WEBUI.md

Web UI

This Web UI lives under examples/docker/server/webui/, but audiocpp_server can also serve it directly when webui_root is configured.

Location:

• examples/docker/server/webui/

Recommended mode is backend-hosted: the browser loads the UI from the same origin as audiocpp_server, so there is no separate proxy hop and no CORS issue.

Integrated Run

For the container examples, set "webui_root": "/app/webui" in server.json and start the backend:

docker compose -f compose.podman.yml up

That exposes both the API and the Web UI at http://127.0.0.1:8880.

Open:

http://127.0.0.1:8880

Expose it to your LAN by publishing the backend port and binding the server to 0.0.0.0 in server.json.

Standalone Run

If you still want to run the Web UI by itself:

cd webui
npm start

Lint the Web UI:

cd webui
npm run lint

The lint script uses npx to run ESLint with the checked-in eslint.config.js, so no separate frontend build step is required for this example app.

Defaults:

• Web UI: http://127.0.0.1:3000
• Proxied backend: http://127.0.0.1:8880

Override host, port, or backend origin manually:

WEBUI_HOST=0.0.0.0 WEBUI_PORT=3000 AUDIOCPP_API_ORIGIN=http://127.0.0.1:8080 npm start

Then open it from another machine using your host machine's LAN IP, for example:

http://192.168.1.50:3000

Current v1 Scope

• Fetch models from GET /v1/models
• Switch between TTS and ASR models based on the server-reported task
• For TTS models, submit text to POST /v1/audio/speech
• For TTS models, populate family-specific voice controls from GET /v1/audio/voices?model=<id>
• For pocket_tts, choose either a built-in/cached voice, a shared WAV sample from samples[], or a locally uploaded WAV for clone mode
• For multi-speaker TTS families such as vibevoice, choose each speaker row from shared WAV samples[], preset-backed voice_presets, or a local upload, then submit ordered voice_samples through POST /v1/audio/speech
• For ASR models, upload WAV audio to POST /v1/audio/transcriptions

Non-goals:

• No sidecar container