longer loses CodeGraph when the shared background server is restarted out from under it — for example when your MCP host (opencode and others) stops and restarts the server as you open another session. Previously the affected session's connection died silently and any request in flight at that moment hung; now CodeGraph keeps that session working by serving it locally, so the tools stay available without restarting the session. (#662)
- React Native native→JS events now connect through the common `sendEvent(context, "X", body)` wrapper. Many libraries (react-native-device-info and others) wrap the event emitter behind a helper whose `.emit(eventName, …)` takes a *variable*, so the matcher — which looked for `.emit("literal", …)` — missed it; the literal event name actually lives in the wrapper call. Now a native method that fires `sendEvent(…, "batteryLevelChanged", …)` links to the JS `addListener('batteryLevelChanged', …)` handler, so editing the native emitter surfaces the JS subscriber. (React Native)
- React Native / Expo cross-language bridges are more complete and more precise. An Expo Module method declared with a generic type — Android's `AsyncFunction<Float>("getBatteryLevelAsync")` — is now indexed (the `<Float>` used to defeat the matcher, so every Android Expo method was dropped and a JS call resolved only to the iOS Swift impl). The iOS and Android implementations of the same JS-visible method — both Expo Modules and classic NativeModules (`@ReactMethod` on Android, the matching method on iOS) — are now linked to each other, so a JS call that resolves to one platform still reaches the other and editing either platform's native code surfaces the JS caller. And a `Type.member` static read in native code (e.g. Android's `BatteryManager.EXTRA_LEVEL`) no longer falsely links to a coincidentally same-named class in another language (a web `BatteryManager`) — type references stay within a language family, while genuine cross-language bridges (config→code, JS↔native calls) are unaffected. (React Native, Expo)
- A TypeScript/JavaScript reference or import no longer gets mis-linked to a same-named class in a native language. In a React Native / Expo repo that has both a TypeScript `TestRunner` type and a Kotlin `TestRunner` class, a TS reference to `TestRunner` — or an `import React` sitting next to a Swift `React` — used to resolve onto the native symbol (the component resolver matched any same-named class regardless of language, and import statements weren't language-checked at all). References and imports now stay within their language family, so they land on the right symbol while genuine cross-language bridges (JS↔native calls, config→code) are untouched. A C/C++ `#include "Foo.h"` likewise no longer resolves to a same-named header from another platform (an iOS Objective-C `Foo.h`). (React Native, Expo, TypeScript, C/C++)
- Native includes and Kotlin Multiplatform imports now resolve to the correct file in multi-platform projects. A C/C++ `#include "Foo.h"` now resolves to the header in the including file's own directory first (the C quoted-include rule), so when a module ships a same-named header per platform (a Windows, an Apple, and an Android `Foo.h` side by side) the local one correctly shows its dependents instead of an arbitrary other-platform header looking like the dependency. And a Kotlin Multiplatform `expect` declaration is no longer reported as having no dependents: a `commonMain` import now resolves to the `commonMain` `expect` (matched within the importing source set) rather than being absorbed by one platform's `actual`. (C/C++, Kotlin)
- `codegraph affected` now reports the tests and files that actually depend on your changes. It used to follow only `import` statements — but those never cross file boundaries in CodeGraph's graph — so it returned **no affected tests for any change, in every language**. It now traces the real cross-file usage graph (calls, references, instantiations, and class `extends` / `implements`), so `git diff --name-only | codegraph affected` surfaces the test files that exercise the changed code. Circular-dependency detection, which had the same blind spot, now works too.
- Blast radius, callers, and `codegraph affected` now recognize far more of the dependencies that were already in your code. A symbol now counts as a dependency whether it's called, used only in a type annotation inside a function body (`const items: Foo[] = []`), imported and placed in a registry array or passed as an argument, used as a JSX component, simply re-exported from a barrel (`export { X } from './x'`), or pulled in as a namespace (`import * as ns from '@/x'`) — including through tsconfig path aliases like `@/`. Previously only called, instantiated, or signature-typed symbols created a cross-file link, so a file that used a dependency in any other way could look like it depended on nothing — and the file that defined a widely-used symbol could look like nothing depended on it. The graph still indexes exactly the same symbols; it just connects the ones that were already there. (TypeScript/JavaScript)
- The same completeness fix now applies to **Python**: a name brought in with `from module import X` is recorded as a dependency on that module even when `X` is only stored in a list/dict, passed as an argument, used as a decorator, or re-exported through an `__init__.py`. Previously Python linked only imports that were called or instantiated, so a module consumed purely by value — or only re-exported — looked like nothing depended on it.
- `codegraph_callers` (and the `callers` command) now lists the places a class is **instantiated**, not just where it's imported. Constructing a class — `Foo(...)` in Python, `new Foo()` elsewhere — is calling its constructor, so asking who calls a class now returns the construction sites, and `codegraph_callees` / trace cross the instantiation the same way. Previously a class's instantiation sites were invisible to `callers`, so "what breaks if I change this class?" could come back empty even when the constructor was called from many places. Works on your existing index — no re-index needed. (#774)
- Rust impact and `codegraph affected` now connect far more of the module graph. Struct literals (`Widget { n: 1 }`) are recorded as instantiations; a `use` / `pub use` brings its item into the dependency graph — so a `pub use` re-export hub (a `mod.rs` re-exporting its submodules) depends on the modules it re-exports — resolved by Rust module path (`crate::`/`self::`/`super::`), so a re-export of a common name like `read` links to the right module instead of a same-named symbol elsewhere; and trait dispatch reaches implementations — a struct whose methods cover a trait's is treated as implementing it, and a call through `&dyn Trait` resolves to the concrete method. Previously a Rust type linked only when called or used in a type position, so structs built by literal, modules surfaced only through `pub use`, and trait-only implementations looked like they had no dependents. (#584 for Rust traits)
- Rust cross-module function calls now resolve to the right file. A call to a sibling submodule's function — `users::router()`, the common router-assembly / handler-registration pattern where `mod users;` makes `users` a child of the current module — is now resolved relative to the current module, not only the crate root. Deeper module-path calls (`database::profiles::find()` — the `db.run(|c| …)` data-access shape) now resolve too; these were being discarded before resolution even ran, because the path's leaf function name was never checked. Previously such a call linked to nothing, so a module reached only as `module::path::function()` looked like it had no dependents; a web app wired this way (Axum, Rocket, and similar) now surfaces its handler and data-access modules' real callers. (Rust)
- Rocket route handlers now connect to where they're mounted. A handler registered in a `routes![a::b::handler, …]` or `catchers![…]` macro used to be invisible — the macro body is a raw token tree, so the handler looked like it had no caller (Rocket mounts it at runtime) and its file showed no dependents. The handler paths are now read out of the macro and linked to the `mount`/`register` call, so editing a Rocket handler surfaces its route registration and a routes module is no longer reported as unused. (Rust, Rocket)
- SvelteKit pages now connect to their server `load` functions. SvelteKit wires a `+page.server.js` / `+page.js` `load` (and form `actions`) to the sibling `+page.svelte`'s `data` by file path, with no import between them — so editing a `load` previously showed no impact on the page it feeds. Each page is now linked to the `load`/`actions` in its own route directory (and likewise for `+layout`), so editing a loader surfaces the page that renders its data, and tracing a page reaches its server-side data source. (Svelte, SvelteKit)
- Nuxt nested components are now connected to where they're used. Nuxt auto-imports a component in a subdirectory by a directory-prefixed name — `components/media/Card.vue` is used in templates as `<MediaCard/>` — but it was tracked by its file name (`Card`), so the usage didn't resolve and the component looked unused. PascalCase component tags (`<MediaCard>`, `<NavBar>`) in a `.vue` template are now matched, falling back to the Nuxt directory-prefixed name, so editing a nested component surfaces every page and component that renders it. (Vue, Nuxt)
- Lua and Luau `require` calls now connect to their module files. A dotted module path (`require("telescope.config")` → `telescope/config.lua` or `.../config/init.lua`) and a Roblox/Luau instance-path require (`require(script.Parent.Signal)` → the `Signal` module) now link to the file they load, so editing a module surfaces every file that requires it. Previously requires resolved to nothing, so a Lua/Luau module looked like it had no dependents. (Lua, Luau)
- Shopify OS 2.0 sections now connect to the JSON templates that use them. Modern Shopify themes define templates as JSON (`templates/*.json`, plus section groups `sections/*.json`) that list sections by `type` rather than with a `{% section %}` Liquid tag, so a section used only from a JSON template was reported as having no dependents. Those JSON files are now read and each section `type` is linked to its `sections/<type>.liquid`, so editing a section surfaces the templates that render it. (Liquid, Shopify)
- Delphi form definitions now connect to their code-behind units. A `.dfm` (VCL) or `.fmx` (FireMonkey) form is owned by its same-named `.pas` unit through the `{$R *.dfm}` directive rather than a `uses` clause, so a form file used only as a definition was reported as having no dependents. The unit is now linked to its form, so editing a form surfaces the unit that owns it. (Pascal/Delphi)
- Swift property wrappers and attributes are now connected. A `@Argument` / `@Published` / `@State` / custom `@propertyWrapper` on a property — and attributes on types, methods, and functions (`@objc`, `@MainActor`, …) — now record a dependency on the wrapper/attribute type. Previously these were dropped entirely (Swift attributes parse differently from other languages, and stored properties weren't being inspected), so the wrapper type looked unused and the file using it depended on nothing — a big gap for SwiftUI and argument-parser-style code.
- Swift Fluent relationship models are no longer orphaned. A type referenced only through a property-wrapper *argument* — `@Siblings(through: AcronymCategoryPivot.self, …)`, the many-to-many pivot/join model — now records a dependency on that type. Previously only the wrapper itself (`Siblings`) and the property's declared type were captured, so a pivot model reached solely through the relationship looked like nothing depended on it and editing it surfaced no impact. (Swift, Vapor/Fluent)
- Java annotations are now connected. Annotation definitions (`@interface Foo`) are indexed as types, and every `@Foo` usage on a class, method, or field is recorded as a dependency on it. Previously neither side was captured — annotation usages were dropped (they live inside the declaration's modifiers) and `@interface` types weren't indexed at all — so annotation-driven code (Spring `@GetMapping`, JPA `@Entity`, Gson `@SerializedName`, …) showed the annotation as having no users and the annotated class as not depending on it.
- Kotlin Multiplatform `expect`/`actual` declarations are now connected. A platform implementation — `actual fun`, `actual class`, or an `actual typealias` in a `jvm` / `native` / `js` / `wasm` source set — is linked to the common `expect` declaration it fulfills (including the common case of an `expect class` fulfilled by an `actual typealias`). Previously a caller in common code resolved to the `expect` declaration, so every platform `actual` looked like it had no dependents and editing one showed an empty blast radius; now changing a platform implementation surfaces the common API and everything that uses it. (Kotlin)
- Scala impact and `codegraph affected` now connect the type graph that typeclass-style code is built on. A parameterized supertype (`trait Monoid[A] extends Semigroup[A] with Serializable`) now links to each parent; a type used in a `val`/`def` signature, as a type argument, or as a context bound (`def f[A: Monoid]`) — including the trailing implicit parameter list (`(implicit M: Monoid[A])`) where typeclass instances are passed — now records a dependency; and `new T[...] { … }` counts as an instantiation. Previously Scala linked only plain calls and bare, non-generic supertypes, so a trait extended with type parameters, used as a type, or required as an implicit looked like nothing depended on it — which on a typeclass-heavy codebase (cats, algebra) was most of the graph. (Scala)
- PHP impact and `codegraph affected` now understand namespaces and `use` imports. Classes are tracked by their namespaced name, so the many same-named classes a framework defines (Laravel has 7+ `Factory` interfaces, several `Dispatcher`s, across namespaces) are told apart instead of collapsing into one arbitrary match. A `use App\Contracts\Cache\Factory;` now records a dependency on exactly that class — so a contract or interface that's imported and constructor-injected (the dependency-injection pattern) is no longer reported as having no dependents — and parameter, property, and return type-hints are recorded too. Previously PHP ignored namespaces entirely and linked only calls, `new`, and inheritance. (PHP)
- Objective-C impact and `codegraph affected` are dramatically more complete. Four gaps are fixed: a single-argument message (`[cache storeImage:key]` — the most common call form) now matches its `storeImage:` method instead of dropping the colon; a class-message receiver (`[SDImageCache sharedCache]`, `[[Foo alloc] init]`) now records a dependency on the class, whose `@interface` lives in the header; `#import "Foo.h"` now resolves to the header file, so a header is no longer reported as having no dependents; and class-method message calls now resolve through the receiver type. Together these took typical libraries from a third-to-half of their files showing real dependents to ~90%. (Objective-C)
- A type referenced only through a static member or enum value now records a dependency. Reading an enum value (`MediaKind.video`), a static constant (`Colors.red`, `JsonScope.EMPTY_DOCUMENT`), or a class constant (`Foo::BAR`) now links to the type — previously only method calls and `new` did, so a type or enum used purely *by value* (enum-heavy APIs, constants classes — a very common pattern) looked like nothing depended on it. Applies to Java, C#, Kotlin, Swift, Scala, Dart, PHP, and C++.
- Dart impact and `codegraph affected` now follow mixins and method type annotations. A `with` mixin — Dart's core composition mechanism, which Flutter is built on — now records a dependency, so editing a mixin surfaces every class that mixes it in (the whole `with` clause used to be dropped, and a class declared `with M` alone even lost its real superclass link). And types used in a method's parameters or return value now link to their definition, so a class or enum referenced only as a type — not constructed or called — is no longer reported as having no dependents. (Dart)
- C++ free functions are now indexed under their real name. A function written with a qualified-type parameter (`std::string TableFileName(const std::string& dbname)`) or an `auto … -> std::string` trailing return type was mistakenly named after that type (`string`), so calls to it never resolved, `codegraph_node` couldn't find it by name, and the file defining it looked like nothing depended on it. The function now keeps its real name, so cross-file calls, callers, and blast radius work — a meaningful gain for any namespaced C++ codebase (this is how most free functions in a library look). (C++)
- Ruby impact and `codegraph affected` now follow mixins and `require`s. `include`, `extend`, and `prepend` of a module — Ruby's primary composition mechanism (ActiveSupport concerns, `Comparable`, `Enumerable`) — now record a dependency on that module, so editing a concern surfaces every class that mixes it in; previously these were read as a call to a method named `include`, so a module whose methods are exercised only by application code looked like nothing depended on it. And `require "lib/foo"` / `require_relative "../foo"` now link to the required file, so a file pulled in only by a `require` (config-loaded components, gems that don't autoload) is no longer reported as having no dependents. Together these took a typical gem from ~71% of its files showing real dependents to ~100%. (Ruby)
- C# `record` types are now indexed. `record`, `record class`, and `record struct` declarations (everywhere in modern C# — DTOs, value objects, CQRS messages, MediatR notifications) were previously skipped entirely, so every reference, generic type argument (`IEnumerable<MyRecord>`), and `new MyRecord(...)` pointed at nothing and the file defining a record looked like it had no callers or dependents. (#237)
- C# is now parsed with an up-to-date grammar that understands C# 12 **primary constructors**. A class or struct written as `class OrderService(IRepo repo, [FromKeyedServices("primary")] ICache cache) { … }` is now indexed reliably — previously the constructor parameter list confused the parser and could drop the whole class (and all of its methods) from the index, most often exactly when a parameter carried an attribute, as in the ASP.NET keyed-dependency-injection pattern. The primary-constructor parameters are also recorded as dependencies, so the services a type is constructed with show up in its blast radius and "who depends on this contract" answers. Method return types, base types, and members all continue to resolve, and `#if`-guarded members in multi-targeting code keep parsing correctly. (#237)
- Go interfaces now connect to their implementations. Go has no `implements` keyword — a type satisfies an interface just by having the right methods — so CodeGraph now infers that link: a struct whose methods cover an interface's method set is treated as implementing it, and a call through the interface (`API.Marshal(...)`) reaches every concrete implementation. This means a type used only via an interface (the common plugin/strategy pattern — e.g. JSON-codec or renderer implementations selected at runtime) is no longer reported as having no callers or no dependents, and impact now flows from an interface method to the implementations behind it. (#584)
- Go now records cross-package struct creation. A composite literal like `render.XML{...}` or `pkga.Widget{...}` — including ones registered in a package-level `var registry = map[string]R{...}` — now links to the package that defines the type. Cross-package function calls and type references already resolved; this closes struct instantiation, so a package whose types are only *constructed* elsewhere (a common pattern for interface implementations) is no longer reported as having no dependents. Go type conversions such as `(*Wrapped)(x)` now link to the converted-to type as well.
- Python now follows whole-module imports — `from . import certs` then `certs.where()`, or `from pkg import sub` then `sub.run()`. Calls and attribute access through an imported submodule now resolve to that submodule, and importing a module is recorded as a dependency on it even when the member you use is itself re-exported from a third-party package. This also fixed Python relative-import path resolution generally (`from .sub.mod import x`), so `codegraph affected` and impact see the real module graph of a package.
- Python now also links a whole-module **absolute** import (`import conduit.apps.signals`) to that module's file, not just `from x import y`. A module imported by its dotted path — common in package setup and side-effect imports — is no longer reported as having no dependents. Standard-library imports (`import os`) correctly create no edge. (Python)
- Python `from package import submodule` now links to that submodule's file, resolved through the import's package so it lands on the right one when same-named modules exist in sibling packages (the FastAPI / Django router-aggregator pattern: `from app.api.routes import authentication` with an unrelated `authentication.py` elsewhere). So a route/handler module pulled in only by an aggregator is no longer reported as having no dependents. (Python)
- Python now records the actual call edge for a function invoked through an imported module — `from pkg import module` (or `import pkg.module`) followed by `module.func()`, a common testing and namespacing pattern. Previously only the module-level dependency was tracked, so `codegraph_callers`, `codegraph_callees`, and impact on the target function came back empty even though the import itself resolved. (#578) (Python)
- Django `include('app.urls')` now records a dependency from the project URLconf onto the included app's `urls.py`, so an app's routes module is no longer reported as having no dependents and editing it surfaces the project that mounts it. (Django)
- A chained method call (`builder.Services.AddCoreServices(...)`) now resolves to its definition. Previously only a single-segment receiver (`obj.method()`) resolved, so a call through a property/member chain — very common for C# extension methods like ASP.NET dependency-injection registration (`AddCoreServices`/`AddWebServices`) and Guard clauses — found no definition. (C#, and any language with chained calls)
- A renamed default import (`import articlesController from './article.controller'` where the module does `export default router`) now records a dependency on the imported module. Previously only named imports linked, so a module consumed only through a default import — the standard Express/NestJS route-controller pattern — looked like nothing depended on it. External packages (`import React from 'react'`) still create no edge. (TypeScript/JavaScript)
- The background file watcher no longer exhausts your machine's file-descriptor budget. On macOS it previously kept **one open file handle per watched file**, so on a large project the running MCP server could pile up tens of thousands of handles and blow past the system-wide limit — at which point *unrelated* apps (your shell, editor, Docker, browser) started failing with "too many open files" until the codegraph process was killed. The watcher now uses a single recursive watch on macOS and Windows, and bounded per-directory watches on Linux, so its cost stays flat no matter how large the project is. (#644, #496, #555, #628, #579)
- Indexing a project with very symbol-dense files (tens of thousands of functions or methods in a single file) no longer runs out of memory. The step that links dynamic call relationships used to load every function and method into memory at once, which could exhaust the heap and abort indexing with "JavaScript heap out of memory" on large or generated codebases; it now streams them, so memory stays flat no matter how many symbols the project has. (#610)
- Indexing a very large repository no longer aborts during its first sync with a "too many SQL variables" error. (#540)
- Files under directories with non-ASCII names (for example CJK characters) are no longer silently skipped during indexing. (#541)
- The `.codegraph/` index folder no longer clutters `git status`: its generated ignore file now excludes everything in the folder except itself, so the database, `daemon.pid`, sockets, and logs stop showing up as untracked changes. (#492, #484)
- Projects initialized by an older version now get that fix automatically: a `.codegraph/.gitignore` written before this change — which listed only the database, cache, and logs and so let the daemon's `daemon.pid` get committed — is upgraded in place the next time you run any CodeGraph command. A `.gitignore` you've customized yourself is left untouched. (#788)
- SAP HANA `.xsjs` / `.xsjslib` files are now indexed as JavaScript. (#556)
- TypeScript `.mts` and `.cts` module files are now indexed instead of being skipped. (#366)
- JavaScript modules that wrap their code in an anonymous function — AMD/RequireJS, NetSuite SuiteScript, IIFE bundles — now have their inner functions and calls indexed, instead of the file coming up nearly empty. (#528)
- Go methods declared on generic types (e.g. `func (s *Stack[T]) Push(...)`) are now correctly attached to their type, so callers, callees, and impact include them. (#583)
- Go methods now attach to their receiver type even when declared in a different file from the `type` itself — the idiomatic split where `type User struct{…}` lives in one file and `func (u *User) Save()` in another. Previously a cross-file method was orphaned from its struct, so the type's member list, callers/callees, and impact missed it; as a knock-on, a struct whose interface-satisfying methods are spread across files now also links to the interfaces it implements. (#583)
- Asking what a symbol impacts no longer drags in every unrelated sibling method of its class — impact now follows real dependencies instead of the structural "contains" relationship, keeping the result focused on what actually depends on the symbol. (#536)
- CodeGraph's MCP server now answers an agent's `resources/list` and `prompts/list` probes with an empty list instead of an error, clearing the `-32601` messages some clients (opencode, Codex) logged on connect. (#621)
- Svelte and Vue components used through a barrel file — `export { default as Button } from './Button.svelte'` re-exported from an `index.ts` and imported elsewhere — are no longer falsely reported as having **0 callers**. CodeGraph now follows the default re-export all the way to the component and resolves the imports that `.svelte` / `.vue` files themselves use, so `codegraph_callers` and `codegraph_impact` see every place a component is used. This also covers components imported from another package in a workspace/monorepo (`@scope/ui/widgets`) and bare directory imports (`import { x } from './'`). Previously a live component consumed only through a barrel looked like dead code. Thanks @nakisen. (#629)
- Components used in a Vue Single-File Component's `<template>` — `<MyButton />`, or the kebab-case `<my-button />` — are now indexed as usages, so `codegraph_callers` and `codegraph_impact` include components that appear only in another component's markup (including through a barrel re-export). Previously only a Vue component's `<script>` block was analyzed, so template-only usages were invisible. (#629)
- PHP: `include` / `require` / `include_once` / `require_once` of a static path now create a file→file dependency edge, so `codegraph_callers` and `codegraph_impact` follow includes in procedural / script-style PHP codebases — previously only namespace `use` statements became dependency edges. Dynamic includes (`include $var`, `require __DIR__ . '/x'`) are skipped. Thanks @atahan150 (#660).

## [0.9.9] - 2026-06-02

### New Features

- `codegraph_explore` is now the primary tool, and one call is usually all an agent needs: it returns the verbatim source of the symbols relevant to your question (a plain question works as the query — you no longer need exact symbol names), grouped by file and Read-equivalent, so the agent answers without falling back to read/grep. The narrower `codegraph_context` and `codegraph_trace` tools were removed in favor of it — explore already surfaces the call flow among the symbols you name (the job trace did), so there's one obvious tool to reach for instead of three.
- `codegraph_explore` now includes a compact "Blast radius" for the symbols you're looking at — who depends on each (just the locations, not their source) and which test files cover it — so before editing, the agent can see what else to update and which tests to run, without a separate impact lookup. Symbols nothing depends on are skipped, so it stays short.
- Functions defined inside a store or handler object — the actions in a Zustand `create((set, get) => ({ … }))` store, and the same shape in Redux, Pinia, MobX, or any exported handler/route map — are now indexed as real symbols. Previously they existed only as object properties, so looking one up by name or asking who calls it returned "not found" and the agent had to read the whole store file to follow the flow; now `codegraph_node`, `codegraph_callers`, and `codegraph_explore` resolve them directly — including calls made through `useStore.getState().fetchUser()` or a destructured `const { fetchUser } = useStore.getState()`.
- `codegraph_explore` now surfaces the *right* definition when a method name is overloaded across types. Asking about, say, `DataRequest`'s `task` and `validate` used to return a same-named method from an unrelated file (or an abstract base stub) and bury the one you meant; explore now recognizes the type you named in the query and leads with that type's own overloads, in full.

### Fixes

- Search ranking no longer lets a common word in your request hijack the results: asking about, say, a "flat object" screen used to surface an unrelated constant that merely happened to be named the same, because the exact-name match outweighed everything else. Ranking now weighs how well each result is corroborated by the rest of your request, so the symbols you actually meant come first (this improves `codegraph_explore`'s results).
- `codegraph_node` now returns *every* definition when a name is ambiguous — an overloaded method, or the same method name on different types — instead of returning one (sometimes the wrong one) with a note listing the rest. Asking for such a symbol now hands back all of the matching definitions with their source in a single call, so the agent stops having to read the file by hand to find the specific overload it wanted (common in Swift, Go, Java, and C#). For a heavily-overloaded name (a `poll`/`validate` with dozens of definitions), pass `file` (and/or `line`) — e.g. the `file:line` shown in a trail — to get that exact definition's body. Large overload sets show the most relevant ones in full and list the remainder by location.
- `codegraph_explore` never returns half a method anymore: when output runs up against its size budget it drops whole methods or whole files (and lists what it dropped, so you can ask for them in another call) instead of cutting off a method body partway. A truncated method was the one case that still sent the agent to read the file for the rest — so the source explore returns is now always complete and usable as-is.

## [0.9.8] - 2026-06-01

### New Features

- `codegraph init` now builds the initial index by default — you no longer need the `-i`/`--index` flag (it's still accepted, so existing commands and scripts keep working). (#483)
- Go: Gin middleware chains now connect end-to-end in `codegraph_trace` and `codegraph_explore` — following a request reaches the middleware and route handlers registered via `.Use()` / `.GET()` instead of dead-ending where the framework dispatches the chain dynamically.
- `codegraph_explore` now sizes its response to the *answer* instead of the file count: it shows the mechanism and the exact methods you asked about in full — even when they're buried deep in a large file — while collapsing the redundant interchangeable implementations of an interface (an HTTP interceptor chain, a query-compiler family) down to signatures. Fewer tokens for a more complete answer, so on the flows that used to occasionally cost more than plain grep/read it's now clearly cheaper — and the win holds across small, medium, and large codebases. Distinct, non-interchangeable code is shown in full as before. Disable with `CODEGRAPH_ADAPTIVE_EXPLORE=0`.
- Swift deferred-validation flows (and similar "handler array" patterns) now connect end-to-end in `codegraph_trace` and `codegraph_explore` — following a request's lifecycle reaches the validators registered with `.validate { … }` instead of dead-ending where the framework runs them by iterating a stored list of closures. Any pattern where closures are appended to a collection and later invoked by looping over it is now traced.
- `codegraph_explore` now spells out the dynamic-dispatch relationships of the symbols you ask about — e.g. "the closures registered here are run by `didCompleteTask`" — so the indirect hops you'd otherwise grep to reconstruct are listed alongside the call flow.
- `codegraph_explore` answers multi-phase questions that span a large "god file" far more completely. For a flow like "build, send, and validate a request" — where one big file holds the build chain and the validate logic lives in others — it now keeps every method *on the flow path* in full, collapses the file's off-path methods to one-line signatures, and guarantees each phase's defining file is shown (instead of truncating at a fixed size and dropping whichever phase came last, which sent you to read it by hand). Incidental files that merely name-drop the flow are still trimmed, so the response stays focused on the code that answers the question.
- CodeGraph is usable as an embedded library again: `require("@colbymchenry/codegraph")` and `import` now resolve the programmatic API — the `CodeGraph` class plus building blocks like `DatabaseConnection`, `QueryBuilder`, `initGrammars`, and `FileLock` — so you can drive the graph directly from your own app (for example an Electron process) instead of only through the CLI or MCP server. Embedding runs on your own runtime, so it needs Node 22.5+ for the built-in SQLite. (#354)

### Fixes

- `codegraph_trace` now resolves an overloaded symbol name to its real implementation instead of an empty protocol/delegate stub. Tracing a flow through a heavily-overloaded API (common in Swift, Java, C#, and Go) could land on an unrelated no-op method that happened to share the name and report "no path"; it now picks the substantive definition the flow actually runs through.
- CodeGraph's MCP server now answers an agent's opening handshake the instant it launches instead of blocking while the index loads, so a fresh session's very first tool call no longer occasionally races a server that's still warming up and falls back to grep/read. The first question in a new session now reliably goes through CodeGraph.
- Indexing a project that contains only config-style files (YAML, Twig, or `.properties`) no longer misleadingly reports "No files found to index" — these files are tracked at the file level and are now counted as indexed. Thanks @luojiyin1987 (#357).

## [0.9.7] - 2026-05-28

### New Features

- Go: gRPC interface stubs now connect to their hand-written implementation, so callers, callees, impact, and trace land on the real method instead of an empty generated stub.
- Generated files (protobuf, gRPC stubs, mocks, build output) now rank last in search, trace, and explore, so results land on your real implementation instead of an auto-generated placeholder.
- When `codegraph_trace` can't find a static path (a dynamic-dispatch break), it now inlines both endpoints' source, callers, and callees in one response, so the agent gets the full picture without a flurry of follow-up calls.
- Trace now picks the right endpoints in large multi-module repos by preferring symbols that share a directory, instead of grabbing an arbitrary same-named symbol from an unrelated module.
- Test files are now deprioritized in `codegraph_explore` (Go, Ruby, JS/TS, Java/Kotlin/Scala), so the explore budget goes to your real implementation source.
- Small projects (under ~500 files) now resolve flow questions in fewer MCP calls, with a leaner tool surface and tuned context and explore output sized for the project.
- `codegraph_context` now auto-traces flow questions like "how does X reach Y" or "trace the path from A to B", splicing the trace into the response so you don't need a separate `codegraph_trace` call.
- `codegraph_context` now inlines a URL-to-handler routing table and the source of your main routes file for routing questions on small projects, so you don't have to go read `routes.rb` or `web.php` yourself.
- `codegraph_context` search now boosts results in the directory of a project's core framework file, so a small same-named extension file no longer outranks the actual framework core.
- Interface-to-implementation linking now works for C#, TypeScript, JavaScript, Swift, and Scala (previously Java/Kotlin only), so investigating an interface method surfaces its concrete implementations.
- MCP tool descriptions are now shorter, trimming per-session overhead while keeping the steering guidance.
- Java and Kotlin imports now resolve by fully-qualified name, so same-name classes in different packages are told apart correctly in multi-module Spring and Android codebases, including across the Java/Kotlin interop boundary.
- Java and C# anonymous classes (`new T() { ... }`) and their overridden methods are now indexed as real class nodes, so an agent sees those hidden overrides in its trail without a Read.
- The installer no longer writes a duplicate `## CodeGraph` instructions block into your agent's instructions file (`CLAUDE.md`, `AGENTS.md`, `GEMINI.md`, Cursor's `.cursor/rules/codegraph.mdc`, or Kiro's steering doc) — the MCP server is now the single source of truth, and re-running `codegraph install` or `codegraph uninstall` strips a block a previous version left behind (#529). If you added your own notes inside the `CODEGRAPH_START`/`CODEGRAPH_END` markers, move them outside the markers first, since the whole marked block is removed.

### Fixes

- MCP tools no longer return results for files that were deleted while no server was running — the first query of a session now waits for the catch-up sync, so you get the correct index instead of stale rows.
- Windows: black console windows no longer flash on every file save or MCP reconnect (#485, #510, #530).
- `codegraph index` and `init -i` now report the true edge count in their summary, instead of undercounting by missing resolution and synthesizer edges.

## [0.9.6] - 2026-05-27

### New Features

- Enterprise Spring and MyBatis flows now trace end-to-end: MyBatis XML mappers are indexed and linked to their Java mapper interfaces, Spring `@Value` and `@ConfigurationProperties` references resolve to the matching keys in your `application.yml`/`.properties` config (including relaxed kebab/camel/snake binding), and field-injected concrete beans like `this.field.method()` resolve through to their implementation (#389).
- Gemini CLI (and the rebranded Antigravity CLI) plus the Antigravity IDE are now supported by `codegraph install`, detected and configured out of the box with sibling settings and MCP servers preserved across re-installs (#399).
- Kiro (CLI and IDE) is now supported by `codegraph install` on macOS, Linux, and Windows, with its own steering file so it loads CodeGraph guidance naturally (#385).

### Fixes

- C/C++: bare `#include "header.h"` directives now connect to the real header file instead of a phantom import, so includes show up as true file-to-file edges; system and stdlib headers are filtered out so they don't false-resolve (#453).
- Java/Kotlin: imports now disambiguate same-name classes across modules using the fully-qualified import path, so callers, callees, and trace land on the right class in multi-module projects instead of guessing by file proximity (#314).
- TypeScript: `type` aliases with object shapes (including function-typed members and intersection types) now surface their members in the graph, so a call like `handle.stop()` resolves to the alias member instead of an unrelated look-alike class in a sibling directory (#359).
- C#: parameter, return, property, and field types now produce reference edges, so callers and callees on a DTO or service type return real results instead of nothing (#381).
- Go: cross-package qualified calls like `pkg.Func()` now resolve to the right package by reading your `go.mod`, so callers, callees, impact, and trace return complete results on Go monorepos instead of almost nothing (#388).
- `codegraph_files` now returns the whole project when an agent passes a root-ish path like `/`, `.`, `./`, `""`, or a Windows-style `\`, and subdirectory filters like `/src`, `./src`, and `src\components` all resolve correctly instead of returning "No files found" (#426).
- The file watcher no longer marks edited files as fresh when another process holds the index lock, so the per-file staleness signal stays accurate until the edit is actually indexed (#449).
- TypeScript/JavaScript: calls inside top-level variable initializers (`const token = getToken()`) and inside inline object-literal methods are no longer dropped, so they show up in callers as expected, including in Vue single-file components (#425).
- Watch sync no longer aborts with a `FOREIGN KEY constraint failed` error in a long-running daemon; a stale lookup now drops a single edge instead of failing the whole sync (#455).
- Hermes: `codegraph install --target hermes` no longer corrupts `~/.hermes/config.yaml`, correctly handling PyYAML's block-style lists and re-installing cleanly even on an already-corrupted file (#456).
- NestJS: route prefixes from `RouterModule.register([...])` (including nested `children`) now propagate to controller routes, so a route shows up at its full path like `GET /admin/users` instead of `GET /` (#459).
- C++: callers now resolve through typed member pointers such as `m_alg->Processing()`, including out-of-line method definitions and the common case of two classes sharing a method name (#445, #454).

## [0.9.5] - 2026-05-25

### New Features

- Running multiple AI agents in the same project no longer multiplies the cost: two Claude Code windows, a worktree agent, or parallel sub-agents now share one background daemon per project with a single file watcher, SQLite connection, and tree-sitter warm-up instead of N independent copies (#411).
- The daemon runs detached so it outlives any single session, meaning closing one editor or terminal never severs the others; it lingers briefly after the last client disconnects so back-to-back sessions skip the startup cost, then exits and cleans up after itself. Tune the idle wait with `CODEGRAPH_DAEMON_IDLE_TIMEOUT_MS` (default five minutes).
- Set `CODEGRAPH_NO_DAEMON=1` to opt out and get one independent server per client, handy for debugging or sandboxes that disallow local sockets; the daemon is also version-pinned, so upgrading CodeGraph never mixes versions over the connection.
- CodeGraph responses now tell the agent which files are pending re-index: when the watcher has seen edits since the last sync, tool responses add a warning banner naming the stale files and their state so the agent reads just those directly while trusting the rest, with zero cost when nothing is pending (#403).
- `CODEGRAPH_WATCH_DEBOUNCE_MS` lets you tune the file-watcher quiet window (default 2000ms) for workspaces with bursty writes like format-on-save chains or large generated outputs, without touching your agent's command line (#403).
- Objective-C indexing: `.m`, `.mm`, and content-sniffed `.h` files now parse with full structural extraction, including full multi-part selectors, properties, imports, and superclass/protocol relationships, so trace, callers, and callees work across iOS codebases (#165).
- Mixed iOS, React Native, and Expo projects now trace end-to-end across language boundaries: Swift to Objective-C auto-bridging, the React Native legacy bridge and TurboModules, native-to-JS event channels, Expo Modules, and Fabric/Codegen view components are all bridged so flows connect through gaps that static parsing alone can't follow (#401).

### Fixes

- TypeScript: types used only in an interface's property or method signatures now produce references edges, so impact and callers on a type include every consumer that imports it just for an interface shape (#432).
- Git worktrees no longer silently borrow another tree's index; running CodeGraph from a worktree nested inside the main checkout used to return the wrong branch's code with no warning, and now both the status command and every read tool call out the conflict and point you to `codegraph init -i` in the worktree (#155).
- The file watcher no longer exhausts the OS file-watch budget on large repos: it now excludes the same directories the indexer ignores (defaults plus your `.gitignore`) before registering watches, so CodeGraph can run alongside your editor or dev server without hitting the per-user watch ceiling (#276).
- The index now stays in sync after `git pull`, branch switches, and edits made outside your editor; change detection is filesystem-based instead of relying on `git status`, so pulled or checked-out code is picked up without a full re-index.
- The MCP server now catches up on connect, reconciling anything that changed while it wasn't running so your first query reflects the current code instead of a stale snapshot.
- Dependency, build, and cache directories like `node_modules`, `vendor`, `dist`, `build`, `target`, `.venv`, `__pycache__`, `Pods`, and `.next` are now excluded by default, so context and search reflect your code instead of third-party noise even in a project with no `.gitignore`; add a `.gitignore` negation to index one anyway (#407).

## [0.9.4] - 2026-05-24

### New Features

- Request-to-handler flows now trace end-to-end across many web stacks, with new or improved route resolution for Express, Rails, Spring (Java and Kotlin), Django/DRF, Laravel, Flask, FastAPI, Gin, chi, ASP.NET, Drupal, Axum, actix, Vapor, Play, Vue/Nuxt, Svelte/SvelteKit, and React Router.
- `codegraph_trace`, `codegraph_callees`, and `codegraph_explore` now follow flows that have no static call edge — callback and observer registration, EventEmitter, React re-renders and JSX children, Flutter `setState` to `build`, C++ virtual overrides, and Java/Kotlin interface-to-implementation dispatch (like Spring's `@Autowired` service calls) — and each bridged hop is labeled inline in trace with where it was wired up.
- `codegraph_trace` now returns a self-contained flow dossier: every hop shows its full body inline plus the destination's own outgoing calls, so a single trace usually answers a "how does X reach Y" question without a follow-up explore, node, or Read.
- `codegraph_explore` now leads with the execution flow when your query names the symbols of a flow, finding the call path among those symbols (including across dynamic-dispatch hops) so you get a trace-quality answer without switching tools.
- `codegraph_node` and `codegraph_trace` now emit line-numbered source (matching `codegraph_explore` and Read), so you can cite or edit exact lines without re-reading the file just to recover line numbers.
- New `CODEGRAPH_MCP_TOOLS` environment variable lets you expose only a chosen subset of codegraph tools over MCP (e.g. `trace,search,node,context`) without editing your client's MCP config; unset exposes all of them.
- Release archives now ship with a `SHA256SUMS` file, and the npm launcher verifies the bundle it downloads against it, aborting on a mismatch (releases published before this change skip verification rather than failing).

### Fixes

- Several static-extraction and resolution correctness fixes underpin the routing work above: C++ inheritance edges that were previously missing, Dart methods that were extracted signature-only, Python handlers named `index`/`get`/`update` that were being silently dropped, and an explore output-budget issue that under-returned source on repos with very large files.
- `codegraph serve --mcp` no longer keeps running after its parent agent is force-killed (OOM, `kill -9`, or container teardown) on Linux, where it used to hold inotify watches, file descriptors, and the SQLite WAL indefinitely; the server now shuts down as soon as its parent process changes, tunable via `CODEGRAPH_PPID_POLL_MS` (#277).
- Installing `@colbymchenry/codegraph` through a registry mirror that hadn't yet mirrored the matching per-platform package no longer fails with `no prebuilt bundle for <platform>`; the launcher now downloads the bundle from GitHub Releases and caches it, with `CODEGRAPH_NO_DOWNLOAD=1` to disable the fallback and `CODEGRAPH_DOWNLOAD_BASE` to point it at your own mirror (#303).
- `install.sh` no longer fails with `403` / "could not resolve latest version" on shared or cloud hosts that exhaust GitHub's unauthenticated API rate limit; it now resolves the version through the unthrottled releases redirect, and `CODEGRAPH_VERSION` accepts a bare version like `0.9.4` as well as `v0.9.4` (#325).

## [0.9.3] - 2026-05-22

### New Features

- New `codegraph uninstall` command cleanly removes CodeGraph from every agent it's configured on — Claude Code, Cursor, Codex CLI, opencode, and Hermes Agent — in one step, asking whether to clean up your global or this project's local config and reporting exactly which agents it touched; it accepts `--location`, `--target`, and `--yes` for scripted or non-interactive use, removes only what `codegraph install` wrote, and leaves your `.codegraph/` index alone (#313).

### Fixes

- Indexing a large multi-language project no longer aborts partway through with a `Fatal process out of memory: Zone` crash on Node.js 22 and 24, even with plenty of RAM free — CodeGraph now launches with a V8 flag that keeps grammar compilation off the optimizing tier, and any launch path that doesn't get the flag directly re-execs once with it automatically (#298, #293). Node 25 stays blocked for now, since its variant of this bug isn't fixed by the same flag.
- Uninstalling from Cursor now deletes the leftover `.cursor/rules/codegraph.mdc` file outright instead of leaving an orphaned, empty rule behind, while keeping any content you added outside CodeGraph's markers.

## [0.9.2] - 2026-05-21

### Breaking Changes

- CodeGraph no longer has a config file: `.codegraph/config.json` and the entire config surface are gone, and the library API for it (the config type, the `config` option on `init()`, and the get/update config exports) has been removed — existing config files are now ignored, and `.gitignore` is the single source of truth for what gets indexed. The `.codegraphignore` marker is also no longer supported; use `.gitignore` instead.

### New Features

- `codegraph install` now supports Hermes Agent (Nous Research), wiring up the CodeGraph MCP server so Hermes can drive the knowledge graph like the other agents.
- Drupal projects (8/9/10/11) are now detected and indexed with framework smarts: routes from `*.routing.yml` link to their controller, form, or entity-handler, and hook implementations across modules are connected to their canonical hook name, so asking for callers of a hook returns every implementation (#268).
- Indexing is now zero-config and honors your `.gitignore` everywhere — in git repos via git, and in non-git projects by reading `.gitignore` files directly — so to keep something out of the graph you just add it to `.gitignore`. Behavior change: committed files that aren't gitignored are now indexed even under `vendor/`, `Pods/`, or a committed `dist/`; add a `.gitignore` negation to exclude them (#283).

### Fixes

- Windows: installing globally and then running any `codegraph` command no longer fails — the launcher now invokes the bundled runtime directly instead of a `.cmd` file that modern Node refuses to spawn, so `codegraph` works regardless of your Node version (#289).

### Security

- The temp-dir marker written on each `codegraph_context` call is now opened safely so it can't follow a symlink, closing a hole where another local user on a shared machine could redirect that write onto a file you can write (#280).

## [0.9.1] - 2026-05-21

### Fixes

- The standalone installers (`curl … | sh` and `irm … | iex`) no longer fail to launch on a machine that has no Node installed.
- Installing with `npm i -g` on Linux x64 now finds its bundle, after the 0.9.0 release silently shipped without the linux-x64 package; the release pipeline now verifies every package reached the npm registry so a release can't pass green-but-broken again.

## [0.9.0] - 2026-05-21

CodeGraph now ships its own self-contained runtime, so it installs on any Node version — or none at all — with no native build step, and the old intermittent "database is locked" errors are gone for good.

### New Features

- One-line standalone installers that need no Node.js: `install.sh` on macOS and Linux, and `install.ps1` on Windows fetch the self-contained bundle and put `codegraph` on your PATH (you can still use `npm`/`npx` on any Node version too).
- CodeGraph now uses real SQLite with full WAL and FTS5 built into its bundled runtime, which fixes the concurrent-read "database is locked" errors at the root, removes the native build step entirely, and runs faster for anyone who had been stuck on the old WASM fallback (#238).
- Lua: CodeGraph now indexes `.lua` projects (Neovim plugins, Kong, OpenResty, game code), surfacing functions, table methods, local variables, `require(...)` imports, and the call edges between them.
- Luau: CodeGraph now indexes `.luau`, Roblox's typed superset of Lua, adding type and `export type` aliases, typed function signatures, generics, and Roblox instance-path requires on top of everything Lua extracts (#232).
- `codegraph status` now reports the effective journal mode, so a "database is locked" report is easy to triage at a glance.

### Fixes

- Re-running `codegraph install` now strips the broken auto-sync hooks that pre-0.8 versions wrote into Claude Code's settings, which had been causing a "Stop hook error: unknown command 'sync-if-dirty'" on every turn. The cleanup is surgical and leaves unrelated hooks untouched. Re-run `codegraph install` once on an affected machine to clear the error.

## [0.8.0] - 2026-05-20

### Breaking Changes

- The minimum supported Node.js version is now 20 (Node 18 is end-of-life); Node 22 LTS and Node 24 get the fast native backend out of the box, other Node versions still run via the slower WASM fallback, and Node 25+ remains blocked (#81). If you're on an older Node, upgrade to 20 or newer.

### New Features

- NestJS: CodeGraph now recognizes NestJS projects and surfaces the route that binds each handler across HTTP controllers, GraphQL resolvers, microservice handlers, and WebSocket gateways, detected automatically from any `@nestjs/*` dependency (#220).
- `codegraph_explore` source now includes line numbers, so an agent can cite `file:line` straight from the result instead of reopening the file to find a line number; set `CODEGRAPH_EXPLORE_LINENUMS=0` to disable.
- On WSL2 `/mnt/*` drives, where the live file watcher is too slow and could break MCP startup, CodeGraph now skips the watcher and offers to keep the index fresh with git hooks instead; new `CODEGRAPH_NO_WATCH=1` (or `serve --mcp --no-watch`) forces the watcher off anywhere, and `CODEGRAPH_FORCE_WATCH=1` overrides the WSL auto-detect when your setup is actually fast.
- CodeGraph now guides agents to answer "how does X work" and architecture questions directly with a couple of codegraph calls instead of delegating to a file-reading sub-agent or a grep-and-read loop, which gives faster, cheaper, `file:line`-cited answers on medium and large repos.
- `codegraph_node` with code on a class, interface, struct, or enum now returns a compact member outline (fields plus method signatures with line numbers) instead of the entire class body; functions and methods still return their full source.
- `codegraph_explore` output now scales with project size, so small projects get tighter responses than your native grep-and-read flow would produce while large codebases keep their fuller budget, and a per-file cap stops a single dense file from collapsing into a whole-file dump (#185). Thanks @essopsp.
- Search ranking now correctly de-prioritizes CamelCase test files (`FooTest.kt`, `BarTests.swift`, `BazSpec.scala`, `QuxTestCase.cs`) and test source-set directories in Kotlin, Swift, Scala, and C#, so real implementations no longer get outranked by tests.

### Fixes

- `codegraph_explore` output is now hard-capped to its size budget, so an oversized response no longer overruns the cap and sits in the agent's context to be re-read every turn.
- Newly created untracked files are no longer reported as pending forever and re-indexed from scratch on every `codegraph sync`; CodeGraph now hash-compares them against the index the same way it does tracked files (#206). Thanks @15290391025.
- `codegraph init -i` now finds source inside nested, independent git repositories that aren't submodules (common in CMake super-repo layouts), instead of reporting "No files found to index" (#193). Thanks @timxx.
- On Node 24, indexing no longer silently drops to the slower fallback backend with a warning that couldn't be cleared; a fresh install on Node 22 or 24 now gets the fast native backend with no compiler, and `codegraph status` should report it (#203). Thanks @Finndersen.
- MCP tools no longer fail with "CodeGraph not initialized" when the index actually exists; when the client doesn't report a workspace root, the server now asks for it via the standard MCP `roots/list` request before falling back, and the error message is actionable when a project still can't be resolved (#196). Thanks @zhangyu1197.
- The MCP server no longer hangs on startup under WSL2 when the project lives on an NTFS `/mnt/*` mount, so the codegraph tools actually appear; CodeGraph auto-skips the watcher there with manual and git-hook sync fallbacks (#199). Thanks @mengfanbo123.
- Claude Code project-local installs now write the MCP server to `.mcp.json` (the file Claude Code actually reads for project-scoped servers) instead of a file it ignores, and re-running `codegraph install` migrates an affected project automatically (#207). Thanks @Jhsmit.
- Source-omission markers in `codegraph_explore` and `codegraph_context` output are now language-neutral instead of C-style comments, so they no longer look wrong inside Python, Ruby, and other non-C source blocks.

## [0.7.10] - 2026-05-19

### Fixes

- CodeGraph tools now reliably appear in your client on slow filesystems (Docker Desktop VirtioFS on macOS, WSL2), where the startup handshake could previously time out and leave the process running with no tools visible (#172). Thanks @sashanclrp and @sgrimm.
- On Windows PowerShell and cmd.exe, terminal output during `codegraph index` and `codegraph sync` no longer turns into garbled characters; CodeGraph now uses plain ASCII glyphs by default on Windows, with `CODEGRAPH_UNICODE=1` to opt back into the Unicode glyphs or `CODEGRAPH_ASCII=1` to force ASCII on any platform (#168). Thanks @starkleek and @Bortlesboat.
- Module-qualified symbol lookups now resolve in the codegraph tools, so you can pass names like `module::symbol` (Rust, C++, Ruby), `Module.symbol` (TypeScript, JavaScript, Python), or `module/symbol`, including multi-level paths and Rust prefixes like `crate`, `super`, and `self` (#173). Thanks @joselhurtado.

## [0.7.9] - 2026-05-17

### New Features

- opencode: the installer now writes an `AGENTS.md` file with CodeGraph usage guidance, so opencode reaches for the `codegraph_*` tools instead of falling back to its native search.
- opencode: your comments and formatting in `opencode.jsonc` now survive install, re-install, and uninstall, because the installer makes surgical edits instead of rewriting the whole file.

### Fixes

- opencode: `codegraph install` now wires up the MCP server in the file opencode actually reads — previously it wrote to a config file opencode ignores by default, so the CodeGraph entry never appeared in any opencode session; re-run `codegraph install --target=opencode` after upgrading so the entry lands in the right place.

## [0.7.7] - 2026-05-17

### New Features

- `codegraph install` now sets up Claude Code, Cursor, Codex CLI, and opencode from one multi-select prompt, with any agents it detects pre-checked, so a single install wires up every editor you use (#137).
- You can install non-interactively for scripting and CI with flags like `--target`, `--location`, `--yes`, `--no-permissions`, and `--print-config`.
- `codegraph init` now auto-wires project-local agent config for any agent you installed globally, so one global `codegraph install` works in every project you open without re-installing per project.
- Agent instructions are now agent-agnostic and tell each agent to trust codegraph results instead of re-verifying with grep, fixing the case where Cursor and Codex fell back to native search even with codegraph available.
- The install prompts are clearer: the agent picker comes first, and the separate "install the CLI on your PATH" and "apply to all projects or just this one" questions no longer both read as "Global".

### Fixes

- Cursor: a globally-installed codegraph no longer reports "not initialized" in every workspace; the installer now passes the correct project path into Cursor's MCP config to work around Cursor launching MCP servers with the wrong working directory.

Thanks @andreinknv for the substantive draft this release was based on.

## [0.7.6] - 2026-05-13

### Fixes

- Fixed the `codegraph` command failing with `permission denied` right after a fresh global install — the 0.7.5 package shipped the CLI without its executable bit, so your shell refused to run it. New installs work out of the box. If you're stuck on 0.7.5, upgrade to 0.7.6 or unblock yourself in place by making the installed binary executable with `chmod +x`.

[0.9.7]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.7
[0.9.6]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.6
[0.9.5]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.5
[0.9.4]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.4
[0.9.3]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.3
[0.9.2]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.2
[0.9.1]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.1
[0.9.0]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.0
[0.8.0]: https://github.com/colbymchenry/codegraph/releases/tag/v0.8.0
[0.7.10]: https://github.com/colbymchenry/codegraph/releases/tag/v0.7.10
[0.7.9]: https://github.com/colbymchenry/codegraph/releases/tag/v0.7.9
[0.7.7]: https://github.com/colbymchenry/codegraph/releases/tag/v0.7.7
[0.7.6]: https://github.com/colbymchenry/codegraph/releases/tag/v0.7.6
[0.9.8]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.8
[0.9.9]: https://github.com/colbymchenry/codegraph/releases/tag/v0.9.9
[1.0.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.0.0
[1.0.1]: https://github.com/colbymchenry/codegraph/releases/tag/v1.0.1
[1.1.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.0
[1.1.1]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.1
[1.1.2]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.2
[1.1.3]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.3
[1.1.4]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.4
[1.1.5]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.5
[1.1.6]: https://github.com/colbymchenry/codegraph/releases/tag/v1.1.6
[1.2.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.2.0
[1.3.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.3.0
[1.3.1]: https://github.com/colbymchenry/codegraph/releases/tag/v1.3.1
[1.4.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.4.0
[1.4.1]: https://github.com/colbymchenry/codegraph/releases/tag/v1.4.1
[1.5.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.5.0
[1.6.0]: https://github.com/colbymchenry/codegraph/releases/tag/v1.6.0