test: execute media and Talk runtime boundaries#101091
Conversation
|
@clawsweeper re-review |
|
🦞🧹 I asked ClawSweeper to review this item again. |
|
Codex review: needs maintainer review before merge. Reviewed July 6, 2026, 5:30 PM ET / 21:30 UTC. Summary PR surface: Source +118, Tests +536, Other +15. Total +669 across 21 files. Reproducibility: yes. at source level: current main still points the affected QA scenarios at canaries or injected boundaries, and source inspection shows the Gateway Origin, Talk acknowledgement, and mock Voice Call realtime blockers. I did not run the runtime QA lane in this read-only review. Review metrics: 1 noteworthy metric.
Stored data model Root-cause cluster Members:
Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything. Merge readiness Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch. Rank-up moves:
Risk before merge
Maintainer options:
Next step before merge
Maintainer decision needed
Security Review detailsBest possible solution: Land this PR after maintainer review and exact-head gates if maintainers accept the combined QA/runtime-support shape; otherwise ask for a split before merging the linked issue's implementation path. Do we have a high-confidence way to reproduce the issue? Yes at source level: current main still points the affected QA scenarios at canaries or injected boundaries, and source inspection shows the Gateway Origin, Talk acknowledgement, and mock Voice Call realtime blockers. I did not run the runtime QA lane in this read-only review. Is this the best way to solve the issue? Yes, this appears to be the best fix shape: it reuses the real Gateway child, plugin runtime, QA evidence writer, and existing browser Talk path instead of adding a parallel harness. The small runtime support changes are directly exercised by the new producers. AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning high; reviewed against c7295e417d5d. Label changesLabel justifications:
Evidence reviewedPR surface: Source +118, Tests +536, Other +15. Total +669 across 21 files. View PR surface stats
What I checked:
Likely related people:
What the crustacean ranks mean
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics. How this review workflow works
|
|
@clawsweeper re-review |
|
🦞🧹 I asked ClawSweeper to review this item again. |
|
@clawsweeper re-review |
|
🦞🧹 I asked ClawSweeper to review this item again. |
d940331 to
198512e
Compare
|
@clawsweeper re-review |
|
🦞🧹 I asked ClawSweeper to review this item again. |
198512e to
71d95d6
Compare
|
@clawsweeper re-review |
|
🦞🧹 I asked ClawSweeper to review this item again. |
* test: execute media and Talk runtime boundaries * test(qa): centralize gateway fixture startup * test(qa): check in voice call plugin fixture
* test: execute media and Talk runtime boundaries * test(qa): centralize gateway fixture startup * test(qa): check in voice call plugin fixture
Closes #101074
What Problem This Solves
Resolves a problem where QA scenarios for WebChat audio delivery, browser Talk lifecycle, active Talk run control, and Voice Call entry points could pass without exercising the runtime boundaries named by their coverage claims.
Why This Change Was Made
This replaces shallow canaries with shared script producers that start the real Gateway and plugin runtime while mocking only external model, speech, realtime, and telephony providers. It reuses the existing Playwright browser Talk test, allows Node Gateway clients to send an explicit WebSocket Origin, returns realtime Talk consult acknowledgement without waiting for the full agent run, and permits the built-in mock Voice Call provider to drive realtime media streams.
User Impact
Maintainers get executable evidence that WebChat TTS reaches scoped media tickets, browser Talk starts and releases audio resources, active Talk RPCs control one real embedded agent run, and Voice Call CLI/RPC/tool entry points share a fake-provider runtime whose webhook media stream carries embedded-agent consult context. The Talk acknowledgement fix also prevents realtime tool calls from blocking until the consulted agent run finishes.
Evidence
tbx_01kww5h9qz0xb9rangk2pqnhsw(pearl-hermit), Actions run28808350733.64182mson Linux Nodev24.13.0.voicecall.initiate,tools.invoke, built-in mock telephony, runtime-issued webhook media stream, and embedded consult transcript/provider context passed.ui/src/e2e/browser-talk-start-stop.e2e.test.ts: Playwright browser start/stop passed.packages/gateway-client/src/client.watchdog.test.ts: 15 passed.src/gateway/server-methods/talk.test.ts: 92 passed.extensions/voice-call/src/config.test.ts: 32 passed.check:changedpassed typecheck, lint, import-cycle, and repository policy gates.git diff --checkpassed.