How xomyhoc/open-webui works
Project type and purpose
• This is a full-stack monolithic application (“Open WebUI”) combining a SvelteKit-based PWA frontend with a Python backend.
• Purpose: provide a self-hosted AI chat platform with Retrieval-Augmented Generation (RAG), plugin support, voice/video calls, image generation, RBAC, multi-model orchestration, and offline operation.
File structure
• Top-level:
• backend/ – Python service (FastAPI style) with routers, models (SQLAlchemy), migrations (Alembic), retrieval modules (vector stores & web scrapers), tasks, utils, storage provider.
• src/ – SvelteKit frontend (routes/, lib/components, lib/apis, stores, workers).
• cypress/ & test/ – end-to-end and unit tests.
• Dockerfile, docker-compose.yaml, kubernetes/ – containerization & K8s deployment.
• config: env files, Caddyfile, Makefile, hatch_build.py, pyproject.toml, package.json, Vite/Svelte/Tailwind configs.
README for additional insights
• It confirms a Docker/Kubernetes deploy, Python 3.11 backend (open-webui serve CLI), Ollama/OpenAI runners, vector DB for RAG, plugin & pipeline architecture, PWA offline support, RBAC, i18n.