feat: Added personality_weight (0.0–1.0) to chat API; modulates system…
- "backend/routers/chat.py" - "backend/chat_service.py" - "backend/tests/test_chat.py" GSD-Task: S02/T01
This commit is contained in:
jlightner
parent
9431aa2095
commit
d1efdbb3fa
13 changed files with 1010 additions and 4 deletions
|
|
@ -362,3 +362,15 @@
|
|||
**Context:** Extracting a creator personality profile from transcripts. Small creators have <5 videos, large creators have 20+. Sampling strategy must adapt to corpus size while ensuring topic diversity.
|
||||
|
||||
**Pattern:** 3 tiers: small (≤5 videos: use all), medium (6-15: sample ~8), large (>15: sample ~10). For medium/large, use Redis classification data to group transcripts by topic and sample proportionally, ensuring the profile captures the creator's full range rather than overrepresenting one topic area.
|
||||
|
||||
## Tiptap v3 useEditor requires immediatelyRender: false for React 18
|
||||
|
||||
**Context:** Tiptap v3's `useEditor` hook by default attempts synchronous rendering on mount, which conflicts with React 18's concurrent features (StrictMode double-mounting, suspense boundaries). This causes hydration mismatches and "flushSync was called from inside a lifecycle method" warnings.
|
||||
|
||||
**Fix:** Pass `immediatelyRender: false` in the useEditor config: `useEditor({ immediatelyRender: false, extensions: [...], content: ... })`. The editor still renders on the first paint — this just defers it to be React 18 compatible. No visual difference.
|
||||
|
||||
## get_optional_user pattern for public-but-auth-aware endpoints
|
||||
|
||||
**Context:** Some API endpoints need to be publicly accessible but provide different behavior for authenticated users (e.g., showing draft posts to the owner). Using the standard `get_current_user` dependency rejects unauthenticated requests with 401.
|
||||
|
||||
**Fix:** Create `get_optional_user` using `OAuth2PasswordBearer(auto_error=False)`. When `auto_error=False`, missing/invalid tokens return `None` instead of raising 401. The endpoint receives `Optional[User]` and branches on whether the user is present and matches the resource owner.
|
||||
|
|
|
|||
|
|
@ -6,7 +6,7 @@ The demo MVP comes together. Chat widget wires to the intelligence layer (INT-1)
|
|||
## Slice Overview
|
||||
| ID | Slice | Risk | Depends | Done | After this |
|
||||
|----|-------|------|---------|------|------------|
|
||||
| S01 | [A] Post Editor + File Sharing | high | — | ⬜ | Creator writes rich text posts with file attachments (presets, sample packs). Followers see posts in feed. Files downloadable via signed URLs. |
|
||||
| S01 | [A] Post Editor + File Sharing | high | — | ✅ | Creator writes rich text posts with file attachments (presets, sample packs). Followers see posts in feed. Files downloadable via signed URLs. |
|
||||
| S02 | [A] Chat Widget ↔ Chat Engine Wiring (INT-1) | high | — | ⬜ | Chat widget on creator profile wired to chat engine. Personality slider adjusts response style. Citations link to sources. |
|
||||
| S03 | [B] Shorts Generation Pipeline v1 | medium | — | ⬜ | Shorts pipeline extracts clips from highlight boundaries in 3 format presets (vertical, square, horizontal) |
|
||||
| S04 | [B] Personality Slider (Full Interpolation) | medium | — | ⬜ | Personality slider at 0.0 gives encyclopedic response. At 1.0 gives creator-voiced response with their speech patterns. |
|
||||
|
|
|
|||
151
.gsd/milestones/M023/slices/S01/S01-SUMMARY.md
Normal file
151
.gsd/milestones/M023/slices/S01/S01-SUMMARY.md
Normal file
|
|
@ -0,0 +1,151 @@
|
|||
---
|
||||
id: S01
|
||||
parent: M023
|
||||
milestone: M023
|
||||
provides:
|
||||
- Post CRUD API at /api/v1/posts
|
||||
- File upload/download API at /api/v1/files
|
||||
- MinIO object storage integration
|
||||
- PostsFeed component for creator profile pages
|
||||
- PostEditor page for creator content authoring
|
||||
- PostsList management page for creators
|
||||
requires:
|
||||
[]
|
||||
affects:
|
||||
- S05
|
||||
key_files:
|
||||
- docker-compose.yml
|
||||
- backend/minio_client.py
|
||||
- backend/routers/posts.py
|
||||
- backend/routers/files.py
|
||||
- backend/models.py
|
||||
- backend/schemas.py
|
||||
- backend/auth.py
|
||||
- backend/main.py
|
||||
- alembic/versions/024_add_posts_and_attachments.py
|
||||
- frontend/src/pages/PostEditor.tsx
|
||||
- frontend/src/pages/PostEditor.module.css
|
||||
- frontend/src/api/posts.ts
|
||||
- frontend/src/api/client.ts
|
||||
- frontend/src/components/PostsFeed.tsx
|
||||
- frontend/src/components/PostsFeed.module.css
|
||||
- frontend/src/pages/PostsList.tsx
|
||||
- frontend/src/pages/PostsList.module.css
|
||||
- frontend/src/pages/CreatorDetail.tsx
|
||||
- frontend/src/App.tsx
|
||||
- frontend/src/pages/CreatorDashboard.tsx
|
||||
- docker/nginx.conf
|
||||
- backend/config.py
|
||||
- backend/requirements.txt
|
||||
key_decisions:
|
||||
- MinIO internal-only (no public port) — files served via API-generated presigned URLs
|
||||
- get_optional_user dependency for public endpoints with auth-aware behavior
|
||||
- MinIO object deletion on post delete is best-effort (logged, non-blocking)
|
||||
- Tiptap v3 with StarterKit + Link + Placeholder for rich text editing
|
||||
- PostsFeed hides section entirely when creator has zero posts
|
||||
- File uploads use run_in_executor for sync MinIO I/O in async handlers
|
||||
patterns_established:
|
||||
- MinIO integration pattern: lazy-init singleton, ensure_bucket on first write, presigned URLs for downloads
|
||||
- requestMultipart API client helper for FormData uploads without Content-Type header
|
||||
- get_optional_user auth dependency for public-but-auth-aware endpoints
|
||||
- Tiptap JSON→HTML rendering with generateHTML + extensions for read-only display
|
||||
observability_surfaces:
|
||||
- File upload errors logged with MinIO error message and object_key
|
||||
- Post ownership violations logged at WARNING
|
||||
- MinIO bucket auto-creation logged on startup
|
||||
- 503 status for MinIO failures, 403 for ownership violations, 404 for missing resources
|
||||
drill_down_paths:
|
||||
- .gsd/milestones/M023/slices/S01/tasks/T01-SUMMARY.md
|
||||
- .gsd/milestones/M023/slices/S01/tasks/T02-SUMMARY.md
|
||||
- .gsd/milestones/M023/slices/S01/tasks/T03-SUMMARY.md
|
||||
- .gsd/milestones/M023/slices/S01/tasks/T04-SUMMARY.md
|
||||
duration: ""
|
||||
verification_result: passed
|
||||
completed_at: 2026-04-04T09:19:26.082Z
|
||||
blocker_discovered: false
|
||||
---
|
||||
|
||||
# S01: [A] Post Editor + File Sharing
|
||||
|
||||
**Full post editor with Tiptap rich text, MinIO-backed file attachments, CRUD API, public feed rendering, and creator management page.**
|
||||
|
||||
## What Happened
|
||||
|
||||
This slice delivered the complete post system — write path, read path, and file storage.
|
||||
|
||||
**T01 (Data Layer):** Added MinIO Docker service (internal-only, no public port) to docker-compose.yml with healthcheck and volume mount. Created minio_client.py with lazy-init singleton, ensure_bucket, upload_file, delete_file, and presigned URL generation. Added Post and PostAttachment SQLAlchemy models with UUID PKs and cascade delete. Created Alembic migration 024. Added Pydantic schemas for all CRUD operations. Bumped nginx client_max_body_size to 100m.
|
||||
|
||||
**T02 (API Layer):** Built posts.py router with 5 CRUD endpoints enforcing creator ownership via auth. Introduced get_optional_user dependency (OAuth2PasswordBearer with auto_error=False) for public list endpoints that show drafts only to owners. Built files.py router with multipart upload (run_in_executor for sync MinIO I/O) and signed download URL generation. Registered both routers and added MinIO bucket auto-creation in the app lifespan handler.
|
||||
|
||||
**T03 (Post Editor):** Installed Tiptap v3 with StarterKit, Link, and Placeholder extensions. Built PostEditor page with formatting toolbar, title input, drag-and-drop file attachment zone, publish toggle, and sequential save flow (create post → upload files). Added requestMultipart helper to the API client for FormData uploads. Lazy-loaded routes at /creator/posts/new and /creator/posts/:postId/edit.
|
||||
|
||||
**T04 (Public Feed + Management):** Created PostsFeed component rendering published posts with Tiptap JSON→HTML conversion and signed-URL download buttons. Integrated into CreatorDetail page (hidden when no posts). Built PostsList management page with status badges, edit/delete actions, and confirmation dialog. Added SidebarNav link and route at /creator/posts.
|
||||
|
||||
## Verification
|
||||
|
||||
All slice verification checks passed:
|
||||
1. `docker compose config --quiet` — exit 0
|
||||
2. All backend imports (Post, PostAttachment, PostCreate, PostRead, PostAttachmentRead, PostListResponse, get_minio_client, generate_download_url, upload_file, ensure_bucket, get_optional_user, posts router, files router) — exit 0
|
||||
3. `cd frontend && npm run build` — exit 0, 179 modules, PostEditor and PostsList lazy-loaded as separate chunks
|
||||
4. Route registration verified: 8 new API routes (5 posts CRUD + 2 files + bucket init), 3 frontend routes (/creator/posts, /creator/posts/new, /creator/posts/:postId/edit)
|
||||
|
||||
## Requirements Advanced
|
||||
|
||||
None.
|
||||
|
||||
## Requirements Validated
|
||||
|
||||
None.
|
||||
|
||||
## New Requirements Surfaced
|
||||
|
||||
None.
|
||||
|
||||
## Requirements Invalidated or Re-scoped
|
||||
|
||||
None.
|
||||
|
||||
## Deviations
|
||||
|
||||
- Added get_optional_user auth dependency (not in plan, needed for public list with draft visibility)
|
||||
- Added delete_file() to minio_client.py (needed for post deletion cleanup)
|
||||
- Tiptap v3 required immediatelyRender: false for React 18 compatibility
|
||||
- SidebarNav links to Posts management page instead of direct New Post link (better UX with management page available)
|
||||
|
||||
## Known Limitations
|
||||
|
||||
- Alembic migration 024 not yet applied to live database (deploy-time)
|
||||
- PostsFeed fetches all posts for a creator in one call (no pagination on public profile)
|
||||
- File upload reads entire file into memory before MinIO write
|
||||
|
||||
## Follow-ups
|
||||
|
||||
- Apply migration 024 on ub01 deploy
|
||||
- Add pagination to PostsFeed if post counts grow
|
||||
- Consider streaming file uploads for very large files
|
||||
|
||||
## Files Created/Modified
|
||||
|
||||
- `docker-compose.yml` — Added MinIO service with healthcheck, volume, internal network
|
||||
- `backend/config.py` — Added MinIO connection settings
|
||||
- `backend/minio_client.py` — Created — lazy-init MinIO client, ensure_bucket, upload, delete, presigned URL
|
||||
- `backend/models.py` — Added Post and PostAttachment models with FKs and cascade
|
||||
- `backend/schemas.py` — Added post CRUD and attachment Pydantic schemas
|
||||
- `backend/requirements.txt` — Added minio package
|
||||
- `backend/auth.py` — Added get_optional_user dependency
|
||||
- `backend/routers/posts.py` — Created — 5 CRUD endpoints with auth and ownership
|
||||
- `backend/routers/files.py` — Created — multipart upload and signed download URL
|
||||
- `backend/main.py` — Registered posts and files routers, added MinIO bucket init
|
||||
- `alembic/versions/024_add_posts_and_attachments.py` — Created — posts and post_attachments tables
|
||||
- `docker/nginx.conf` — Bumped client_max_body_size to 100m
|
||||
- `frontend/src/api/client.ts` — Added requestMultipart helper
|
||||
- `frontend/src/api/posts.ts` — Created — TypeScript types and API functions for posts
|
||||
- `frontend/src/pages/PostEditor.tsx` — Created — Tiptap editor with toolbar, file attachments, save flow
|
||||
- `frontend/src/pages/PostEditor.module.css` — Created — dark theme editor styles
|
||||
- `frontend/src/components/PostsFeed.tsx` — Created — public feed with HTML rendering and downloads
|
||||
- `frontend/src/components/PostsFeed.module.css` — Created — feed card styles
|
||||
- `frontend/src/pages/PostsList.tsx` — Created — creator post management page
|
||||
- `frontend/src/pages/PostsList.module.css` — Created — management page styles
|
||||
- `frontend/src/pages/CreatorDetail.tsx` — Integrated PostsFeed component
|
||||
- `frontend/src/App.tsx` — Added PostEditor, PostsList lazy imports and routes
|
||||
- `frontend/src/pages/CreatorDashboard.tsx` — Added Posts link to SidebarNav
|
||||
84
.gsd/milestones/M023/slices/S01/S01-UAT.md
Normal file
84
.gsd/milestones/M023/slices/S01/S01-UAT.md
Normal file
|
|
@ -0,0 +1,84 @@
|
|||
# S01: [A] Post Editor + File Sharing — UAT
|
||||
|
||||
**Milestone:** M023
|
||||
**Written:** 2026-04-04T09:19:26.083Z
|
||||
|
||||
## UAT: S01 — Post Editor + File Sharing
|
||||
|
||||
### Preconditions
|
||||
- Chrysopedia stack running on ub01 (docker compose up -d)
|
||||
- Migration 024 applied (alembic upgrade head)
|
||||
- MinIO container healthy (docker ps shows chrysopedia-minio healthy)
|
||||
- At least one creator account with login credentials
|
||||
- A test file (any .zip or .wav, 1-50MB)
|
||||
|
||||
---
|
||||
|
||||
### TC-01: Post Creation (Happy Path)
|
||||
1. Log in as creator at /login
|
||||
2. Navigate to /creator/posts via sidebar → "Posts" link
|
||||
3. **Expected:** PostsList page renders, shows "New Post" button
|
||||
4. Click "New Post"
|
||||
5. **Expected:** PostEditor renders with title input, Tiptap editor with toolbar, file attachment area, publish toggle
|
||||
6. Enter title: "Test Post — Sample Pack Release"
|
||||
7. Type rich text in editor: a paragraph, then bold text, then a bullet list
|
||||
8. **Expected:** Toolbar buttons (B, I, H2, H3, list, link, code) toggle formatting visually
|
||||
9. Toggle "Published" on
|
||||
10. Attach a file by clicking the file attachment area and selecting the test file
|
||||
11. **Expected:** File appears in attachment list below editor with filename and remove button
|
||||
12. Click "Save"
|
||||
13. **Expected:** Post created, files uploaded, redirects to creator dashboard or posts list
|
||||
|
||||
### TC-02: Post Listing (Creator Management)
|
||||
1. Navigate to /creator/posts
|
||||
2. **Expected:** PostsList shows the post from TC-01 with title, "Published" badge, created date, Edit and Delete buttons
|
||||
3. **Expected:** "New Post" button visible at top
|
||||
|
||||
### TC-03: Post Editing
|
||||
1. From PostsList, click "Edit" on the TC-01 post
|
||||
2. **Expected:** PostEditor loads with existing title, body content, and attached file
|
||||
3. Change the title to "Updated — Sample Pack v2"
|
||||
4. Click "Save"
|
||||
5. **Expected:** Post updates successfully
|
||||
|
||||
### TC-04: Public Feed Display
|
||||
1. Log out (or open incognito)
|
||||
2. Navigate to the creator's public profile page (/creators/{slug})
|
||||
3. **Expected:** "Posts" section appears below technique grid
|
||||
4. **Expected:** Post card shows title, rendered rich text (bold, lists visible), timestamp
|
||||
5. **Expected:** File attachment shows with download button (filename, size)
|
||||
|
||||
### TC-05: File Download via Signed URL
|
||||
1. On the public creator profile, click the download button on the post attachment
|
||||
2. **Expected:** Browser initiates file download (MinIO presigned URL)
|
||||
3. **Expected:** Downloaded file matches the original upload
|
||||
|
||||
### TC-06: Draft Visibility
|
||||
1. Log in as creator, create a new post, leave "Published" toggled OFF, save
|
||||
2. Navigate to /creator/posts
|
||||
3. **Expected:** New draft post shows with "Draft" badge
|
||||
4. Log out, visit creator's public profile
|
||||
5. **Expected:** Draft post is NOT visible in the PostsFeed
|
||||
|
||||
### TC-07: Post Deletion
|
||||
1. Log in as creator, navigate to /creator/posts
|
||||
2. Click "Delete" on a post
|
||||
3. **Expected:** Confirmation dialog appears
|
||||
4. Confirm deletion
|
||||
5. **Expected:** Post removed from list, no longer visible on public profile
|
||||
|
||||
### TC-08: Ownership Enforcement
|
||||
1. Log in as a different user (not the post creator)
|
||||
2. Attempt PUT /api/v1/posts/{post_id} via curl with a valid auth token for the wrong user
|
||||
3. **Expected:** 403 Forbidden response
|
||||
4. Attempt DELETE /api/v1/posts/{post_id} with wrong user token
|
||||
5. **Expected:** 403 Forbidden response
|
||||
6. Attempt POST /api/v1/files/upload with wrong user's post_id
|
||||
7. **Expected:** 403 Forbidden response
|
||||
|
||||
### TC-09: Edge Cases
|
||||
1. Create a post with no attachments → save succeeds
|
||||
2. Create a post with empty body → save succeeds (title-only post)
|
||||
3. Upload a file with special characters in filename (e.g., "my file (v2).zip") → filename sanitized, upload succeeds
|
||||
4. Visit public profile for creator with zero posts → Posts section hidden (no empty state)
|
||||
5. Attempt GET /api/v1/posts/nonexistent-uuid → 404 response
|
||||
16
.gsd/milestones/M023/slices/S01/tasks/T04-VERIFY.json
Normal file
16
.gsd/milestones/M023/slices/S01/tasks/T04-VERIFY.json
Normal file
|
|
@ -0,0 +1,16 @@
|
|||
{
|
||||
"schemaVersion": 1,
|
||||
"taskId": "T04",
|
||||
"unitId": "M023/S01/T04",
|
||||
"timestamp": 1775294250736,
|
||||
"passed": true,
|
||||
"discoverySource": "task-plan",
|
||||
"checks": [
|
||||
{
|
||||
"command": "cd frontend",
|
||||
"exitCode": 0,
|
||||
"durationMs": 6,
|
||||
"verdict": "pass"
|
||||
}
|
||||
]
|
||||
}
|
||||
|
|
@ -1,6 +1,75 @@
|
|||
# S02: [A] Chat Widget ↔ Chat Engine Wiring (INT-1)
|
||||
|
||||
**Goal:** Wire chat widget UI to chat engine via creator-scoped router with personality system. INT-1 complete.
|
||||
**Goal:** Chat widget on creator profile sends personality_weight (0.0–1.0) to the chat engine. Backend modulates the system prompt: 0.0 = encyclopedic, >0 = creator voice injected from personality_profile JSONB. Slider UI visible in chat panel header.
|
||||
**Demo:** After this: Chat widget on creator profile wired to chat engine. Personality slider adjusts response style. Citations link to sources.
|
||||
|
||||
## Tasks
|
||||
- [x] **T01: Added personality_weight (0.0–1.0) to chat API; modulates system prompt with creator voice profile and scales LLM temperature** — Thread personality_weight through the chat API and modulate the system prompt based on creator personality profile.
|
||||
|
||||
## Failure Modes
|
||||
|
||||
| Dependency | On error | On timeout | On malformed response |
|
||||
|---|---|---|---|
|
||||
| Creator DB query | Log warning, fall back to encyclopedic prompt | Same — DB timeout treated as missing profile | N/A — query returns model or None |
|
||||
|
||||
## Negative Tests
|
||||
|
||||
- **Malformed inputs**: personality_weight outside 0.0–1.0 → 422, personality_weight as string → 422
|
||||
- **Error paths**: creator name doesn't match any DB record → encyclopedic fallback, personality_profile is null → encyclopedic fallback
|
||||
- **Boundary conditions**: weight exactly 0.0 → no profile query needed, weight exactly 1.0 → full personality injection
|
||||
|
||||
## Steps
|
||||
|
||||
1. In `backend/routers/chat.py`, add `personality_weight: float = Field(default=0.0, ge=0.0, le=1.0)` to `ChatRequest`. Pass it to `service.stream_response()`.
|
||||
2. In `backend/chat_service.py`, add `personality_weight: float = 0.0` param to `stream_response()`. When weight > 0 and creator is not None:
|
||||
- Import `Creator` from models, `select` from sqlalchemy
|
||||
- Query `select(Creator).where(Creator.name == creator)` using the existing `db` session
|
||||
- If creator found and `personality_profile` is not None, build a personality injection block from the profile dict (extract signature_phrases, tone descriptors, teaching_style, energy, formality from the nested vocabulary/tone/style_markers structure)
|
||||
- Append the personality block to the system prompt: 'Respond in {creator}'s voice. {profile_summary}. Use their signature phrases: {phrases}. Match their {formality} {energy} tone.'
|
||||
- If creator not found or profile is null, log at DEBUG and use the standard encyclopedic prompt
|
||||
3. Scale LLM temperature with weight: `temperature = 0.3 + (personality_weight * 0.2)` (range 0.3–0.5).
|
||||
4. Extend the `logger.info('chat_request ...')` line to include `weight=body.personality_weight`.
|
||||
5. Add tests to `backend/tests/test_chat.py`:
|
||||
- Test personality_weight is accepted and forwarded (mock verifies stream_response called with weight)
|
||||
- Test system prompt includes personality context when weight > 0 and profile exists (mock Creator query, capture messages sent to OpenAI)
|
||||
- Test encyclopedic fallback when weight > 0 but personality_profile is null
|
||||
- Test 422 for personality_weight outside [0.0, 1.0]
|
||||
6. Run `cd backend && python -m pytest tests/test_chat.py -v` — all tests pass.
|
||||
|
||||
## Must-Haves
|
||||
|
||||
- [ ] ChatRequest schema has personality_weight with ge=0.0, le=1.0, default=0.0
|
||||
- [ ] stream_response queries Creator.personality_profile when weight > 0
|
||||
- [ ] Null profile or missing creator falls back to encyclopedic prompt silently
|
||||
- [ ] Temperature scales with weight (0.3 at 0.0, up to 0.5 at 1.0)
|
||||
- [ ] All existing tests still pass
|
||||
- [ ] New tests for weight forwarding, prompt injection, null fallback, validation
|
||||
- Estimate: 1h
|
||||
- Files: backend/routers/chat.py, backend/chat_service.py, backend/tests/test_chat.py
|
||||
- Verify: cd backend && python -m pytest tests/test_chat.py -v
|
||||
- [ ] **T02: Frontend — personality slider UI + API wiring in ChatWidget** — Add a personality weight slider to ChatWidget and thread it through the streamChat() API call.
|
||||
|
||||
## Steps
|
||||
|
||||
1. In `frontend/src/api/chat.ts`, add `personalityWeight?: number` parameter to `streamChat()` function signature (after `conversationId`). Include `personality_weight: personalityWeight ?? 0` in the JSON body.
|
||||
2. In `frontend/src/components/ChatWidget.tsx`:
|
||||
- Add `const [personalityWeight, setPersonalityWeight] = useState(0);` state
|
||||
- In the panel header (the `<div className={styles.header}>` block), add a slider row below the title: `<div className={styles.sliderRow}>` containing a label span ('Encyclopedic' on left, 'Creator Voice' on right), and an `<input type='range' min={0} max={1} step={0.1}>` bound to personalityWeight state
|
||||
- Pass `personalityWeight` to `streamChat()` in the `sendMessage` callback (the 5th argument)
|
||||
3. In `frontend/src/components/ChatWidget.module.css`, add styles:
|
||||
- `.sliderRow` — flex row with gap, padding, align-items center, subtle border-bottom to separate from messages
|
||||
- `.sliderLabel` — small muted text (var(--text-secondary) or similar)
|
||||
- `.slider` — custom range input styling matching the dark theme (cyan accent track/thumb matching existing brand colors)
|
||||
4. Run `cd frontend && npm run build` — zero errors.
|
||||
|
||||
## Must-Haves
|
||||
|
||||
- [ ] streamChat() accepts and sends personalityWeight in POST body
|
||||
- [ ] ChatWidget renders a range slider in the panel header
|
||||
- [ ] Slider labels show 'Encyclopedic' at 0 and 'Creator Voice' at 1
|
||||
- [ ] Slider value flows through to API call
|
||||
- [ ] Styles match existing dark theme
|
||||
- [ ] Frontend builds with zero TypeScript/compilation errors
|
||||
- Estimate: 45m
|
||||
- Files: frontend/src/api/chat.ts, frontend/src/components/ChatWidget.tsx, frontend/src/components/ChatWidget.module.css
|
||||
- Verify: cd frontend && npm run build
|
||||
|
|
|
|||
89
.gsd/milestones/M023/slices/S02/S02-RESEARCH.md
Normal file
89
.gsd/milestones/M023/slices/S02/S02-RESEARCH.md
Normal file
|
|
@ -0,0 +1,89 @@
|
|||
# S02 Research: Chat Widget ↔ Chat Engine Wiring (INT-1)
|
||||
|
||||
## Summary
|
||||
|
||||
Most of the chat plumbing already exists and works. The ChatWidget component (`frontend/src/components/ChatWidget.tsx`) renders a floating bubble on the CreatorDetail page, streams SSE responses via `frontend/src/api/chat.ts`, and renders citations with links. The backend (`backend/routers/chat.py` → `backend/chat_service.py`) accepts a query + optional creator name, retrieves context via SearchService, streams an LLM completion with citations, and supports multi-turn memory via Redis.
|
||||
|
||||
**What's missing is the personality slider.** The current system prompt is pure encyclopedic — no creator voice injection. The `personality_weight` parameter doesn't exist anywhere in the stack. S04 owns "full interpolation" (the sophisticated blending logic), but S02 needs to:
|
||||
|
||||
1. Add a slider UI to ChatWidget
|
||||
2. Thread `personality_weight` (0.0–1.0) through the API
|
||||
3. Implement basic system prompt variation: at weight 0.0 use the existing encyclopedic prompt; at weight > 0 inject the creator's personality profile data into the system prompt to flavor the response
|
||||
|
||||
## Recommendation
|
||||
|
||||
**Targeted research depth.** The tech is all familiar — React slider input, FastAPI field addition, system prompt templating. The integration seam is clean: one new field flows frontend → API → chat_service → system prompt.
|
||||
|
||||
## Implementation Landscape
|
||||
|
||||
### What Exists
|
||||
|
||||
| Component | File | Status |
|
||||
|---|---|---|
|
||||
| Chat bubble + panel UI | `frontend/src/components/ChatWidget.tsx` (491 lines) | ✅ Complete — renders on CreatorDetail, has streaming, citations, suggestions, conversation memory |
|
||||
| Chat SSE client | `frontend/src/api/chat.ts` | ✅ Complete — POST with `{query, creator, conversation_id}`, parses sources/token/done/error events |
|
||||
| Chat API route | `backend/routers/chat.py` | ✅ Complete — `POST /api/v1/chat` accepts `ChatRequest(query, creator, conversation_id)` |
|
||||
| Chat service | `backend/chat_service.py` | ✅ Complete — search → build context → stream LLM. Has Redis-backed multi-turn. System prompt is hardcoded encyclopedic |
|
||||
| Chat tests | `backend/tests/test_chat.py` | ✅ Complete — 11 tests covering SSE format, citations, creator forwarding, validation, errors, conversation memory |
|
||||
| Personality profile model | `backend/models.py:134` — `Creator.personality_profile` JSONB | ✅ Extracted and stored per creator |
|
||||
| Personality profile schema | `backend/schemas.py:766` — `PersonalityProfile` with `VocabularyProfile`, `ToneProfile`, `StyleMarkersProfile` | ✅ |
|
||||
| Personality profile display | `frontend/src/components/PersonalityProfile.tsx` | ✅ Collapsible section on CreatorDetail |
|
||||
| ChatPage (standalone) | `frontend/src/pages/ChatPage.tsx` | ✅ No creator context — no personality slider needed here |
|
||||
| CSS module | `frontend/src/components/ChatWidget.module.css` (388 lines) | ✅ Dark theme, styled |
|
||||
|
||||
### What Needs Building
|
||||
|
||||
1. **Slider UI in ChatWidget** — `<input type="range">` with 0.0–1.0 range, styled label ("Encyclopedic ↔ Creator Voice"), positioned in the chat panel header or above the input row. Only shown when the widget is creator-scoped (always true for ChatWidget, since it only appears on CreatorDetail).
|
||||
|
||||
2. **`personality_weight` in API contract** — Add to `ChatRequest` schema (`float`, 0.0–1.0, default 0.0), `streamChat()` TS function, and the fetch body.
|
||||
|
||||
3. **System prompt interpolation in ChatService** — When `personality_weight > 0` and a creator is specified:
|
||||
- Query `Creator.personality_profile` from DB (the `db: AsyncSession` is already passed to `stream_response`)
|
||||
- Build a personality injection block from the profile (signature phrases, tone descriptors, teaching style, etc.)
|
||||
- Append it to the system prompt with instructions like "Respond in {creator}'s voice: {profile_summary}. Use their signature phrases: {phrases}. Match their {formality} tone and {energy} energy."
|
||||
- Scale the temperature slightly with the weight (0.3 at 0.0, up to ~0.5 at 1.0)
|
||||
|
||||
4. **Tests** — Add test cases for personality_weight parameter forwarding, system prompt injection when weight > 0, and behavior when personality_profile is null.
|
||||
|
||||
### Key Files to Modify
|
||||
|
||||
| File | Changes |
|
||||
|---|---|
|
||||
| `backend/routers/chat.py` | Add `personality_weight: float = 0.0` to `ChatRequest`, pass to service |
|
||||
| `backend/chat_service.py` | Accept `personality_weight`, query Creator model for profile, build voice-injected prompt variant |
|
||||
| `frontend/src/api/chat.ts` | Add `personalityWeight` param to `streamChat()`, include in POST body |
|
||||
| `frontend/src/components/ChatWidget.tsx` | Add slider state, pass to `streamChat()`, render slider UI |
|
||||
| `frontend/src/components/ChatWidget.module.css` | Styles for slider row |
|
||||
| `backend/tests/test_chat.py` | New tests for personality_weight behavior |
|
||||
|
||||
### Natural Task Seams
|
||||
|
||||
**T01: Backend — personality_weight API + prompt interpolation**
|
||||
- Add field to `ChatRequest`
|
||||
- Modify `ChatService.stream_response()` to accept weight, query Creator, build voice prompt
|
||||
- Add/update tests
|
||||
- Verify: `pytest backend/tests/test_chat.py`
|
||||
|
||||
**T02: Frontend — slider UI + API wiring**
|
||||
- Add slider component to ChatWidget header
|
||||
- Thread personality_weight through `streamChat()`
|
||||
- Style the slider
|
||||
- Verify: `npm run build` (type-check + bundle)
|
||||
|
||||
### Constraints & Risks
|
||||
|
||||
- **Creator lookup by name, not ID:** The current API accepts `creator: str` (name string). To fetch the personality profile, the service needs to query `Creator` by name/slug. The `db` session is already available in `stream_response()`. Query: `select(Creator).where(Creator.name == creator)`.
|
||||
|
||||
- **Null personality profiles:** Not all creators have extracted profiles. When `personality_weight > 0` but `personality_profile` is null, fall back to the encyclopedic prompt silently. Don't error.
|
||||
|
||||
- **S04 boundary:** S02 builds the plumbing and a basic two-mode prompt (encyclopedic vs voice-injected). S04 does the sophisticated interpolation (blending between the two at continuous weight values, speech pattern matching, etc.). S02 should use a simple approach: weight == 0 → pure encyclopedic; weight > 0 → append personality context to the prompt. This is enough for the demo.
|
||||
|
||||
- **ChatPage (standalone) doesn't need a slider** — it has no creator context. No changes needed there.
|
||||
|
||||
- **Existing test pattern:** Tests use a standalone ASGI client with mocked DB session and Redis. The personality_weight tests should follow the same pattern — mock the Creator query result.
|
||||
|
||||
### Verification Strategy
|
||||
|
||||
- `cd backend && python -m pytest tests/test_chat.py -v` — all existing tests pass + new personality tests
|
||||
- `cd frontend && npm run build` — zero TypeScript/build errors
|
||||
- Manual: open CreatorDetail page, move slider, send chat message, observe personality-flavored response vs encyclopedic response
|
||||
66
.gsd/milestones/M023/slices/S02/tasks/T01-PLAN.md
Normal file
66
.gsd/milestones/M023/slices/S02/tasks/T01-PLAN.md
Normal file
|
|
@ -0,0 +1,66 @@
|
|||
---
|
||||
estimated_steps: 32
|
||||
estimated_files: 3
|
||||
skills_used: []
|
||||
---
|
||||
|
||||
# T01: Backend — personality_weight API field + system prompt modulation + tests
|
||||
|
||||
Thread personality_weight through the chat API and modulate the system prompt based on creator personality profile.
|
||||
|
||||
## Failure Modes
|
||||
|
||||
| Dependency | On error | On timeout | On malformed response |
|
||||
|---|---|---|---|
|
||||
| Creator DB query | Log warning, fall back to encyclopedic prompt | Same — DB timeout treated as missing profile | N/A — query returns model or None |
|
||||
|
||||
## Negative Tests
|
||||
|
||||
- **Malformed inputs**: personality_weight outside 0.0–1.0 → 422, personality_weight as string → 422
|
||||
- **Error paths**: creator name doesn't match any DB record → encyclopedic fallback, personality_profile is null → encyclopedic fallback
|
||||
- **Boundary conditions**: weight exactly 0.0 → no profile query needed, weight exactly 1.0 → full personality injection
|
||||
|
||||
## Steps
|
||||
|
||||
1. In `backend/routers/chat.py`, add `personality_weight: float = Field(default=0.0, ge=0.0, le=1.0)` to `ChatRequest`. Pass it to `service.stream_response()`.
|
||||
2. In `backend/chat_service.py`, add `personality_weight: float = 0.0` param to `stream_response()`. When weight > 0 and creator is not None:
|
||||
- Import `Creator` from models, `select` from sqlalchemy
|
||||
- Query `select(Creator).where(Creator.name == creator)` using the existing `db` session
|
||||
- If creator found and `personality_profile` is not None, build a personality injection block from the profile dict (extract signature_phrases, tone descriptors, teaching_style, energy, formality from the nested vocabulary/tone/style_markers structure)
|
||||
- Append the personality block to the system prompt: 'Respond in {creator}'s voice. {profile_summary}. Use their signature phrases: {phrases}. Match their {formality} {energy} tone.'
|
||||
- If creator not found or profile is null, log at DEBUG and use the standard encyclopedic prompt
|
||||
3. Scale LLM temperature with weight: `temperature = 0.3 + (personality_weight * 0.2)` (range 0.3–0.5).
|
||||
4. Extend the `logger.info('chat_request ...')` line to include `weight=body.personality_weight`.
|
||||
5. Add tests to `backend/tests/test_chat.py`:
|
||||
- Test personality_weight is accepted and forwarded (mock verifies stream_response called with weight)
|
||||
- Test system prompt includes personality context when weight > 0 and profile exists (mock Creator query, capture messages sent to OpenAI)
|
||||
- Test encyclopedic fallback when weight > 0 but personality_profile is null
|
||||
- Test 422 for personality_weight outside [0.0, 1.0]
|
||||
6. Run `cd backend && python -m pytest tests/test_chat.py -v` — all tests pass.
|
||||
|
||||
## Must-Haves
|
||||
|
||||
- [ ] ChatRequest schema has personality_weight with ge=0.0, le=1.0, default=0.0
|
||||
- [ ] stream_response queries Creator.personality_profile when weight > 0
|
||||
- [ ] Null profile or missing creator falls back to encyclopedic prompt silently
|
||||
- [ ] Temperature scales with weight (0.3 at 0.0, up to 0.5 at 1.0)
|
||||
- [ ] All existing tests still pass
|
||||
- [ ] New tests for weight forwarding, prompt injection, null fallback, validation
|
||||
|
||||
## Inputs
|
||||
|
||||
- ``backend/routers/chat.py` — existing ChatRequest schema and route handler`
|
||||
- ``backend/chat_service.py` — existing ChatService with stream_response() and _SYSTEM_PROMPT_TEMPLATE`
|
||||
- ``backend/tests/test_chat.py` — existing 11 tests with standalone ASGI client pattern`
|
||||
- ``backend/models.py` — Creator model with personality_profile JSONB column`
|
||||
- ``backend/schemas.py` — PersonalityProfile, VocabularyProfile, ToneProfile, StyleMarkersProfile schemas`
|
||||
|
||||
## Expected Output
|
||||
|
||||
- ``backend/routers/chat.py` — ChatRequest with personality_weight field, passed to service`
|
||||
- ``backend/chat_service.py` — stream_response accepts personality_weight, queries Creator, builds voice prompt`
|
||||
- ``backend/tests/test_chat.py` — new tests for personality_weight behavior (forwarding, prompt injection, null fallback, validation)`
|
||||
|
||||
## Verification
|
||||
|
||||
cd backend && python -m pytest tests/test_chat.py -v
|
||||
79
.gsd/milestones/M023/slices/S02/tasks/T01-SUMMARY.md
Normal file
79
.gsd/milestones/M023/slices/S02/tasks/T01-SUMMARY.md
Normal file
|
|
@ -0,0 +1,79 @@
|
|||
---
|
||||
id: T01
|
||||
parent: S02
|
||||
milestone: M023
|
||||
provides: []
|
||||
requires: []
|
||||
affects: []
|
||||
key_files: ["backend/routers/chat.py", "backend/chat_service.py", "backend/tests/test_chat.py"]
|
||||
key_decisions: ["Personality injection intensity varies with weight: <0.4 subtle, 0.4-0.8 voice, >=0.8 full embody", "Temperature range 0.3-0.5 keeps responses grounded while allowing expressiveness"]
|
||||
patterns_established: []
|
||||
drill_down_paths: []
|
||||
observability_surfaces: []
|
||||
duration: ""
|
||||
verification_result: "cd backend && python -m pytest tests/test_chat.py -v — 22 passed (13 existing + 9 new), 0 failed"
|
||||
completed_at: 2026-04-04T09:28:21.555Z
|
||||
blocker_discovered: false
|
||||
---
|
||||
|
||||
# T01: Added personality_weight (0.0–1.0) to chat API; modulates system prompt with creator voice profile and scales LLM temperature
|
||||
|
||||
> Added personality_weight (0.0–1.0) to chat API; modulates system prompt with creator voice profile and scales LLM temperature
|
||||
|
||||
## What Happened
|
||||
---
|
||||
id: T01
|
||||
parent: S02
|
||||
milestone: M023
|
||||
key_files:
|
||||
- backend/routers/chat.py
|
||||
- backend/chat_service.py
|
||||
- backend/tests/test_chat.py
|
||||
key_decisions:
|
||||
- Personality injection intensity varies with weight: <0.4 subtle, 0.4-0.8 voice, >=0.8 full embody
|
||||
- Temperature range 0.3-0.5 keeps responses grounded while allowing expressiveness
|
||||
duration: ""
|
||||
verification_result: passed
|
||||
completed_at: 2026-04-04T09:28:21.556Z
|
||||
blocker_discovered: false
|
||||
---
|
||||
|
||||
# T01: Added personality_weight (0.0–1.0) to chat API; modulates system prompt with creator voice profile and scales LLM temperature
|
||||
|
||||
**Added personality_weight (0.0–1.0) to chat API; modulates system prompt with creator voice profile and scales LLM temperature**
|
||||
|
||||
## What Happened
|
||||
|
||||
Added personality_weight field to ChatRequest with Pydantic validation (ge=0.0, le=1.0, default=0.0), forwarded to ChatService.stream_response(). When weight > 0 and creator is provided, queries Creator.personality_profile JSONB, extracts voice cues (signature phrases, tone, teaching style), and appends a personality injection block to the system prompt. Temperature scales linearly from 0.3 to 0.5. Graceful fallback to encyclopedic prompt on DB error, missing creator, or null profile. Wrote 9 new tests covering all paths.
|
||||
|
||||
## Verification
|
||||
|
||||
cd backend && python -m pytest tests/test_chat.py -v — 22 passed (13 existing + 9 new), 0 failed
|
||||
|
||||
## Verification Evidence
|
||||
|
||||
| # | Command | Exit Code | Verdict | Duration |
|
||||
|---|---------|-----------|---------|----------|
|
||||
| 1 | `cd backend && python -m pytest tests/test_chat.py -v` | 0 | ✅ pass | 1150ms |
|
||||
|
||||
|
||||
## Deviations
|
||||
|
||||
Simplified test_personality_weight_accepted_and_forwarded from wraps-based approach to capture-kwargs approach due to unbound method limitation.
|
||||
|
||||
## Known Issues
|
||||
|
||||
None.
|
||||
|
||||
## Files Created/Modified
|
||||
|
||||
- `backend/routers/chat.py`
|
||||
- `backend/chat_service.py`
|
||||
- `backend/tests/test_chat.py`
|
||||
|
||||
|
||||
## Deviations
|
||||
Simplified test_personality_weight_accepted_and_forwarded from wraps-based approach to capture-kwargs approach due to unbound method limitation.
|
||||
|
||||
## Known Issues
|
||||
None.
|
||||
48
.gsd/milestones/M023/slices/S02/tasks/T02-PLAN.md
Normal file
48
.gsd/milestones/M023/slices/S02/tasks/T02-PLAN.md
Normal file
|
|
@ -0,0 +1,48 @@
|
|||
---
|
||||
estimated_steps: 19
|
||||
estimated_files: 3
|
||||
skills_used: []
|
||||
---
|
||||
|
||||
# T02: Frontend — personality slider UI + API wiring in ChatWidget
|
||||
|
||||
Add a personality weight slider to ChatWidget and thread it through the streamChat() API call.
|
||||
|
||||
## Steps
|
||||
|
||||
1. In `frontend/src/api/chat.ts`, add `personalityWeight?: number` parameter to `streamChat()` function signature (after `conversationId`). Include `personality_weight: personalityWeight ?? 0` in the JSON body.
|
||||
2. In `frontend/src/components/ChatWidget.tsx`:
|
||||
- Add `const [personalityWeight, setPersonalityWeight] = useState(0);` state
|
||||
- In the panel header (the `<div className={styles.header}>` block), add a slider row below the title: `<div className={styles.sliderRow}>` containing a label span ('Encyclopedic' on left, 'Creator Voice' on right), and an `<input type='range' min={0} max={1} step={0.1}>` bound to personalityWeight state
|
||||
- Pass `personalityWeight` to `streamChat()` in the `sendMessage` callback (the 5th argument)
|
||||
3. In `frontend/src/components/ChatWidget.module.css`, add styles:
|
||||
- `.sliderRow` — flex row with gap, padding, align-items center, subtle border-bottom to separate from messages
|
||||
- `.sliderLabel` — small muted text (var(--text-secondary) or similar)
|
||||
- `.slider` — custom range input styling matching the dark theme (cyan accent track/thumb matching existing brand colors)
|
||||
4. Run `cd frontend && npm run build` — zero errors.
|
||||
|
||||
## Must-Haves
|
||||
|
||||
- [ ] streamChat() accepts and sends personalityWeight in POST body
|
||||
- [ ] ChatWidget renders a range slider in the panel header
|
||||
- [ ] Slider labels show 'Encyclopedic' at 0 and 'Creator Voice' at 1
|
||||
- [ ] Slider value flows through to API call
|
||||
- [ ] Styles match existing dark theme
|
||||
- [ ] Frontend builds with zero TypeScript/compilation errors
|
||||
|
||||
## Inputs
|
||||
|
||||
- ``frontend/src/api/chat.ts` — existing streamChat() with query, callbacks, creator, conversationId params`
|
||||
- ``frontend/src/components/ChatWidget.tsx` — existing ChatWidget with header, messages, input form`
|
||||
- ``frontend/src/components/ChatWidget.module.css` — existing dark-theme styles`
|
||||
- ``backend/routers/chat.py` — T01 output: ChatRequest now accepts personality_weight field`
|
||||
|
||||
## Expected Output
|
||||
|
||||
- ``frontend/src/api/chat.ts` — streamChat() sends personality_weight in POST body`
|
||||
- ``frontend/src/components/ChatWidget.tsx` — slider state + UI in header + passed to streamChat()`
|
||||
- ``frontend/src/components/ChatWidget.module.css` — slider row styles matching dark theme`
|
||||
|
||||
## Verification
|
||||
|
||||
cd frontend && npm run build
|
||||
|
|
@ -21,9 +21,11 @@ import uuid
|
|||
from typing import Any, AsyncIterator
|
||||
|
||||
import openai
|
||||
from sqlalchemy import select
|
||||
from sqlalchemy.ext.asyncio import AsyncSession
|
||||
|
||||
from config import Settings
|
||||
from models import Creator
|
||||
from search_service import SearchService
|
||||
|
||||
logger = logging.getLogger("chrysopedia.chat")
|
||||
|
|
@ -95,12 +97,43 @@ class ChatService:
|
|||
except Exception:
|
||||
logger.warning("chat_history_save_error cid=%s", conversation_id, exc_info=True)
|
||||
|
||||
async def _inject_personality(
|
||||
self,
|
||||
system_prompt: str,
|
||||
db: AsyncSession,
|
||||
creator_name: str,
|
||||
weight: float,
|
||||
) -> str:
|
||||
"""Query creator personality_profile and append a voice block to the system prompt.
|
||||
|
||||
Falls back to the unmodified prompt on DB error, missing creator, or null profile.
|
||||
"""
|
||||
try:
|
||||
result = await db.execute(
|
||||
select(Creator).where(Creator.name == creator_name)
|
||||
)
|
||||
creator_row = result.scalars().first()
|
||||
except Exception:
|
||||
logger.warning("chat_personality_db_error creator=%r", creator_name, exc_info=True)
|
||||
return system_prompt
|
||||
|
||||
if creator_row is None or creator_row.personality_profile is None:
|
||||
logger.debug("chat_personality_skip creator=%r reason=%s",
|
||||
creator_name,
|
||||
"not_found" if creator_row is None else "null_profile")
|
||||
return system_prompt
|
||||
|
||||
profile = creator_row.personality_profile
|
||||
voice_block = _build_personality_block(creator_name, profile, weight)
|
||||
return system_prompt + "\n\n" + voice_block
|
||||
|
||||
async def stream_response(
|
||||
self,
|
||||
query: str,
|
||||
db: AsyncSession,
|
||||
creator: str | None = None,
|
||||
conversation_id: str | None = None,
|
||||
personality_weight: float = 0.0,
|
||||
) -> AsyncIterator[str]:
|
||||
"""Yield SSE-formatted events for a chat query.
|
||||
|
||||
|
|
@ -151,6 +184,15 @@ class ChatService:
|
|||
# ── 3. Stream LLM completion ────────────────────────────────────
|
||||
system_prompt = _SYSTEM_PROMPT_TEMPLATE.format(context_block=context_block)
|
||||
|
||||
# Inject creator personality voice when weight > 0
|
||||
if personality_weight > 0 and creator:
|
||||
system_prompt = await self._inject_personality(
|
||||
system_prompt, db, creator, personality_weight,
|
||||
)
|
||||
|
||||
# Scale temperature with personality weight: 0.3 (encyclopedic) → 0.5 (full personality)
|
||||
temperature = 0.3 + (personality_weight * 0.2)
|
||||
|
||||
messages: list[dict[str, str]] = [
|
||||
{"role": "system", "content": system_prompt},
|
||||
]
|
||||
|
|
@ -165,7 +207,7 @@ class ChatService:
|
|||
model=self.settings.llm_model,
|
||||
messages=messages,
|
||||
stream=True,
|
||||
temperature=0.3,
|
||||
temperature=temperature,
|
||||
max_tokens=2048,
|
||||
)
|
||||
|
||||
|
|
@ -245,3 +287,47 @@ def _build_context_block(items: list[dict[str, Any]]) -> str:
|
|||
lines.append("")
|
||||
|
||||
return "\n".join(lines)
|
||||
|
||||
|
||||
def _build_personality_block(creator_name: str, profile: dict[str, Any], weight: float) -> str:
|
||||
"""Build a personality voice injection block from a creator's personality_profile JSONB.
|
||||
|
||||
The ``weight`` (0.0–1.0) determines how strongly the personality should
|
||||
come through. At low weights the instruction is softer ("subtly adopt");
|
||||
at high weights it is emphatic ("fully embody").
|
||||
"""
|
||||
vocab = profile.get("vocabulary", {})
|
||||
tone = profile.get("tone", {})
|
||||
style = profile.get("style_markers", {})
|
||||
|
||||
phrases = vocab.get("signature_phrases", [])
|
||||
descriptors = tone.get("descriptors", [])
|
||||
teaching_style = tone.get("teaching_style", "")
|
||||
energy = tone.get("energy", "moderate")
|
||||
formality = tone.get("formality", "conversational")
|
||||
|
||||
parts: list[str] = []
|
||||
|
||||
# Intensity qualifier
|
||||
if weight >= 0.8:
|
||||
parts.append(f"Fully embody {creator_name}'s voice and style.")
|
||||
elif weight >= 0.4:
|
||||
parts.append(f"Respond in {creator_name}'s voice.")
|
||||
else:
|
||||
parts.append(f"Subtly adopt {creator_name}'s communication style.")
|
||||
|
||||
if teaching_style:
|
||||
parts.append(f"Teaching style: {teaching_style}.")
|
||||
if descriptors:
|
||||
parts.append(f"Tone: {', '.join(descriptors[:5])}.")
|
||||
if phrases:
|
||||
parts.append(f"Use their signature phrases: {', '.join(phrases[:6])}.")
|
||||
parts.append(f"Match their {formality} {energy} tone.")
|
||||
|
||||
# Style markers
|
||||
if style.get("uses_analogies"):
|
||||
parts.append("Use analogies when helpful.")
|
||||
if style.get("audience_engagement"):
|
||||
parts.append(f"Audience engagement: {style['audience_engagement']}.")
|
||||
|
||||
return " ".join(parts)
|
||||
|
|
|
|||
|
|
@ -29,6 +29,7 @@ class ChatRequest(BaseModel):
|
|||
query: str = Field(..., min_length=1, max_length=1000)
|
||||
creator: str | None = None
|
||||
conversation_id: str | None = None
|
||||
personality_weight: float = Field(default=0.0, ge=0.0, le=1.0)
|
||||
|
||||
|
||||
@router.post("")
|
||||
|
|
@ -45,7 +46,7 @@ async def chat(
|
|||
- ``event: done`` — completion metadata with cascade_tier, conversation_id
|
||||
- ``event: error`` — error message (on failure)
|
||||
"""
|
||||
logger.info("chat_request query=%r creator=%r cid=%r", body.query, body.creator, body.conversation_id)
|
||||
logger.info("chat_request query=%r creator=%r cid=%r weight=%.2f", body.query, body.creator, body.conversation_id, body.personality_weight)
|
||||
|
||||
redis = await get_redis()
|
||||
service = ChatService(settings, redis=redis)
|
||||
|
|
@ -56,6 +57,7 @@ async def chat(
|
|||
db=db,
|
||||
creator=body.creator,
|
||||
conversation_id=body.conversation_id,
|
||||
personality_weight=body.personality_weight,
|
||||
),
|
||||
media_type="text/event-stream",
|
||||
headers={
|
||||
|
|
|
|||
|
|
@ -563,3 +563,307 @@ async def test_single_turn_fallback_no_redis_history(chat_client, mock_redis):
|
|||
assert len(captured_messages) == 2
|
||||
assert captured_messages[0]["role"] == "system"
|
||||
assert captured_messages[1]["role"] == "user"
|
||||
|
||||
|
||||
# ── Personality weight tests ─────────────────────────────────────────────────
|
||||
|
||||
|
||||
_FAKE_PERSONALITY_PROFILE = {
|
||||
"vocabulary": {
|
||||
"signature_phrases": ["let's gooo", "that's fire"],
|
||||
"jargon_level": "mixed",
|
||||
"filler_words": [],
|
||||
"distinctive_terms": ["sauce", "vibes"],
|
||||
"sound_descriptions": ["crispy", "punchy"],
|
||||
},
|
||||
"tone": {
|
||||
"formality": "casual",
|
||||
"energy": "high",
|
||||
"humor": "occasional",
|
||||
"teaching_style": "hands-on demo-driven",
|
||||
"descriptors": ["enthusiastic", "direct", "encouraging"],
|
||||
},
|
||||
"style_markers": {
|
||||
"explanation_approach": "example-first",
|
||||
"uses_analogies": True,
|
||||
"analogy_examples": ["like cooking a steak"],
|
||||
"sound_words": ["brrr", "thwack"],
|
||||
"self_references": "I always",
|
||||
"audience_engagement": "asks rhetorical questions",
|
||||
"pacing": "fast",
|
||||
},
|
||||
"summary": "High-energy producer who teaches by doing.",
|
||||
}
|
||||
|
||||
|
||||
def _mock_creator_row(name: str, profile: dict | None):
|
||||
"""Build a mock Creator ORM row with just the fields personality injection needs."""
|
||||
row = MagicMock()
|
||||
row.name = name
|
||||
row.personality_profile = profile
|
||||
return row
|
||||
|
||||
|
||||
def _mock_db_execute(creator_row):
|
||||
"""Return a mock db.execute that yields a scalars().first() result."""
|
||||
mock_scalars = MagicMock()
|
||||
mock_scalars.first.return_value = creator_row
|
||||
mock_result = MagicMock()
|
||||
mock_result.scalars.return_value = mock_scalars
|
||||
return AsyncMock(return_value=mock_result)
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_weight_accepted_and_forwarded(chat_client):
|
||||
"""personality_weight is accepted in the request and forwarded to stream_response."""
|
||||
search_result = _fake_search_result()
|
||||
|
||||
captured_kwargs = {}
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_kwargs.update(kwargs)
|
||||
return _mock_openai_stream(["ok"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "creator": "Keota", "personality_weight": 0.7},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
events = _parse_sse(resp.text)
|
||||
event_types = [e["event"] for e in events]
|
||||
assert "done" in event_types
|
||||
# Temperature should reflect the weight: 0.3 + 0.7*0.2 = 0.44
|
||||
assert captured_kwargs.get("temperature") == pytest.approx(0.44)
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_prompt_injected_when_weight_and_profile(chat_client):
|
||||
"""System prompt includes personality context when weight > 0 and profile exists."""
|
||||
search_result = _fake_search_result()
|
||||
creator_row = _mock_creator_row("Keota", _FAKE_PERSONALITY_PROFILE)
|
||||
|
||||
captured_messages = []
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_messages.extend(kwargs.get("messages", []))
|
||||
return _mock_openai_stream(["personality answer"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
# We need to mock db.execute inside the service — override the session
|
||||
mock_session = AsyncMock()
|
||||
mock_session.execute = _mock_db_execute(creator_row)
|
||||
|
||||
async def _mock_get_session():
|
||||
yield mock_session
|
||||
|
||||
app.dependency_overrides[get_session] = _mock_get_session
|
||||
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "snare tips", "creator": "Keota", "personality_weight": 0.7},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
assert len(captured_messages) >= 2
|
||||
|
||||
system_prompt = captured_messages[0]["content"]
|
||||
# Personality block should be appended
|
||||
assert "Keota" in system_prompt
|
||||
assert "let's gooo" in system_prompt
|
||||
assert "hands-on demo-driven" in system_prompt
|
||||
assert "casual" in system_prompt
|
||||
assert "high" in system_prompt
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_encyclopedic_fallback_null_profile(chat_client):
|
||||
"""When weight > 0 but personality_profile is null, falls back to encyclopedic prompt."""
|
||||
search_result = _fake_search_result()
|
||||
creator_row = _mock_creator_row("NullCreator", None)
|
||||
|
||||
captured_messages = []
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_messages.extend(kwargs.get("messages", []))
|
||||
return _mock_openai_stream(["encyclopedic answer"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
mock_session = AsyncMock()
|
||||
mock_session.execute = _mock_db_execute(creator_row)
|
||||
|
||||
async def _mock_get_session():
|
||||
yield mock_session
|
||||
|
||||
app.dependency_overrides[get_session] = _mock_get_session
|
||||
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "reverb tips", "creator": "NullCreator", "personality_weight": 0.5},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
system_prompt = captured_messages[0]["content"]
|
||||
# Should be the standard encyclopedic prompt, no personality injection
|
||||
assert "Chrysopedia" in system_prompt
|
||||
assert "NullCreator" not in system_prompt
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_encyclopedic_fallback_missing_creator(chat_client):
|
||||
"""When weight > 0 but creator doesn't exist in DB, falls back to encyclopedic prompt."""
|
||||
search_result = _fake_search_result()
|
||||
|
||||
captured_messages = []
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_messages.extend(kwargs.get("messages", []))
|
||||
return _mock_openai_stream(["encyclopedic answer"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
mock_session = AsyncMock()
|
||||
mock_session.execute = _mock_db_execute(None) # No creator found
|
||||
|
||||
async def _mock_get_session():
|
||||
yield mock_session
|
||||
|
||||
app.dependency_overrides[get_session] = _mock_get_session
|
||||
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "bass tips", "creator": "GhostCreator", "personality_weight": 0.8},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
system_prompt = captured_messages[0]["content"]
|
||||
assert "Chrysopedia" in system_prompt
|
||||
assert "GhostCreator" not in system_prompt
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_weight_zero_skips_profile_query(chat_client):
|
||||
"""When weight is 0.0, no Creator query is made even if creator is set."""
|
||||
search_result = _fake_search_result()
|
||||
|
||||
captured_kwargs = {}
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_kwargs.update(kwargs)
|
||||
return _mock_openai_stream(["ok"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
mock_session = AsyncMock()
|
||||
mock_session.execute = AsyncMock() # Should NOT be called
|
||||
|
||||
async def _mock_get_session():
|
||||
yield mock_session
|
||||
|
||||
app.dependency_overrides[get_session] = _mock_get_session
|
||||
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "creator": "Keota", "personality_weight": 0.0},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
# DB execute should not have been called for Creator lookup
|
||||
mock_session.execute.assert_not_called()
|
||||
# Temperature should be 0.3 (base)
|
||||
assert captured_kwargs.get("temperature") == pytest.approx(0.3)
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_temperature_scales_with_weight(chat_client):
|
||||
"""Temperature scales: 0.3 at weight=0.0, 0.5 at weight=1.0."""
|
||||
search_result = _fake_search_result()
|
||||
creator_row = _mock_creator_row("Keota", _FAKE_PERSONALITY_PROFILE)
|
||||
|
||||
captured_kwargs = {}
|
||||
mock_openai_client = MagicMock()
|
||||
|
||||
async def _capture_create(**kwargs):
|
||||
captured_kwargs.update(kwargs)
|
||||
return _mock_openai_stream(["warm"])
|
||||
|
||||
mock_openai_client.chat.completions.create = AsyncMock(side_effect=_capture_create)
|
||||
|
||||
with (
|
||||
patch("chat_service.SearchService.search", new_callable=AsyncMock, return_value=search_result),
|
||||
patch("chat_service.openai.AsyncOpenAI", return_value=mock_openai_client),
|
||||
):
|
||||
mock_session = AsyncMock()
|
||||
mock_session.execute = _mock_db_execute(creator_row)
|
||||
|
||||
async def _mock_get_session():
|
||||
yield mock_session
|
||||
|
||||
app.dependency_overrides[get_session] = _mock_get_session
|
||||
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "creator": "Keota", "personality_weight": 1.0},
|
||||
)
|
||||
|
||||
assert resp.status_code == 200
|
||||
assert captured_kwargs.get("temperature") == pytest.approx(0.5)
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_weight_above_1_returns_422(chat_client):
|
||||
"""personality_weight > 1.0 fails Pydantic validation with 422."""
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "personality_weight": 1.5},
|
||||
)
|
||||
assert resp.status_code == 422
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_weight_below_0_returns_422(chat_client):
|
||||
"""personality_weight < 0.0 fails Pydantic validation with 422."""
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "personality_weight": -0.1},
|
||||
)
|
||||
assert resp.status_code == 422
|
||||
|
||||
|
||||
@pytest.mark.asyncio
|
||||
async def test_personality_weight_string_returns_422(chat_client):
|
||||
"""personality_weight as a non-numeric string fails validation with 422."""
|
||||
resp = await chat_client.post(
|
||||
"/api/v1/chat",
|
||||
json={"query": "test", "personality_weight": "high"},
|
||||
)
|
||||
assert resp.status_code == 422
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue