invariant.spec.ts 7.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145
  1. import { describe, expect, it } from 'vitest'
  2. import { Context } from '@deepseek-ai/cordis'
  3. import SessionStore, { SessionId } from '@deepseek-ai/dsh-session'
  4. import InvariantRegistry from '@deepseek-ai/dsh-invariants'
  5. import * as AgentLoopInvariant from '@deepseek-ai/dsh-agent-loop/invariant'
  6. import { createUserMessage, markAgentLoopRequest, type GenerateOptions } from '@deepseek-ai/dsh-llm'
  7. async function setup(): Promise<Context> {
  8. const ctx = new Context()
  9. await ctx.plugin(SessionStore)
  10. await ctx.plugin(InvariantRegistry)
  11. await ctx.plugin(AgentLoopInvariant)
  12. return ctx
  13. }
  14. function dispatch(ctx: Context, options: unknown): void {
  15. void ctx.waterfall('llm/stream', options as never, () => (async function* () {})() as never)
  16. }
  17. function loopRequest<T extends object>(options: T): Readonly<T> {
  18. markAgentLoopRequest(options as GenerateOptions)
  19. return Object.freeze(options)
  20. }
  21. async function requestSetup() {
  22. const ctx = await setup()
  23. const session = ctx.sessions.create(SessionId('req-check'))
  24. session.append('turn/start', { turn: 1 })
  25. session.append('user/message', createUserMessage({
  26. content: [{ type: 'text', text: 'hi' }], source: { kind: 'user' },
  27. }), { surfaceOp: 'append' })
  28. const boundary = session.deriveMessages()
  29. session.append('step/start', { turn: 1, step: 1 })
  30. session.append('request/header', { header: { config: { provider: 'mock', model: 'm' } }, reason: 'initial' })
  31. return { ctx, session, boundary }
  32. }
  33. describe('request-reconstruction invariant', () => {
  34. it('accepts a frozen request equal to the boundary derivation and folded header', async () => {
  35. const { ctx, session, boundary } = await requestSetup()
  36. const options = loopRequest({ model: 'm', messages: Object.freeze(boundary), sessionId: session.id })
  37. expect(() => { dispatch(ctx, options) }).not.toThrow()
  38. })
  39. it('includes context appended inside the open step before dispatch', async () => {
  40. const { ctx, session } = await requestSetup()
  41. session.append('user/message', createUserMessage({
  42. content: [{ type: 'text', text: '[step context]' }], source: { kind: 'plugin', plugin: 'x' },
  43. }), { surfaceOp: 'append' })
  44. const options = loopRequest({
  45. model: 'm',
  46. messages: Object.freeze(session.deriveMessages()),
  47. sessionId: session.id,
  48. })
  49. expect(() => { dispatch(ctx, options) }).not.toThrow()
  50. })
  51. it('requires the messages to equal the boundary derivation exactly (no unlogged prefix)', async () => {
  52. const { ctx, session, boundary } = await requestSetup()
  53. const extra = { role: 'user' as const, content: [{ type: 'text' as const, text: '<system-reminder>catalog</system-reminder>' }] }
  54. expect(() => { dispatch(ctx, loopRequest({ model: 'm', messages: Object.freeze([...boundary]), sessionId: session.id })) })
  55. .not.toThrow()
  56. expect(() => { dispatch(ctx, loopRequest({ model: 'm', messages: Object.freeze([extra, ...boundary]), sessionId: session.id })) })
  57. .toThrow(/diverges from the dispatch-time durable derivation/)
  58. expect(() => { dispatch(ctx, loopRequest({ model: 'm', messages: Object.freeze([...boundary, extra]), sessionId: session.id })) })
  59. .toThrow(/diverges from the dispatch-time durable derivation/)
  60. })
  61. it('rejects message and header divergence', async () => {
  62. const { ctx, session, boundary } = await requestSetup()
  63. const divergent = [...boundary, { role: 'user', content: [{ type: 'text', text: 'phantom' }] }]
  64. expect(() => { dispatch(ctx, loopRequest({ model: 'm', messages: Object.freeze(divergent), sessionId: session.id })) })
  65. .toThrow(/diverges from the dispatch-time durable derivation/)
  66. expect(() => { dispatch(ctx, loopRequest({ model: 'other', messages: Object.freeze(boundary), sessionId: session.id })) })
  67. .toThrow(/diverges from the folded request header/)
  68. // The system prompt is surface node 0 inside `messages`; a `system` field is an unlogged prefix.
  69. expect(() => { dispatch(ctx, loopRequest({ model: 'm', system: 'unlogged', messages: Object.freeze(boundary), sessionId: session.id })) })
  70. .toThrow(/diverges from the folded request header/)
  71. })
  72. it('rejects loop requests with no boundary or header', async () => {
  73. const ctx = await setup()
  74. const session = ctx.sessions.create(SessionId('req-bare'))
  75. session.append('turn/start', { turn: 1 })
  76. const bare = loopRequest({ model: 'm', messages: Object.freeze([]), sessionId: session.id })
  77. expect(() => { dispatch(ctx, bare) }).toThrow(/no step\/start/)
  78. session.append('step/start', { turn: 1, step: 1 })
  79. expect(() => { dispatch(ctx, bare) }).toThrow(/no request\/header event/)
  80. })
  81. it('rejects an unfrozen messages array but skips requests outside the loop contract', async () => {
  82. const { ctx, session, boundary } = await requestSetup()
  83. expect(() => { dispatch(ctx, loopRequest({ model: 'm', messages: [...boundary], sessionId: session.id })) })
  84. .toThrow(/frozen messages array/)
  85. expect(() => { dispatch(ctx, { model: 'summarizer', messages: [], sessionId: session.id }) }).not.toThrow()
  86. expect(() => { dispatch(ctx, Object.freeze({ model: 'm', messages: Object.freeze([]) })) }).not.toThrow()
  87. expect(() => { dispatch(ctx, Object.freeze({ model: 'm', messages: Object.freeze([]), sessionId: SessionId('ghost') })) })
  88. .not.toThrow()
  89. const directSession = ctx.sessions.create(SessionId('direct-one-shot'))
  90. expect(() => {
  91. dispatch(ctx, Object.freeze({ model: 'one-shot', messages: Object.freeze([]), sessionId: directSession.id }))
  92. }).not.toThrow()
  93. })
  94. it('rejects malformed requests carrying the loop marker', async () => {
  95. const { ctx, session } = await requestSetup()
  96. const messages: GenerateOptions['messages'] = []
  97. Object.freeze(messages)
  98. expect(() => {
  99. dispatch(ctx, markAgentLoopRequest({ provider: 'p', model: 'm', messages, sessionId: session.id }))
  100. }).toThrow(/request must be frozen/)
  101. expect(() => {
  102. dispatch(ctx, loopRequest({ model: 'm', messages: Object.freeze([]) }))
  103. }).toThrow(/carry a session id/)
  104. expect(() => {
  105. dispatch(ctx, loopRequest({
  106. model: 'm',
  107. messages: Object.freeze([]),
  108. sessionId: SessionId('missing-loop-session'),
  109. }))
  110. }).toThrow(/live session id/)
  111. })
  112. it('prepends ahead of a short-circuiting stream listener', async () => {
  113. const ctx = new Context()
  114. await ctx.plugin(SessionStore)
  115. ctx.on('llm/stream', () => (async function* () {})() as never)
  116. await ctx.plugin(InvariantRegistry)
  117. await ctx.plugin(AgentLoopInvariant)
  118. const session = ctx.sessions.create(SessionId('prepend-check'))
  119. session.append('turn/start', { turn: 1 })
  120. session.append('user/message', createUserMessage({
  121. content: [{ type: 'text', text: 'hi' }], source: { kind: 'user' },
  122. }), { surfaceOp: 'append' })
  123. session.append('step/start', { turn: 1, step: 1 })
  124. session.append('request/header', { header: { config: { provider: 'mock', model: 'm' } }, reason: 'initial' })
  125. const divergent = loopRequest({
  126. model: 'm',
  127. messages: Object.freeze([{ role: 'user', content: [{ type: 'text', text: 'phantom' }] }]),
  128. sessionId: session.id,
  129. })
  130. expect(() => { dispatch(ctx, divergent) }).toThrow(/diverges from the dispatch-time durable derivation/)
  131. })
  132. })