File size: 1,205 Bytes
e317d56 52b0ede e317d56 52b0ede e317d56 52b0ede e317d56 46cc63a e317d56 52b0ede e317d56 52b0ede e317d56 52b0ede e317d56 52b0ede e317d56 52b0ede e317d56 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 | # Architecture — youtube_hate_detector
## Runtime (production)
```mermaid
flowchart LR
Browser[React SPA]
API[FastAPI :8000]
MS[ModelService]
YT[YouTube Data API]
Browser -->|HTTP JSON| API
API --> MS
API --> YT
```
- **UI:** `frontend/` built to `frontend/dist`, served by FastAPI `StaticFiles` in production.
- **Inference:** Only `ModelService` in `src/service/` loads models.
- **Catalog:** `configs/model_catalog.yaml` — baselines (LR, frozen BERT) + production meta-stack.
- **Suggested videos:** `configs/suggested_videos.yaml` — YouTube video IDs for the right rail.
## Local development
| Process | Command | Port |
|---------|---------|------|
| API | `uv run uvicorn src.api.main:app --reload` | 8000 |
| UI | `cd frontend && npm run dev` | 5173 (proxies API) |
## Docker
Single service `youtube_hate_detector-app` on port **8000** (API + static UI).
## API layout
```
src/api/
main.py # app factory, CORS, static mount
schemas.py # Pydantic models
services.py # predict helpers
youtube.py # comment fetch + metadata
state.py # shared app state
routes/
health.py
models.py
predict.py
videos.py
```
|