Feat/work history faiss remember #7
Conversation
- Add OpenAI embeddings + FAISS-backed vector store with auto-migration from legacy JSON\n- Endpoints: /work-history (stats), /work-history/add, /work-history/import-path, /work-history/search, /work-history (DELETE), and /session/{id}/voice-transcript/memorize\n- Voice: inject top matches; keyword trigger ("remember that"); short spoken acknowledgement\n- UI: Remember button for current question\n- Logs: knowledge.* for import/add/search/memorize\n- Deps: faiss-cpu, numpy; config reads OPENAI_EMBEDDING_MODEL\n
…ranscripts with chosen name
…ext+voice; UI selector; session persistence
…helpful/discovery) with template files under app/prompts/
- replace the playwright testing plan with a helium-focused strategy
…ct Discovery in persona dropdown and JS state\n- Update voice session payload/log defaults to discovery\n\nfeat(server): set Discovery as default coach persona\n\n- New sessions saved with coach_persona=discovery\n- Fallbacks/use-sites now default to discovery for prompts, voice, and status\n\ntest(ui): add UI test asserting default persona\n\n- New Helium/Selenium test verifies #coach-persona defaults to discovery\n- Drives a fresh session via uploads to reach the interview view\n\nrefactor(agent): introduce InterviewAgentConfig + docstrings\n\n- Replace long __init__ arg list with dataclass config\n- Add docstrings and split persona prompt builders\n\ndocs: update README to note Discovery default\n\nchore: pytest.ini to scope tests and filter noisy warnings\n\nchore(log): docstrings for request logging middleware\n\nmeta: capture analyzer report after refactor
…ia .gitignore exception\n\n- Fixes runtime: ModuleNotFoundError: app.config when running uvicorn on this branch\n- Provides OpenAI, uploads, and knowledge store settings with safe defaults
…c in default ALLOWED_EXTENSIONS, matching server expectations\n- Update analyzer report after rerun
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| meta = dict(req.metadata or {}) | ||
| meta.update({ | ||
| "session_id": session_id, | ||
| "question_index": qidx, | ||
| "question": ((session.get("questions") or [None])[qidx] if (qidx is not None) else None), | ||
| "source": "voice_memorize", |
There was a problem hiding this comment.
The /session/{session_id}/voice-transcript/memorize endpoint indexes session["questions"][qidx] when building metadata. If the client supplies a question_index that has no corresponding entry (for example, memorizing before questions are generated or providing a custom transcript with a high index), the expression ((session.get("questions") or [None])[qidx]) raises IndexError, returning a 500 instead of a graceful 4xx response or None. This means a legitimate memorize request with valid text fails whenever the session lacks that index. Consider guarding the lookup with a length check or .get-style access so out-of-range indices do not crash the endpoint.
Useful? React with 👍 / 👎.
Merge Checklist
|
|
Updated base branch to Conflicting files (from local rebase/merge attempt):
Suggested resolution
Notes
|
No description provided.