Does OvertChat run the AI model itself?
No. It connects to inference servers you already run, such as vLLM, llama.cpp or SGLang, and can also use hosted providers or custom compatible endpoints.
Chat with the local models your homelab already runs, such as vLLM, llama.cpp or SGLang. It adds web search, voice chat and a separate account for each person. The installer can set up SearXNG, Kokoro and Parakeet for search and speech, so a local model needs no provider API keys. Clients exist for web, Android, iOS and desktop.
Compare OvertChat with other AI apps. Add or remove apps to see how they differ.
| Requirement | OvertChat | |
|---|---|---|
| Price | Free | Free |
| Licence | MIT | MIT |
| Web UI | Yes | Yes |
| Minimum hardware | Not stated | 1 vCPU · 1 GiB RAM |
| Docker Containers | 2 (app +1) | 6 (app + Admin panel, MongoDB, Meilisearch, pgvector, RAG API) |
| Database | SQLite | MongoDB |
| Architecture | x86 + ARM | x86 + ARM |
| Tech stack | Node | Node.js |
| Docker | Yes | Yes |
| Podman | Yes | – |
| iOS | Yes | – |
| Android | Yes | – |
| Windows | Yes | – |
| macOS | Yes | – |
| Linux | Yes | – |
Official client for phone and desktop.
No. It connects to inference servers you already run, such as vLLM, llama.cpp or SGLang, and can also use hosted providers or custom compatible endpoints.
Yes. The first account you create is the administrator, and public signup stays closed after that; you add everyone else in Settings under Users. Each person gets separate conversations and projects on web, Android and iOS.
Not with the bundled services. The installer can add SearXNG, Kokoro text to speech and Parakeet speech to text, so a local model works without provider keys. Search queries go to whichever search providers are enabled.
Yes outside localhost. Browsers only allow microphone access on HTTPS pages, so put OvertChat behind a reverse proxy with a certificate for realtime voice.
Get a ChatGPT-style chat UI that talks to many AI providers and local models, with user accounts so the household can share one install.
Run local LLMs, image generation, and voice AI on your own hardware with no cloud and no required GPU.
Turn your own PDFs, notes and web pages into a research notebook that answers with citations, then into reports, flashcards, quizzes or slide decks. better-vault is a self-hosted NotebookLM alternative that keeps the notebooks in your browser, works with your own API key or a local model, and serves it from one Node process on your network.
Run an AI assistant on your own devices that answers you in the chat apps you already use.
Download Wikipedia, Khan Academy courses and regional maps onto an x86 machine once, and Project NOMAD serves them to the rest of your network as an offline knowledge server. A local AI chat through Ollama sits alongside.
A NotebookLM-style research assistant for your own documents: it reads PDFs, Office files, images and audio, answers with citations, and can edit or convert what it finds. It runs as a desktop app or one Docker container, with a local model if you want the work to stay on your hardware.