fix(browser): recognize Chinese Deep Research controls

[oraclexing]

Recognize Simplified Chinese Deep Research menu labels, the selected-tool pill, and the add-files fallback. Preserve the existing compact English DeepResearch label and keep selection scoped to the visible popover so a neighboring connector is not selected.

Maintainer repair preserves the contributor history and moves release-note ownership to the final notes PR. The implementation has no CHANGELOG.md delta against main.

Validation at ce5ca1c14aa8e1ef202905902315210ff96c8c5d:

• 62 focused tests and the full 2,233-test suite passed (45 skipped); build, formatting, lint/typecheck, and docs checks passed.
• Signed-in built CLI on a real zh-CN ChatGPT page selected 深度研究, submitted one public-source research prompt, and reached the research-plan/researching state. The original Auto-detect language preference was restored.
• The full research report did not complete within 40 minutes. This live evidence proves localized activation and submission; it does not claim completed report capture.
• Local and branch P0–P2 autoreview are clean.
• CI: https://github.com/steipete/oracle/actions/runs/34175639564 — all four jobs succeeded on this head.

Deferred changelog:

Browser: recognize Chinese Deep Research menu, selected-tool, and add-files controls while preserving compact English labels; thanks @oraclexing.

[clawsweeper]

🦞👀
ClawSweeper picked this up.

Pull request received. I will update this pull request when review starts.

ClawSweeper review complete

ClawSweeper finished reviewing this revision. The review result is being finalized.

View the workflow run.

[clawsweeper]

Codex review: needs maintainer review before merge. Reviewed September 7, 2026, 9:53 PM ET / September 8, 2026, 01:53 UTC.

ClawSweeper review

What this changes

Recognizes Simplified Chinese Deep Research menu, selected-tool, and add-files controls, preserves compact English labels, and adds activation tests and documentation.

Merge readiness

✅ Ready for maintainer review

This PR remains useful and has no identified blocking defect. Current main lacks Chinese activation matching, and the maintainer’s signed-in evidence covers the changed behavior. The earlier changelog concern is resolved.

Priority: P2
Reviewed head: ce5ca1c14aa8e1ef202905902315210ff96c8c5d

Review scores

Measure	Result	What it means
Overall readiness	🐚 platinum hermit (4/6)	A focused repair with relevant signed-in proof, English regression coverage, and no remaining actionable finding.
Proof confidence	🦞 diamond lobster (5/6)	Sufficient (live_output): The captured maintainer observations tie the changed activation owner to the built CLI on a signed-in zh-CN page: Chinese tool selection, one committed prompt, and a researching plan were observed. The 40-minute timeout leaves final report capture unproven, which is outside this activation change.
Patch quality	🐚 platinum hermit (4/6)	No actionable review findings were identified.

Verification

Check	Result	Evidence
Real behavior	Verified	Sufficient (live_output): The captured maintainer observations tie the changed activation owner to the built CLI on a signed-in zh-CN page: Chinese tool selection, one committed prompt, and a researching plan were observed. The 40-minute timeout leaves final report capture unproven, which is outside this activation change.
Evidence reviewed	8 items	Applicable repository policy: Read the full root AGENTS.md; no nested AGENTS.md files were found. Browser verification guidance informed the proof assessment. The owner's explicit deferred-release-note decision resolves the prior changelog feedback. Current main still needs localization: Main's activation expression recognizes English menu and pill labels only; its shared constants are also English. The existing production path therefore does not already supply the proposed Chinese matching. Introduced behavior and regression coverage: The complete pinned introduction delta changes four files. Activation retains visible-popover filtering and pill confirmation; three executable expression cases cover Chinese, compact English, and descriptive English rows while checking that the neighboring GitHub connector receives no click.
Findings	None	None.
Security	None	None.

How this fits together

Oracle’s browser engine selects ChatGPT’s Deep Research tool before submitting a research prompt. This change updates how it recognizes the menu entry and verifies the selected tool.

flowchart LR
  A[Research request] --> B[Open composer tools]
  B --> C[Match English or Chinese label]
  C --> D[Select Deep Research]
  D --> E[Verify selected tool]
  E --> F[Submit research prompt]

Loading

Before merge

None.

Agent review details

Security

None.

Review metrics

Metric	Value	Why it matters
Production and test delta	production +35/-19 (net +16); tests +106/-2 (net +104)	The small production increase supports localized matching, with three activation cases covering Chinese and existing English labels.

Technical review

Best possible solution:

Keep localized recognition in the existing activation path, preserving English matching and selected-tool verification without changing user settings.

Do we have a high-confidence way to reproduce the issue?

Yes, from source: current main cannot match the documented Chinese menu and pill labels. This review did not execute a failing main run.

Is this the best way to solve the issue?

Yes. Extending the existing label matching is a focused repair, and the signed-in observations exercise activation and submission through the production CLI.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning medium; reviewed against fe2c8d990562.

Labels

Label justifications:

• P2: Repairs a bounded localization failure in the optional Deep Research browser workflow.
• rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
• status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (live_output): The captured maintainer observations tie the changed activation owner to the built CLI on a signed-in zh-CN page: Chinese tool selection, one committed prompt, and a researching plan were observed. The 40-minute timeout leaves final report capture unproven, which is outside this activation change.
• proof: sufficient: Contributor real behavior proof is sufficient. The captured maintainer observations tie the changed activation owner to the built CLI on a signed-in zh-CN page: Chinese tool selection, one committed prompt, and a researching plan were observed. The 40-minute timeout leaves final report capture unproven, which is outside this activation change.

Evidence

What I checked:

• Applicable repository policy: Read the full root AGENTS.md; no nested AGENTS.md files were found. Browser verification guidance informed the proof assessment. The owner's explicit deferred-release-note decision resolves the prior changelog feedback. (AGENTS.md, ce5ca1c14aa8)
• Current main still needs localization: Main's activation expression recognizes English menu and pill labels only; its shared constants are also English. The existing production path therefore does not already supply the proposed Chinese matching. (src/browser/actions/deepResearch.ts:1131, fe2c8d990562)
• Introduced behavior and regression coverage: The complete pinned introduction delta changes four files. Activation retains visible-popover filtering and pill confirmation; three executable expression cases cover Chinese, compact English, and descriptive English rows while checking that the neighboring GitHub connector receives no click. (tests/browser/deepResearch.test.ts:190, ce5ca1c14aa8)
• Production entrypoint: Both browser execution paths call the changed activation owner before the existing prompt-submission flow. Report completion and capture are separate, unchanged stages. (src/browser/index.ts:1829, ce5ca1c14aa8)
• Signed-in behavior proof: The captured owner comment fix(browser): recognize Chinese Deep Research controls #428 (comment) reports the built CLI on this head selecting 深度研究 on a real zh-CN page, committing one public-source prompt, and reaching the researching plan. Auto-detect was restored. The report timed out after 40 minutes, so final capture is explicitly unproven. Captured context sourceRevision: 88f89eb318b1b516f61de6e1fb9827d16f62a45b8a500447ddc65409a26b7540. (ce5ca1c14aa8)
• Earlier finding resolved: The final branch has no CHANGELOG.md delta against main. The owner explicitly assigns the credited localization entry to the final notes PR, superseding the earlier request to move the entry into Unreleased. (CHANGELOG.md, ce5ca1c14aa8)

Likely related people:

• steipete: Suggested for follow-up; no historical authorship or introduction is verified. (role: unverified routing candidate; confidence: low)
• wbzjt: Suggested for follow-up; no historical authorship or introduction is verified. (role: unverified routing candidate; confidence: low)

Rating scale

Score	Internal tier	Crab rank	Meaning
6/6	S	🦀 challenger crab	Exceptional readiness
5/6	A	🦞 diamond lobster	Very strong readiness
4/6	B	🐚 platinum hermit	Good normal PR; ordinary maintainer review
3/6	C	🦐 gold shrimp	Useful, but confidence is limited
2/6	D	🦪 silver shellfish	Proof or implementation needs work
1/6	F	🧂 unranked krab	Not merge-ready
N/A	NA	🌊 off-meta tidepool	Rating does not apply

Overall follows the weaker of proof and patch quality.
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.

Workflow

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

History

Review history (6 earlier review cycles)

• reviewed 2026-08-28T10:58:17.574Z sha bb5ca9a :: needs maintainer review before merge. :: none
• reviewed 2026-09-02T02:57:53.954Z sha c7fe620 :: needs real behavior proof before merge. :: none
• reviewed 2026-09-02T03:46:48.305Z sha c7fe620 :: needs maintainer review before merge. :: none
• reviewed 2026-09-08T00:00:31.692Z sha 7f58349 :: needs changes before merge. :: [P3] Move the localization entry into Unreleased
• reviewed 2026-09-08T00:10:45.311Z sha 7f58349 :: blocked before merge. :: [P3] Move the localization entry into Unreleased
• reviewed 2026-09-08T01:12:06.699Z sha ce5ca1c :: needs maintainer review before merge. :: none

[oraclexing]

Rebased onto current main and updated the PR body with exact-head full-suite results plus signed-in activation evidence. The proof distinguishes Deep Research activation from the separate legacy send-path commit failure tracked in #431.

@clawsweeper re-review

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event exact_review_queue).
Result: when the review finishes, ClawSweeper will create the durable review comment if needed or update the existing comment in place.

[oraclexing]

Added real signed-in Simplified Chinese proof on exact head c7fe620adb9d78971ec36a9f7a3aba86ad093bc9: live menu row 深度研究 获取详细报告, production activateDeepResearch click capture, localized selected pill 深度研究, Chat on / Work off, and zero user turns. The original Auto-detect language setting was restored and the dedicated Chrome/DevTools resources were closed. Full DOM evidence and screenshot hashes are in the PR body.

@clawsweeper re-review

[clawsweeper]

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event exact_review_queue).
Result: when the review finishes, ClawSweeper will create the durable review comment if needed or update the existing comment in place.

[steipete]

Maintainer proof for ce5ca1c14aa8e1ef202905902315210ff96c8c5d:

• Exact-head CI is green: https://github.com/steipete/oracle/actions/runs/34175639564
• Local full suite: 2,233 passed / 45 skipped; 62 focused Deep Research tests, build, lint/typecheck, formatting, and docs checks passed.
• Built CLI with --engine browser --browser-research deep --browser-attach-running --browser-approval-wait 5m activated the real Simplified Chinese control. DOM evidence showed lang=zh-CN, the 深度研究 selection, one committed synthetic/public-source prompt, and the researching plan. Auto-detect was restored afterward.
• Research report completion timed out at 40 minutes; activation and submission are proved, final report capture is not claimed.
• Both local and committed-branch P0–P2 autoreview are clean. The maintainer repair also restores compact English DeepResearch matching and removes the incorrectly released changelog line; the deferred entry belongs in the final notes PR.