Self-host faster-chat

This self-hosted web application lets you chat with AI models while keeping full control over your data. It supports 17 built-in providers including OpenAI, Anthropic, Google, Groq, Mistral, and others, plus 50 more discoverable through models.dev. You can also run it entirely offline with local inference engines like Ollama or LM Studio. The tech stack centers on Preact for a tiny 3 KB runtime, Hono for the backend, TanStack Router and Query for client state, Tailwind CSS 4.1 for styling, and Bun for fast builds. Conversations are stored in a server-side SQLite database with IndexedDB caching in the browser. Core capabilities include streaming chat, cross-conversation memory that the AI builds automatically, web search via Brave Search with citations, image upload for vision models, image generation through DALL-E 3 and FLUX, ChatGPT conversation import, markdown with code highlighting and LaTeX, voice input and output, 15-plus themes with dark and light modes, keyboard shortcuts, and a responsive layout. Administration features cover multi-user authentication with three roles, encrypted API key storage, a provider hub that auto-discovers models, one-click Ollama model downloads from the UI, and white-labeling options. Deployment uses Docker Compose on port 8787 with an optional Caddy reverse proxy for automatic HTTPS. The project is in maintenance mode with no plans for RAG, plugins, PostgreSQL, or native mobile apps.