fix: GPT session re-auth crash + cleaner ImplementAndJudge card

[elkaix]

Summary

Two independent, small fixes (one commit each).

1. fix(core): GPT/ChatGPT mid-session credential crash → typed 401

When a ChatGPT (openai_codex) session token is invalidated mid-session — classically because the same OAuth credentials were reused on another machine, rotating the single-use refresh token so this machine's refresh is rejected — OAuthManager._apply_access_token(runtime, ref, "") blanks the live OpenAI client's api_key to "". The client was already constructed with a valid token, so this slips past the SDK's construction-time credential check and the next responses.create raises a bare TypeError from _validate_headers ("Could not resolve authentication method…"). That TypeError is neither OpenAIError nor httpx.HTTPError, so it escaped the provider and the soul's 401/connection recovery and surfaced as a fatal "Unexpected error".

The provider now converts that unresolved-credential TypeError into a typed APIStatusError(401) (gated on not self._client.api_key, so unrelated TypeErrors still propagate). APIStatusError subclasses ChatProviderError, so it routes into the existing refresh → re-auth path and the shell's "Type /login to re-authenticate" message instead of crashing.

Reproduced end-to-end (identical TypeError); fix verified for both stream=True and stream=False (the SDK raises at request-build time, so generate() catches both).

Out of scope (noted, not changed here): _apply_access_token(runtime, ref, "") discards a possibly-still-valid access token just because the refresh token died — a separate token-lifecycle change.

2. fix(tui): cleaner ImplementAndJudge card

The ImplementAndJudge chain tool had no dedicated renderer, so it fell back to the generic ImplementAndJudge(4 args: acceptance, base_prompt, brief, scope) arg dump. Added a dedicated renderer:

⏺ Implement & Judge — Rewrite Logo.astro GSAP: matchMedia + scoped context

Internal tool name stays ImplementAndJudge (registry/config/tests unchanged); only the on-screen label changes. The judge subagent name was already consistent across specs/profiles/docs — no naming changes needed.

Tests

• packages/pythinker-core/tests/test_openai_responses_auth.py: blanked-credential → APIStatusError(401) for both stream modes; negative case (TypeError with non-empty key still propagates).
• tests/ui_and_conv/test_tui_card_tool_renderers.py: clean label + brief, no arg dump; running state; a name-sync invariant so a rename can't silently fall back to the generic renderer.

Verification

make check-pythinker-core ✅ · make check-pythinker-code ✅ (ruff + format + pyright + ty) · pythinker-core suite 305 passed · TUI renderer suite 157 passed.

Summary by CodeRabbit

Release Notes

• Bug Fixes

  • Fixed crash when GPT/ChatGPT session tokens expire; users are now redirected to login instead of encountering an error.

• UI Improvements

  • Enhanced "Implement & Judge" chain display to show a cleaner format with brief descriptions instead of raw argument details.

[coderabbitai]

📝 Walkthrough

Walkthrough

Two independent fixes: OpenAIResponses.generate() now catches TypeError caused by a blank api_key and converts it to a typed APIStatusError(401), routing expired session tokens into the standard re-auth flow. Separately, a new implement_judge TUI renderer displays ImplementAndJudge chains as "⏺ Implement & Judge — <brief>" with a collapsible result block instead of a raw argument dump.

Changes

OpenAI Session Token 401 Fix

Layer / File(s)	Summary
TypeError-to-401 conversion in OpenAIResponses.generate() packages/pythinker-core/src/pythinker_core/contrib/chat_provider/openai_responses.py, packages/pythinker-core/tests/test_openai_responses_auth.py	Imports APIStatusError, adds a TypeError catch that raises APIStatusError(401) when api_key is blank, re-raises otherwise. Tests cover both streaming/non-streaming 401 conversion and the pass-through of unrelated TypeErrors.

ImplementAndJudge TUI Renderer

Layer / File(s)	Summary
Renderer implementation and wiring src/pythinker_code/ui/shell/tool_renderers/implement_judge.py, src/pythinker_code/ui/shell/tool_renderers/__init__.py, tests/ui_and_conv/test_tui_card_tool_renderers.py	New module implements _compact(), _render_call(), _label_header(), _render_result(), and exports IMPLEMENT_JUDGE_RENDERER. The __init__.py lazy-imports and registers it. Three smoke tests assert friendly label rendering in completed and running states, and name-constant consistency.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Suggested labels

bug

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 21.43% which is insufficient. The required threshold is 70.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title follows conventional commits format with valid 'fix' type and clearly describes both fixes: GPT re-auth crash and cleaner ImplementAndJudge card display.
Description check	✅ Passed	Description includes detailed explanation of both fixes, reproduction/verification steps, test coverage, and verification results. All required template sections are present and complete.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/gpt-reauth-crash-and-iaj-card

---

_{Comment @coderabbitai help to get the list of available commands.}

[codecov]

Codecov Report

❌ Patch coverage is 74.07407% with 14 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...er_code/ui/shell/tool_renderers/implement_judge.py	72.00%	10 Missing and 4 partials ⚠️

📢 Thoughts on this report? Let us know!

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@packages/pythinker-core/tests/test_openai_responses_auth.py`:
- Around line 20-21: Add the `@pytest.mark.asyncio` marker to both async test
functions in the file. The functions `test_blanked_credential_raises_typed_401`
and `test_typeerror_with_valid_key_propagates` are defined as `async def` but
lack the required marker. Since the project does not configure `asyncio_mode =
"auto"` in pytest configuration, these async functions will not be properly
awaited by pytest and will silently pass without executing their assertions.
Place `@pytest.mark.asyncio` on a new line immediately before each async
function definition, above any other decorators like `@pytest.mark.parametrize`.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 82f5d8a7-c1fd-42eb-b680-d00470dd2341

📥 Commits

Reviewing files that changed from the base of the PR and between 9677c57 and 2d4b1b0.

📒 Files selected for processing (6)

• CHANGELOG.md
• packages/pythinker-core/src/pythinker_core/contrib/chat_provider/openai_responses.py
• packages/pythinker-core/tests/test_openai_responses_auth.py
• src/pythinker_code/ui/shell/tool_renderers/__init__.py
• src/pythinker_code/ui/shell/tool_renderers/implement_judge.py
• tests/ui_and_conv/test_tui_card_tool_renderers.py

[elkaix]

All findings triaged on head 2d4b1b0:

• Inline (asyncio markers): skipped — root pytest.ini already sets asyncio_mode = auto; matches existing pythinker-core async test convention; CI green.
• Docstring coverage warning: skipped — CodeRabbit advisory only; not enforced by repo CI/lint gates; renderer follows existing tool-renderer docstring pattern.
• Codecov patch note: skipped — codecov/patch check passed.

No code changes required.