deepseek-harness/.github/workflows/e2e.yml

name: E2E (real DeepSeek API)

# Real-API end-to-end suite (`pnpm run test:e2e`). Unlike ci.yml this job
# consumes the DEEPSEEK_API_KEY_EXTERNAL secret (mapped to the env var
# DEEPSEEK_API_KEY the tests read) and hits the external API at
# https://api.deepseek.com/anthropic (DEEPSEEK_BASE_URL is pinned explicitly so a
# stray repo-root .env can't redirect the run).
#
# pull_request IS a trigger, but GitHub withholds repo secrets from BOTH forked
# PRs and Dependabot PRs (the latter are same-repo, fork==false, but still
# keyless). The job-level `if:` below skips the whole job for those untrusted
# PRs. The Dependabot test keys on the PR AUTHOR (pull_request.user.login), not
# github.actor (the run trigger) — a maintainer reopening a Dependabot PR would
# make github.actor human while the PR is still keyless. When the job runs, the
# secret is expected, so the preflight hard-fails if it's missing (a missing
# secret would otherwise make the self-skipping suite report a false green). A
# job skipped by a job-level `if:` reports as a SUCCESSFUL check, so this
# workflow is safe to use as a required status check if desired.
#
# SECURITY — NEVER change this trigger to `pull_request_target`. `pull_request`
# checks out and runs the PR's HEAD code WITHOUT secrets for forks, which is what
# keeps an untrusted fork from exfiltrating the key. `pull_request_target` runs
# in the BASE repo's context WITH secrets while still able to check out untrusted
# fork code — a textbook key-leak vector, especially once this repo is public.
# The fork/secret model and its public-repo implications are recorded in
# .agents/notes/implemented/testing/2026-06-19-real-api-e2e-ci.md.
#
# Note: scheduled triggers are auto-disabled after 60 days of repo inactivity;
# push/pull_request/workflow_dispatch act as backstops.
on:
  workflow_dispatch:
  push:
    branches: [main, master]
  pull_request:
  schedule:
    # 00:17 UTC nightly = 08:17 Asia/Shanghai — off the top-of-hour cron stampede.
    - cron: '17 0 * * *'

# Keep only the newest validation in each workflow/ref, including master
# pushes, nightly runs, and manual dispatches.
concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true

# Least privilege: this job only reads the repo to run tests.
permissions:
  contents: read

env:
  # CI runs must never report to the production telemetry endpoint baked
  # into apps/cli/cordis.yml (AppCLIEntry disables the row when set).
  DSH_TELEMETRY_DISABLED: '1'

jobs:
  e2e:
    runs-on: >-
      ${{ vars.DSH_CI_FAILOVER_LINUX == 'blacksmith'
          && 'blacksmith-4vcpu-ubuntu-2404' || 'ubuntu-latest' }}
    name: e2e
    # Run on every trusted event. Skip untrusted PRs (forks + Dependabot) where
    # the secret is withheld — they would otherwise hard-fail the preflight.
    if: >-
      github.event_name != 'pull_request'
      || !(github.event.pull_request.head.repo.fork || github.event.pull_request.user.login == 'dependabot[bot]')
    # Profile e2e files can each own several complete dsh subprocess trees, so
    # four file workers preserve process/PTY headroom. Tests retain 120s/test
    # and retry 2; 45m still bounds retry storms against a slow API.
    timeout-minutes: 45
    steps:
      - uses: actions/checkout@v6

      - uses: pnpm/action-setup@v4

      - uses: actions/setup-node@v6
        with:
          node-version: 24
          cache: pnpm

      - name: Install (immutable)
        run: pnpm install --frozen-lockfile

      # The with-key escalation e2e self-skips without a usable runner. Prepare
      # bwrap so trusted CI exercises it; the functional probe remains the
      # authority on whether the runner can use the sandbox.
      - name: Prepare bubblewrap (unrestrict userns)
        run: bash scripts/prepare-ci-bubblewrap.sh

      # Guard against a false green: the e2e suites self-skip when the key is
      # absent, so a missing/misconfigured secret would otherwise pass as
      # "all skipped". This job only runs on trusted events (the `if:` above
      # excludes keyless PRs), so the secret MUST be present — fail loudly if not.
      - name: Preflight (require DEEPSEEK_API_KEY)
        env:
          # Repo secret DEEPSEEK_API_KEY_EXTERNAL holds the external-API key;
          # the tests read process.env.DEEPSEEK_API_KEY, so map it across here.
          DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY_EXTERNAL }}
        run: |
          set -euo pipefail
          if [ -z "${DEEPSEEK_API_KEY:-}" ]; then
            echo "::error::DEEPSEEK_API_KEY is empty. The e2e suite would self-skip and"
            echo "::error::report a false green. Configure the repo secret DEEPSEEK_API_KEY_EXTERNAL."
            exit 1
          fi
          echo "DEEPSEEK_API_KEY present."

      # The e2e suites boot the example bins in `lib` mode (DSH_EXAMPLE_MODE=lib):
      # the built artifact under plain Node, resolving plugins through real package
      # exports — the shape a real consumer runs. That requires a prior build.
      - name: Build (lib for the e2e example bins)
        run: pnpm run build:official

      # Real-API end-to-end tests only. The keyless gates (lint/typecheck/
      # coverage/snapshot/etc.) already run in ci.yml (pull requests) and
      # ci-master.yml (master push standby).
      # DEEPSEEK_BASE_URL is pinned to the external API; the secret is scoped to
      # this step (and preflight) only — never exposed to checkout/setup/install.
      - name: E2E tests (real DeepSeek API)
        env:
          DEEPSEEK_API_KEY: ${{ secrets.DEEPSEEK_API_KEY_EXTERNAL }}
          DEEPSEEK_BASE_URL: https://api.deepseek.com/anthropic
          DSH_E2E_MAX_WORKERS: 4
          DSH_EXAMPLE_MODE: lib
        run: pnpm run test:e2e