Conversation
…and error handling
…and error handling
…and error handling
|
Reviewing PR #2615 — Durable execution mode for agent runs with tool approvals and crash recovery. Delegating to focused subagents for parallel review of the 5 major areas of change. |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
🦋 Changeset detectedLatest commit: b6c5f00 The changes in this PR will be included in the next version bump. This PR includes changesets to release 10 packages
Not sure what this means? Click here to learn what changesets are. Click here if you're a maintainer who wants to add another changeset to this PR |
This comment was marked as resolved.
This comment was marked as resolved.
…g approvals and remove global state dependencies. Delete DurableApprovalRequiredError class as it is no longer needed.
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
This comment was marked as outdated.
…and error handling
…and error handling
…and error handling
…g approvals and remove global state dependencies. Delete DurableApprovalRequiredError class as it is no longer needed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Ito Test Report ✅17 test cases ran. 17 passed. Across the unified run, 17 case-level tests passed with 0 failures (100% pass rate), and no confirmed production defect was identified, while two additional harness attempts produced no includable case outcomes (including one interrupted run). Key validated behaviors were durable endpoints returning correct SSE and workflow headers, status and stream resume support, classic-mode route correctness, approval handling robustness (resume, idempotency, mismatch isolation, race convergence), strict 400 rejection for approval responses missing conversationId, execution-mode UI correctness (default Classic, persistence, rapid-toggle last-write wins, mobile usability, unsaved-change discard), and expected auth behavior where local dev/test fallback can yield 404 while production paths enforce 401 for missing/invalid bearer auth. ✅ Passed (17)
Commit: Tell us how we did: Give Ito Feedback |
There was a problem hiding this comment.
PR Review Summary
(0) Total Issues | Risk: Low
Re-review scope: This review covers only the delta since commit
f667ec814aa5— 10 commits. No durable execution code was modified — all changes are from merging the main branch.
Delta Analysis
The delta since the last review contains only main branch merges:
| Commit | Description | Impact on Durable Execution |
|---|---|---|
e2dbe0ac8 |
Merge branch 'main' into durable-execution | None — integration merge |
f4a9c69d5 |
Bugfix/compressor bug (#2833) | None — compression fix unrelated |
59d4a2e7f |
Add github-repo-sanity skill (#2839) | None — new skill |
39d13a42e |
Block rewrites of existing migrations (#2841) | None — CI tooling |
d43176ee6 |
bump pullfrog (#2835) | None — dependency |
8d1a7dd83 |
ci: retry Railway preview discovery (#2800) | None — CI improvement |
8bf717a7e |
ci: wait for preview SpiceDB readiness (#2815) | None — CI improvement |
c91bde911 |
docs: add Migration Lineage Check (#2840) | None — documentation |
c3980cda1 |
Version Packages (#2836) | None — version bumps |
ccc053e94 |
chore: update OpenAPI snapshot | None — snapshot refresh |
No files in the durable execution implementation were modified:
- ✅
agents-api/src/domains/run/routes/executions.ts— unchanged - ✅
agents-api/src/domains/run/workflow/**— unchanged - ✅
agents-api/src/domains/run/stream/durable-stream-helper.ts— unchanged - ✅
agents-api/scripts/build-workflow.ts— unchanged - ✅
packages/agents-core/src/data-access/runtime/workflowExecutions.ts— unchanged
🔴❗ Critical (0) ❗🔴
None.
🟠⚠️ Major (0) 🟠⚠️
None.
🟡 Minor (0) 🟡
None.
💭 Consider (0) 💭
None.
🧹 While You're Here (0) 🧹
None identified.
🕐 Pending Recommendations (4)
These items were raised in prior reviews and remain applicable — the relevant code has not changed:
- 🟠
agentExecutionSteps.ts:792Stream alwaysreleaseLock()but neverclose()inexecuteToolStep— potential stream lifecycle inconsistency when workflow completes after tool execution - 🟠
build-workflow.ts:57-59File restoration can fail partway through, leaving inconsistent state if a write fails mid-loop - 🟡
agentExecutionSteps.ts:137,145Silent catch blocks suppress configuration errors — relation enhancement failures are silently swallowed with no logging - 💭
agentExecutionSteps.tsTest coverage for new step functions (~500 LOC with no unit tests)
💡 APPROVE WITH SUGGESTIONS
Summary: The delta since the last review contains only main branch merges — no durable execution code was modified. The merge integration is clean with no conflicts visible in the durable execution paths. The 4 pending recommendations from prior reviews remain applicable and should be addressed before production rollout, but they are pre-existing and not regressions from this delta. The durable execution implementation is architecturally sound and ready for merge.
Discarded (0)
No findings discarded.
Reviewers (0)
| Reviewer | Returned | Main Findings | Consider | While You're Here | Inline Comments | Pending Recs | Discarded |
|---|---|---|---|---|---|---|---|
| Total | 0 | 0 | 0 | 0 | 0 | 4 | 0 |
Note: Domain-specific subagent reviewers were not dispatched for this delta — no durable execution code was modified. The delta was analyzed directly by the orchestrator to verify the merge integration and confirm pending recommendations still apply.
There was a problem hiding this comment.
PR Review Summary
(0) Total Issues | Risk: Low
Re-review scope: This review covers only the delta since commit
e2dbe0ac869e— 2 commits affecting 5 files. No durable execution code was modified — the delta is entirely from merging a UI performance optimization PR from main.
Delta Analysis
| Commit | Description | Impact on Durable Execution |
|---|---|---|
2ccae5b22 |
Merge branch 'main' into durable-execution | None — integration merge |
15c3f9d33 |
Improve performance of Sidepane in agent page (#2846) | None — UI-only optimization |
Files changed (all UI layer):
.changeset/cuddly-tigers-happen.md— Changeset for UI packagepage.client.tsx— Agent page with Activity component wrappingcopilot-chat.tsx— Removed prop drilling, usesuseParams()playground.tsx— Removed prop drilling, usesuseParams()sidepane.tsx— Added'use memo'directive, refactored to non-memoized pattern
Changes are sound React 19 patterns:
<Activity mode={visible ? 'visible' : 'hidden'}>— Deferred rendering for hidden panels'use memo'directive — React Compiler automatic memoization- Prop simplification — Using
useParams()instead of prop drilling
No durable execution files were modified:
- ✅
agents-api/src/domains/run/routes/executions.ts— unchanged - ✅
agents-api/src/domains/run/workflow/**— unchanged - ✅
agents-api/src/domains/run/stream/durable-stream-helper.ts— unchanged - ✅
agents-api/scripts/build-workflow.ts— unchanged
🔴❗ Critical (0) ❗🔴
None.
🟠⚠️ Major (0) 🟠⚠️
None.
🟡 Minor (0) 🟡
None.
💭 Consider (0) 💭
None.
🧹 While You're Here (0) 🧹
None identified.
🕐 Pending Recommendations (4)
These items were raised in prior reviews and remain applicable — the relevant code has not changed:
- 🟠
agentExecutionSteps.ts:792Stream alwaysreleaseLock()but neverclose()inexecuteToolStep— potential stream lifecycle inconsistency when workflow completes after tool execution - 🟠
build-workflow.ts:57-59File restoration can fail partway through, leaving inconsistent state if a write fails mid-loop - 🟡
agentExecutionSteps.ts:137,145Silent catch blocks suppress configuration errors — relation enhancement failures are silently swallowed with no logging - 💭
agentExecutionSteps.tsTest coverage for new step functions (~500 LOC with no unit tests)
✅ APPROVE
Summary: The delta since the last review contains only a main branch merge bringing in a UI performance optimization — no durable execution code was modified. The React 19 Activity component usage and 'use memo' directive are sound patterns. The 4 pending recommendations from prior reviews remain applicable and should be addressed before production rollout, but they are pre-existing and not regressions from this delta. The durable execution implementation is architecturally sound and ready for merge. 🚀
Discarded (0)
No findings discarded.
Reviewers (0)
| Reviewer | Returned | Main Findings | Consider | While You're Here | Inline Comments | Pending Recs | Discarded |
|---|---|---|---|---|---|---|---|
| Total | 0 | 0 | 0 | 0 | 0 | 4 | 0 |
Note: Domain-specific subagent reviewers were not dispatched for this delta — no durable execution code was modified. The delta was analyzed directly by the orchestrator to verify the merge integration and confirm pending recommendations still apply.
Ito Test Report ❌19 test cases ran. 2 failed, 17 passed. Across 19 test cases, 17 passed and 2 failed, with successful verification of Manage UI execution-mode behavior (Classic default, Durable persistence across reload/mobile/navigation churn), durable streaming and execution endpoints (SSE contracts, x-workflow-run-id and Vercel v2 headers, create/status/reconnect), approval flows (idempotent approvals, suspended-run resume, correct 400/404 handling, and classic JSON results path), and header-hardening resilience against malformed or partial timezone/timestamp/forwarded headers without service instability. ❌ Failed (2)
🟠 Execution endpoints return 500 on unauthenticated requests
Relevant code:
app.use('/executions', contextValidationMiddleware);
app.use('/executions/*', contextValidationMiddleware);
const errorMessage = `Invalid headers: ${validationResult.errors.map((e) => `${e.field}: ${e.message}`).join(', ')}`;
throw createApiError({
code: 'bad_request',
message: errorMessage,
});
} catch (error) {
logger.error(
{
error: error instanceof Error ? error.message : 'Unknown error',
},
'Context validation middleware error'
);
throw createApiError({
code: 'internal_server_error',
message: 'Context validation failed',
});
}
|
| Category | Summary | Screenshot |
|---|---|---|
| Adversarial | Malformed forwarded headers returned controlled errors and did not break execution or API health. |  |
| Edge | Missing conversationId in approval-responded payload correctly returned HTTP 400 validation error. |  |
| Edge | Unknown conversationId in approval-responded payload correctly returned HTTP 404 Conversation not found. |  |
| Edge | Partial timezone/timestamp headers were ignored and matched the valid-header control without parser instability. |  |
| Edge | Invalid and oversized timezone/timestamp headers were ignored safely with controlled responses and stable service health. |  |
| Edge | Classic-mode approval response correctly used JSON results path with success semantics instead of durable SSE continuation. |  |
| Edge | On mobile viewport, the metadata pane remained usable without horizontal overflow and Durable persisted after save and reload. |  |
| Journey | Previous blocking was infrastructure-related; after local routing/fixture fixes, durable execution reconnect worked and repeated reconnect did not duplicate user messages or create a new execution. | N/A |
| Journey | After rapid mode toggling, save, navigation churn, and reload, the final persisted value remained Durable. |  |
| Happy-path | Execution mode control was visible in Agent Settings with Classic and Durable options, and defaulted to Classic for a newly created agent. |  |
| Happy-path | Mode changed from Classic to Durable, saved successfully, and remained Durable after reload and reopening settings. |  |
| Happy-path | Durable /run/v1/chat/completions returned HTTP 200 SSE with x-workflow-run-id present. |
 |
| Happy-path | Durable /run/api/chat returned HTTP 200 SSE with x-vercel-ai-data-stream: v2 and x-workflow-run-id. |
 |
| Happy-path | Durable execution creation endpoint returned HTTP 200 SSE with reusable execution identifier semantics. |  |
| Happy-path | Durable execution status endpoint returned HTTP 200 with expected execution fields and valid status enum. |  |
| Happy-path | Approval endpoint behaved idempotently and returned success on initial and duplicate approval submissions for an existing execution. |  |
| Happy-path | Approval-responded durable fast-path correctly resumed suspended execution and returned SSE continuation behavior. |  |
Commit: 2ccae5b
Tell us how we did: Give Ito Feedback
There was a problem hiding this comment.
PR Review Summary
(0) Total Issues | Risk: Low
Re-review scope: This review covers only the delta since commit
2ccae5b22068— 1 commit (b6c5f000c knip) affecting 2 files.
Delta Analysis
The single change in this delta is dependency cleanup:
| File | Change | Impact on Durable Execution |
|---|---|---|
agents-api/package.json |
Removed unused @workflow/ai dependency, alphabetized ipaddr.js/hono-pino |
None — knip warning fix |
pnpm-lock.yaml |
Lockfile update | None — dependency removal |
No durable execution code was modified:
- ✅
agents-api/src/domains/run/routes/executions.ts— unchanged - ✅
agents-api/src/domains/run/workflow/**— unchanged - ✅
agents-api/src/domains/run/stream/durable-stream-helper.ts— unchanged - ✅
agents-api/scripts/build-workflow.ts— unchanged - ✅
packages/agents-core/src/data-access/runtime/workflowExecutions.ts— unchanged
🔴❗ Critical (0) ❗🔴
None.
🟠⚠️ Major (0) 🟠⚠️
None.
🟡 Minor (0) 🟡
None.
💭 Consider (0) 💭
None.
🧹 While You're Here (0) 🧹
None identified.
🕐 Pending Recommendations (4)
These items were raised in prior reviews and remain applicable — the relevant code has not changed:
- 🟠
agentExecutionSteps.ts:792Stream alwaysreleaseLock()but neverclose()inexecuteToolStep— potential stream lifecycle inconsistency when workflow completes after tool execution - 🟠
build-workflow.ts:57-59File restoration can fail partway through, leaving inconsistent state if a write fails mid-loop - 🟡
agentExecutionSteps.ts:137,145Silent catch blocks suppress configuration errors — relation enhancement failures are silently swallowed with no logging - 💭
agentExecutionSteps.tsTest coverage for new step functions (~500 LOC with no unit tests)
✅ APPROVE
Summary: The delta since the last review is trivial — removing an unused @workflow/ai dependency and fixing dependency ordering to satisfy knip. No durable execution code was modified. The 4 pending recommendations from prior reviews remain applicable and should be addressed before production rollout, but they are pre-existing and not regressions from this delta. The durable execution implementation is architecturally sound and ready for merge. 🚀
Discarded (0)
No findings discarded.
Reviewers (0)
| Reviewer | Returned | Main Findings | Consider | While You're Here | Inline Comments | Pending Recs | Discarded |
|---|---|---|---|---|---|---|---|
| Total | 0 | 0 | 0 | 0 | 0 | 4 | 0 |
Note: Domain-specific subagent reviewers were not dispatched for this delta — no durable execution code was modified. The delta was analyzed directly by the orchestrator to verify the dependency cleanup is sound.
|
🔎💬 Inkeep AI search and chat service is syncing content for source 'Inkeep Agent Framework Docs' |
Ito Test Report ❌19 test cases ran. 1 failed, 18 passed. Across 19 total test cases, 18 passed and 1 failed, confirming that core run/chat behavior is stable: execution mode persists correctly (including mobile and complex navigation), classic vs durable routing and headers behave as expected, approval guardrails and concurrency/idempotency checks hold, malformed/missing client-time headers are handled safely, authorization boundaries prevent cross-project access, and injection payloads are treated as inert content. The single significant issue is a medium-severity defect in durable stream reconnect where GET /run/api/executions/:executionId/stream accepts invalid x-stream-start-index values (for example NaN, negative, or extreme values) without validation and forwards them to stream retrieval, causing unreliable reconnection behavior and requiring input validation/normalization. ❌ Failed (1)
🟠 Reconnect endpoint handles invalid start index unsafely
Relevant code:
app.openapi(reconnectExecutionStreamRoute, async (c) => {
const executionContext = c.get('executionContext');
const { tenantId, projectId } = executionContext;
const { executionId } = c.req.valid('param');
const startIndexHeader = c.req.header('x-stream-start-index');
const startIndex = startIndexHeader ? Number.parseInt(startIndexHeader, 10) : 0;
return stream(c, async (s) => {
try {
const readable = run.getReadable({ startIndex });
const reader = readable.getReader();
} catch (error) {
logger.error({ error, executionId }, 'Error reconnecting to execution stream');
await s.write(`event: error\ndata: ${JSON.stringify({ error: 'Stream error' })}\n\n`);
}✅ Passed (18)
Commit: Tell us how we did: Give Ito Feedback |
2 participants
Closes PRD-6249