fix(agent-loop): report turn failures at source

This commit is contained in:
_Kerman
2026-07-31 13:54:36 +08:00
parent 7a463dbe44
commit 741fc7dc79
12 changed files with 125 additions and 44 deletions

View File

@@ -1,6 +1,6 @@
import { describe, expect, it } from 'vitest'
import { Context } from 'cordis'
import LlmService, { createUserMessage, freezeMessage, CallId, MessageSource, ProviderRequestId, StreamChunk } from '@deepseek-ai/dsh-llm'
import LlmService, { createUserMessage, freezeMessage, CallId, LlmError, MessageSource, ProviderRequestId, StreamChunk } from '@deepseek-ai/dsh-llm'
import SessionStore, { Session, SessionEvent, SessionId, TurnEndReason, type UserMessage } from '@deepseek-ai/dsh-session'
import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
import ToolRegistry, { defineContentToolFixture, type PostToolDecision } from '@deepseek-ai/dsh-tools'
@@ -594,12 +594,20 @@ describe('a finish-error stream chunk ends the turn as error, not completed', ()
const agent = ctx.agentLoop.create(SessionId('a-finish-error'), { provider: 'mock', model: 'mock' })
const reasons: TurnEndReason[] = []
const errors: unknown[] = []
ctx.on('agent/error', (_agent, turn, step, error) => {
expect({ turn, step }).toEqual({ turn: 1, step: 1 })
errors.push(error)
})
ctx.on('session/event', (_s, event) => { if (event.type === 'turn/end') reasons.push(event.data.reason) })
send(agent, 'go')
await waitForIdle(ctx, agent)
expect(reasons).toEqual([{ kind: 'error', error: failure }])
expect(errors).toHaveLength(1)
expect(errors[0]).toBeInstanceOf(LlmError)
expect((errors[0] as LlmError).failure).toEqual(failure)
const events = [...agent.session.events]
const turnEnd = events.find(event => event.type === 'turn/end')
@@ -809,6 +817,7 @@ describe('turn and step boundary recovery', () => {
expect(adapter.requests).toHaveLength(1)
expect(errors.map(error => error.message)).toEqual([
'reject first step-end',
'invariant violated by "@deepseek-ai/dsh-session": turn/end 1 while step 1 is still open',
])
expect(boundaryCounts(agent)).toMatchObject({
@@ -842,6 +851,7 @@ describe('turn and step boundary recovery', () => {
kind: 'error',
error: { message: 'provider 500', code: 'SERVER' },
})
expect(threw).toBe(true)
// loop survives: a second turn runs to completion (invariants oracle would
// throw on its turn/start if turn 1 had been left open).
@@ -1012,7 +1022,9 @@ describe('turn and step boundary recovery', () => {
expect(e.some(x => x.type === 'step/end')).toBe(true)
expect(e.some(x => x.type === 'turn/end')).toBe(true)
expect(e.at(-1)?.type).toBe('turn/end')
expect(errors).toEqual([])
expect(errors).toHaveLength(1)
expect(errors[0]).toBeInstanceOf(LlmError)
expect((errors[0] as LlmError).failure).toEqual({ message: 'provider 500', code: 'SERVER' })
// loop survives.
send(agent, 'again')

View File

@@ -181,7 +181,10 @@ describe('durable error rendering', () => {
const turnEnd = agent.session.events.find(e => e.type === 'turn/end')
expect(turnEnd).toBeDefined()
if (turnEnd?.type === 'turn/end' && turnEnd.data.reason.kind === 'error') {
expect(turnEnd.data.reason.error).toBe('server overloaded')
expect(turnEnd.data.reason.error).toEqual({
message: 'server overloaded',
code: 'RATE_LIMIT',
})
}
})
})

View File

@@ -1,6 +1,6 @@
import { describe, expect, it } from 'vitest'
import { Context } from 'cordis'
import LlmService, { createUserMessage, CallId, StreamChunk } from '@deepseek-ai/dsh-llm'
import LlmService, { createUserMessage, CallId, LlmError, StreamChunk } from '@deepseek-ai/dsh-llm'
import SessionStore, { SessionId, TurnEndReason } from '@deepseek-ai/dsh-session'
import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
import ToolRegistry, { defineContentToolFixture } from '@deepseek-ai/dsh-tools'
@@ -200,7 +200,9 @@ describe('agent loop', () => {
await waitForIdle(ctx, agent)
expect(adapter.requests).toHaveLength(0) // the request was never sent
expect(errors).toEqual([])
expect(errors.map(error => error.message)).toEqual([
'prompt variable "{{cwd}}" has no value for this assembly (section "deployment:persona")',
])
const turnEnd = agent.session.events.find(e => e.type === 'turn/end')
expect(turnEnd?.type === 'turn/end' && turnEnd.data.reason.kind).toBe('error')
expect(turnEnd?.type === 'turn/end' && turnEnd.data.reason.kind === 'error'
@@ -346,7 +348,7 @@ describe('agent loop', () => {
expect(JSON.stringify(adapter.requests[0]?.messages)).toContain('second idle steer')
})
it('contains a throwing step observer and carries steering into a replacement turn', async () => {
it('stops after a throwing step observer and retains steering until a later wakeup', async () => {
const adapter = new MockAdapter([textResponse('recovered')])
const ctx = await harness(adapter)
const agent = ctx.agentLoop.create(SessionId('failed-steering'), { provider: 'mock', model: 'mock' })
@@ -361,6 +363,13 @@ describe('agent loop', () => {
send(agent, 'prompt')
await waitForIdle(ctx, agent)
expect(adapter.requests).toHaveLength(0)
expect(agent.session.events.filter(event => event.type === 'turn/start')).toHaveLength(1)
expect(agent.inbox.nextStep).toHaveLength(1)
send(agent, 'resume')
await waitForIdle(ctx, agent)
expect(adapter.requests).toHaveLength(1)
expect(agent.session.events.filter(event => event.type === 'turn/start')).toHaveLength(2)
expect(agent.session.events.some(event => event.type === 'steering/message')).toBe(false)
@@ -658,7 +667,7 @@ describe('agent loop', () => {
send(agent, 'first')
await waitForIdle(ctx, agent)
// The first turn failed at step 1 before a model call.
expect(errors).toEqual([])
expect(errors.map(error => error.message)).toEqual(['boom in pre-step'])
expect(adapter.requests.length).toBe(0)
const firstTurnEnd = agent.session.events.find(e => e.type === 'turn/end')
expect(firstTurnEnd?.type === 'turn/end' && firstTurnEnd.data.reason).toMatchObject({ kind: 'error' })
@@ -1085,20 +1094,24 @@ describe('agent loop', () => {
const ctx = await harness(adapter)
const agent = ctx.agentLoop.create(SessionId('a1'), { provider: 'mock', model: 'mock' })
const errors: Error[] = []
const errors: unknown[] = []
const reasons: TurnEndReason[] = []
ctx.on('agent/error', (_agent, _turn, _step, error) => {
if (error instanceof Error) errors.push(error)
errors.push(error)
})
ctx.on('session/event', (_s, event) => { if (event.type === 'turn/end') reasons.push(event.data.reason) })
send(agent, 'hi')
await waitForIdle(ctx, agent)
expect(errors).toEqual([])
expect(errors).toHaveLength(1)
expect(errors[0]).toBeInstanceOf(LlmError)
expect((errors[0] as LlmError).failure).toEqual({
message: 'MockAdapter: script exhausted',
code: 'UNKNOWN',
})
expect(reasons[0]).toMatchObject({ kind: 'error' })
// The durable failure lives entirely on turn/end.reason (with the failing
// step), not a standalone error event.
// The durable failure and live relay describe the same failed turn.
const turnEnd = agent.session.events.find(e => e.type === 'turn/end')
expect(turnEnd?.type === 'turn/end' && turnEnd.data.reason).toMatchObject({ kind: 'error' })
})

View File

@@ -359,7 +359,12 @@ describe('request stability across the loop', () => {
await waitForIdle(ctx, agent)
expect(agent.session.events.findLast(event => event.type === 'turn/end')).toMatchObject({
data: { reason: { kind: 'error', error: failure.message } },
data: {
reason: {
kind: 'error',
error: failure instanceof LlmError ? failure.failure : failure.message,
},
},
})
expect(adapter.requests).toHaveLength(0)
},