fix: route iOS OpenAI realtime Talk through WebRTC#98563
Conversation
|
@clawsweeper re-review Current head: fdf6532 All currently reported checks are passing or intentionally skipped; PR body is updated with the focused transport-routing evidence and validation. |
|
🦞🧹 I asked ClawSweeper to review this item again. |
|
Codex review: needs changes before merge. Reviewed July 2, 2026, 3:40 PM ET / 19:40 UTC. Summary PR surface: Source +30, Tests +117, Docs +2, Other +725. Total +874 across 12 files. Reproducibility: yes. by source inspection and linked live evidence: current main resolves the iOS OpenAI Realtime selection to Gateway Relay, while the canonical issue reports the matching relay/session.error/Unknown Talk session failure. I did not run a local iPhone repro 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
Security Review findings
Review detailsBest possible solution: Rebase the branch to remove or align the superseded fixture hunk, then land the iOS WebRTC routing fix once maintainers accept the default-route compatibility and auth-provider ownership shift. Do we have a high-confidence way to reproduce the issue? Yes, by source inspection and linked live evidence: current main resolves the iOS OpenAI Realtime selection to Gateway Relay, while the canonical issue reports the matching relay/session.error/Unknown Talk session failure. I did not run a local iPhone repro in this read-only review. Is this the best way to solve the issue? Yes for the central bug: resolving native WebRTC in the iOS routing layer is the right owner boundary because the Gateway already exposes client-owned WebRTC sessions. The branch still needs a conflict repair and maintainer acceptance of the default-route compatibility shift before merge. Full review comments:
Overall correctness: patch is incorrect AGENTS.md: found and applied where relevant. Codex review notes: model internal, reasoning high; reviewed against a004e18b4bbd. Label changesLabel changes:
Label justifications:
Evidence reviewedPR surface: Source +30, Tests +117, Docs +2, Other +725. Total +874 across 12 files. View PR surface stats
Acceptance criteria:
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
|
|
Proof media/link comment removed after merge for privacy. Validation evidence was summarized in the sanitized PR body. |
1 similar comment
|
Proof media/link comment removed after merge for privacy. Validation evidence was summarized in the sanitized PR body. |
8c183d4 to
cb26f4a
Compare
cb26f4a to
6c199bb
Compare
fc19506 to
9486162
Compare
9486162 to
f20ed11
Compare
|
Merged via squash.
|
Related: #91007
What Problem This Solves
Fixes an issue where iOS users trying to use OpenAI Realtime Talk could see the app enter the Gateway Relay path even though the Gateway advertised a WebRTC realtime configuration.
Why This Change Was Made
The Gateway already carries the requested realtime transport. This change makes the iOS Talk route honor OpenAI Realtime configs with
transport: "webrtc"by using the existing native WebRTC session path instead of mapping them to Gateway Relay.User Impact
iOS Talk can use the intended Native WebRTC path for OpenAI Realtime configurations that request WebRTC, while existing relay/provider-websocket behavior remains available for configurations that explicitly need it.
Evidence
059023e7800350bd59d922b81465c5004f90509d; branch-wide review and hosted/local validation were completed before merge.TalkModeManagerTests: 32 passed.TalkConfigParsingTests: 7 passed.git diff --check,pnpm native:i18n:check, docs formatting, MDX syntax check, and docs link audit passed.AI-assisted: yes. The implementation and review were done with Codex, with local review findings addressed before publication.