light-heart-labs--dreamserver/ods/extensions/services/open-webui

open-webui

Primary chat interface for ODS

Overview

Open WebUI is the main user-facing web application bundled with ODS. It provides a full-featured chat UI backed by the local llama-server LLM, with integrated web search via SearXNG, image generation via ComfyUI, and voice input/output via Whisper (STT) and Kokoro (TTS).

Open WebUI is served at http://localhost:3000 and communicates with llama-server through the OpenAI-compatible API.

Features

• Chat interface: Multi-turn conversations with the local LLM
• Web search: Integrated SearXNG metasearch for grounded answers
• Image generation: ComfyUI backend using SDXL Lightning 4-step (1024×1024)
• Voice input: Speech-to-text via Whisper (/v1/audio/transcriptions)
• Voice output: Text-to-speech via Kokoro (/v1/audio/speech)
• User authentication: Optional login system (enabled by default)
• Document Q&A: Upload files and chat with their contents (requires Qdrant)
• Model selection: Switch between models at runtime

Configuration

Environment variables (set in .env):

Variable	Default	Description
WEBUI_SECRET	(required)	Session signing key — generate with openssl rand -hex 32
WEBUI_PORT	3000	External port (maps to internal 8080)
WEBUI_AUTH	true	Enable user authentication (true/false)
TIMEZONE	UTC	System timezone for timestamps
LLM_API_URL	http://llama-server:8080	LLM backend URL (internal Docker hostname)

Voice configuration

Open WebUI connects to Whisper and Kokoro automatically using internal Docker hostnames. To change the default models:

Variable (in docker-compose.nvidia.yml)	Default	Description
AUDIO_STT_MODEL (NVIDIA overlay)	deepdml/faster-whisper-large-v3-turbo-ct2	Whisper model on NVIDIA
AUDIO_STT_MODEL (base)	Systran/faster-whisper-base	Whisper model on AMD/CPU
AUDIO_TTS_VOICE	af_heart	Kokoro voice name

Architecture

Browser
  │
  ▼
Open WebUI (:3000)
  ├── LLM Chat ──────────────▶ llama-server:8080 (OpenAI API)
  ├── Web Search ─────────────▶ SearXNG:8080
  ├── Image Generation ───────▶ ComfyUI:8188
  ├── Speech-to-Text ─────────▶ Whisper:8000
  └── Text-to-Speech ─────────▶ Kokoro TTS:8880

Data Persistence

User accounts, chat history, and uploaded documents are stored in data/open-webui/. This volume is mounted at /app/backend/data inside the container.

First Use

1. Open http://localhost:3000 in your browser
2. Create an admin account (first user automatically becomes admin)
3. Start chatting — the LLM is ready when llama-server reports healthy

Troubleshooting

Open WebUI not loading:

docker compose ps open-webui
docker compose logs open-webui

"Connection refused" to LLM:

• Verify llama-server is healthy: curl http://localhost:8080/health
• Check LLM_API_URL in .env

Voice input not working:

• Confirm Whisper is running: curl http://localhost:9000/health
• Browser must have microphone permission

Image generation not available:

• Requires ComfyUI service to be running
• Enable via ods enable comfyui

Authentication issues:

• To disable auth (local use only): set WEBUI_AUTH=false in .env and restart
• To reset admin password: remove data/open-webui/ (loses all chat history)

Files

• manifest.yaml — Service metadata and feature definitions

License

Part of ODS — Local AI Infrastructure