# Voice-GPU Developer Guide (Web Publish 2026-05-24)

Workspace: `E:\Voice-GPU`  
Core service: `E:\Voice-GPU\WebApp`

## Core architecture

- FastAPI backend: `WebApp/app/main.py`
- Synthesis service: `WebApp/app/omnivoice_service.py`
- Queue + persistence: `WebApp/app/job_queue.py` + SQLite
- Auth/usage controls: `WebApp/app/api_access.py`
- Frontend static: `WebApp/web`

## Runtime

- Supervisor: `WebApp/scripts/phase03_supervisor.ps1`
- Uvicorn bind: `127.0.0.1:8015`
- Mode: `real`
- Device: `cuda:0`
- Model: `k2-fsa/OmniVoice`

## Important endpoints

- `GET /api/health`
- `GET /api/system/version`
- `GET /v1/models`
- `POST /v1/audio/speech`
- `GET /api/nonverbal/tags`
- `POST /api/nonverbal/validate`
- `GET/POST /api/pronunciation/lexicons`
- `POST /api/web/jobs/tts`

## Current auth default

In `WebApp/app/api_access.py`:

- `OMNIVOICE_API_AUTH_ENABLED` default is currently `false`.
- Re-enable strict key gate with env: `OMNIVOICE_API_AUTH_ENABLED=1`.

## Full real verification script

Run:

```powershell
python WebApp/scripts/public_frontend_backend_full_verify.py
```

Latest evidence:

- `E:\Voice-GPU\WebApp\docs\reports\frontend_backend_full_real_20260524_194707\report.html`
- `E:\Voice-GPU\WebApp\docs\reports\frontend_backend_full_real_20260524_194707\summary.json`