Skip to content

test(campaign): follow retained attempt receipt paths #910

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
drewstone merged 1 commit into from
Oct 1, 2026
Merged

test(campaign): follow retained attempt receipt paths #910

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 13 additions & 3 deletions tests/campaign/presets.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,8 +1,9 @@
import { existsSync, mkdtempSync, rmSync } from 'node:fs'
import { existsSync, mkdtempSync, readFileSync, rmSync } from 'node:fs'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { makeProposalFinding } from '../../src/analyst/types'
import { cellAttemptDirectory } from '../../src/campaign/cell-schedule'
import {
buildLoopProvenanceRecord,
type CodeSurface,
Expand Down Expand Up @@ -462,8 +463,17 @@ describe('SurfaceProposer → runImprovementLoop → defaultProductionGate', ()
expect(h1Attempts).toBe(4)
for (const arm of ['holdout-baseline', 'holdout-winner']) {
const cellDir = join(exhaustedRunDir, arm, 'h1_0')
expect(existsSync(join(cellDir, 'failure-receipt.attempt-1.json'))).toBe(true)
expect(existsSync(join(cellDir, 'failure-receipt.json'))).toBe(true)
const attempt = JSON.parse(readFileSync(join(cellDir, 'latest-attempt.json'), 'utf8'))
for (const number of [1, 2]) {
expect(
existsSync(
join(
cellAttemptDirectory(cellDir, { runAttemptId: attempt.runAttemptId, number }),
'failure-receipt.json',
),
),
).toBe(true)
}
}
rmSync(exhaustedRunDir, { recursive: true, force: true })
})
Expand Down
63 changes: 36 additions & 27 deletions tests/campaign/run-campaign.test.ts
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,8 @@
import { existsSync, mkdtempSync, readdirSync, readFileSync, rmSync } from 'node:fs'
import { existsSync, mkdtempSync, readFileSync, rmSync } from 'node:fs'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
import { cellAttemptDirectory } from '../../src/campaign/cell-schedule'
import { campaignCoverage } from '../../src/campaign/coverage'
import {
type CampaignCellFailureReceipt,
Expand Down Expand Up @@ -50,6 +51,17 @@ const DISPATCH: DispatchFn<FakeScenario, FakeArtifact> = async (scenario, ctx) =
return paid.value
}

function failureReceiptPath(cellDir: string, number?: number): string {
const attempt = JSON.parse(readFileSync(join(cellDir, 'latest-attempt.json'), 'utf8'))
return join(
cellAttemptDirectory(cellDir, {
runAttemptId: attempt.runAttemptId,
number: number ?? attempt.number,
}),
'failure-receipt.json',
)
}

const SCENARIOS: FakeScenario[] = [
{ id: 'a', kind: 'chat', intent: 'help with X' },
{ id: 'b', kind: 'chat', intent: 'do Y' },
Expand Down Expand Up @@ -791,7 +803,7 @@ describe('runCampaign — core primitive', () => {
expect(ledger.list()[0]).toMatchObject({ phase: 'search.baseline', actor: 'worker' })

const failureReceipt = JSON.parse(
readFileSync(join(runDir, 'a_0', 'failure-receipt.json'), 'utf8'),
readFileSync(failureReceiptPath(join(runDir, 'a_0')), 'utf8'),
) as CampaignCellFailureReceipt<FakeArtifact>
const failedCallIds = ledger
.list({ tags: { scenarioId: 'a' } })
Expand Down Expand Up @@ -827,9 +839,7 @@ describe('runCampaign — core primitive', () => {
accountingComplete: true,
},
})
expect(result.artifactsByPath['a:0/failure-receipt.json']).toBe(
join(runDir, 'a_0', 'failure-receipt.json'),
)
expect(Object.values(result.artifactsByPath)).toContain(failureReceiptPath(join(runDir, 'a_0')))
})

it('atomically reserves capped calls without constraining free dispatches', async () => {
Expand Down Expand Up @@ -1052,7 +1062,7 @@ describe('runCampaign — core primitive', () => {
})

const failureReceipt = JSON.parse(
readFileSync(join(runDir, 'a_0', 'failure-receipt.json'), 'utf8'),
readFileSync(failureReceiptPath(join(runDir, 'a_0')), 'utf8'),
) as CampaignCellFailureReceipt<FakeArtifact>
expect(result.cells[0]?.error).toBe(
"judge 'paid-judge' failed: judge provider returned malformed output",
Expand Down Expand Up @@ -1091,12 +1101,6 @@ describe('runCampaign — core primitive', () => {
})
expect(failureReceipt.cost.totalCostUsd).toBeCloseTo(0.3, 9)
})

it('writes spans.jsonl per cell', async () => {
await runCampaign({ scenarios: SCENARIOS.slice(0, 1), dispatch: DISPATCH, runDir })
const cellDirs = readdirSync(runDir).filter((d) => d.startsWith('a_'))
expect(cellDirs.length).toBeGreaterThan(0)
})
})

describe('FsLabeledScenarioStore', () => {
Expand Down Expand Up @@ -1654,7 +1658,12 @@ describe('runCampaign — dispatchTimeoutMs (the no-silent-hang guard)', () => {
})
expect(result.aggregates.cost.totalCostUsd).toBe(0.25)
const failureReceipt = JSON.parse(
storage.read(join(lateCostRunDir, 'late-cost_0', 'failure-receipt.json'))!,
storage.read(
join(
cellAttemptDirectory(join(lateCostRunDir, 'late-cost_0'), result.cells[0]!.attempt!),
'failure-receipt.json',
),
)!,
) as CampaignCellFailureReceipt<FakeArtifact>
expect(failureReceipt).toMatchObject({
cell: {
Expand Down Expand Up @@ -1755,7 +1764,7 @@ describe('runCampaign — dispatchTimeoutMs (the no-silent-hang guard)', () => {
() => {
siblingAborted = true
failureReceiptExistedBeforeSiblingAbort = existsSync(
join(failFastRunDir, 'fails_0', 'failure-receipt.json'),
failureReceiptPath(join(failFastRunDir, 'fails_0')),
)
setTimeout(() => {
siblingStopped = true
Expand All @@ -1781,7 +1790,7 @@ describe('runCampaign — dispatchTimeoutMs (the no-silent-hang guard)', () => {
expect(failureReceiptExistedBeforeSiblingAbort).toBe(true)
expect(laterStarted).toBe(false)
expect(
JSON.parse(readFileSync(join(failFastRunDir, 'fails_0', 'failure-receipt.json'), 'utf8')),
JSON.parse(readFileSync(failureReceiptPath(join(failFastRunDir, 'fails_0')), 'utf8')),
).toMatchObject({
failure: {
stage: 'dispatch',
Expand All @@ -1791,7 +1800,7 @@ describe('runCampaign — dispatchTimeoutMs (the no-silent-hang guard)', () => {
cost: { pendingCalls: 0 },
})
expect(
JSON.parse(readFileSync(join(failFastRunDir, 'sibling_0', 'failure-receipt.json'), 'utf8')),
JSON.parse(readFileSync(failureReceiptPath(join(failFastRunDir, 'sibling_0')), 'utf8')),
).toMatchObject({
failure: {
stage: 'dispatch',
Expand Down Expand Up @@ -1835,7 +1844,7 @@ describe('runCampaign — dispatchTimeoutMs (the no-silent-hang guard)', () => {
expect(outcome).toEqual({ kind: 'rejected', error: judgeError })
expect(dispatches).toBe(1)
expect(
JSON.parse(readFileSync(join(judgeFailFastRunDir, 'a_0', 'failure-receipt.json'), 'utf8')),
JSON.parse(readFileSync(failureReceiptPath(join(judgeFailFastRunDir, 'a_0')), 'utf8')),
).toMatchObject({
failure: {
stage: 'judge',
Expand Down Expand Up @@ -1924,7 +1933,7 @@ describe('runCampaign — cellRetry (bounded in-run cell retry)', () => {
// The failed attempt keeps its own receipt, so the 503 stays auditable
// after the retry succeeded.
const attemptReceipt = JSON.parse(
readFileSync(join(runDir, 'flaky_0', 'failure-receipt.attempt-1.json'), 'utf8'),
readFileSync(failureReceiptPath(join(runDir, 'flaky_0'), 1), 'utf8'),
) as CampaignCellFailureReceipt<FakeArtifact>
expect(attemptReceipt).toMatchObject({
schemaVersion: 1,
Expand Down Expand Up @@ -1997,10 +2006,10 @@ describe('runCampaign — cellRetry (bounded in-run cell retry)', () => {
expect(cell.error).toContain('HTTP 503')
expect(cell.costUsd).toBeCloseTo(0.03, 9)
// Every attempt's spend stays auditable, not just the last one.
expect(existsSync(join(runDir, 'flaky_0', 'failure-receipt.attempt-1.json'))).toBe(true)
expect(existsSync(join(runDir, 'flaky_0', 'failure-receipt.attempt-2.json'))).toBe(true)
expect(existsSync(failureReceiptPath(join(runDir, 'flaky_0'), 1))).toBe(true)
expect(existsSync(failureReceiptPath(join(runDir, 'flaky_0'), 2))).toBe(true)
const finalReceipt = JSON.parse(
readFileSync(join(runDir, 'flaky_0', 'failure-receipt.json'), 'utf8'),
readFileSync(failureReceiptPath(join(runDir, 'flaky_0')), 'utf8'),
) as CampaignCellFailureReceipt<FakeArtifact>
expect(finalReceipt.cell.retryAttempts).toBe(2)
expect(finalReceipt.cost.totalCostUsd).toBeCloseTo(0.03, 9)
Expand Down Expand Up @@ -2033,8 +2042,8 @@ describe('runCampaign — cellRetry (bounded in-run cell retry)', () => {
expect(dispatches).toBe(1)
expect(result.cells[0]).toMatchObject({ errorStage: 'judge', errorJudge: 'unstable-judge' })
expect(result.cells[0]!.retryAttempts).toBeUndefined()
expect(existsSync(join(runDir, 'a_0', 'failure-receipt.json'))).toBe(true)
expect(existsSync(join(runDir, 'a_0', 'failure-receipt.attempt-1.json'))).toBe(false)
expect(existsSync(failureReceiptPath(join(runDir, 'a_0')))).toBe(true)
expect(existsSync(failureReceiptPath(join(runDir, 'a_0'), 2))).toBe(false)
})

it('abortOnCellError does not fire while a retryable failure has attempts left', async () => {
Expand Down Expand Up @@ -2078,10 +2087,10 @@ describe('runCampaign — cellRetry (bounded in-run cell retry)', () => {
expect(outcome).toEqual({ kind: 'rejected', error: attemptErrors[1] })
expect(calls).toBe(2)
expect(
JSON.parse(readFileSync(join(runDir, 'flaky_0', 'failure-receipt.attempt-1.json'), 'utf8')),
JSON.parse(readFileSync(failureReceiptPath(join(runDir, 'flaky_0'), 1), 'utf8')),
).toMatchObject({ failure: { error: { message: 'HTTP 503 first' } } })
expect(
JSON.parse(readFileSync(join(runDir, 'flaky_0', 'failure-receipt.json'), 'utf8')),
JSON.parse(readFileSync(failureReceiptPath(join(runDir, 'flaky_0')), 'utf8')),
).toMatchObject({ failure: { error: { message: 'HTTP 503 second' } } })
})

Expand All @@ -2103,7 +2112,7 @@ describe('runCampaign — cellRetry (bounded in-run cell retry)', () => {
expect(calls).toBe(1)
expect(result.cells[0]!.error).toBeDefined()
expect(result.cells[0]!.retryAttempts).toBeUndefined()
expect(existsSync(join(runDir, 'a_0', 'failure-receipt.json'))).toBe(true)
expect(existsSync(join(runDir, 'a_0', 'failure-receipt.attempt-1.json'))).toBe(false)
expect(existsSync(failureReceiptPath(join(runDir, 'a_0')))).toBe(true)
expect(existsSync(failureReceiptPath(join(runDir, 'a_0'), 2))).toBe(false)
})
})
Loading