AllYourBot/hostedgpt

An open version of ChatGPT you can host anywhere or run locally.

HostedGPT – Self‑hosted Chat UI for many LLM providers

What it is – HostedGPT is an open‑source Ruby on Rails web app that gives you a ChatGPT‑like interface you can run on your own server, laptop, or any cloud host. It does not provide its own language model; instead it forwards user messages to whichever LLM API you configure (OpenAI, Anthropic Claude, Google Gemini, Ollama, Groq, etc.) and returns the response.

Key capabilities

  • Multi‑provider support – Switch between providers or even between models mid‑conversation, making it easy to compare outputs.
  • Web‑search tool – If you supply a free Brave Search API key, the assistant can perform live web searches.
  • Google integration (experimental) – Attach a Google account so the bot can read/write Gmail, Tasks and (soon) Calendar.
  • Conversation history – All chats are stored, searchable and shareable.
  • Mobile‑friendly UI – Works in a browser and can be added to a phone’s home screen.
  • German localisation – UI is available in German out of the box.

How to get it running

  1. Deploy – One‑click deployment options for Render, Fly.io, and Heroku (the latter currently broken). Free Render tier works for 90 days, after which a $7/month plan is usually needed.
  2. Docker – The recommended local setup is docker compose up --build; it pulls a PostgreSQL container and the Rails app.
  3. Bare‑metal – You can also run the app directly with Ruby, PostgreSQL, ImageMagick, and rbenv.
  4. Configuration – Most behaviour is toggled via environment variables (e.g., default API keys, Brave search key, Cloudflare R2 storage, Google‑tools flag, voice‑chat flag, authentication method).

Who it’s for

  • Developers who want a self‑hosted ChatGPT‑style UI without paying multiple subscription fees.
  • Teams that need to keep conversation data on‑premise or integrate the chat with internal tools (Google apps, file storage, etc.).
  • Anyone who wants to experiment with different LLM providers side‑by‑side.

Limitations

  • The app itself does not contain an LLM; you must supply valid API keys and pay per‑request to the chosen provider.
  • Some optional features (Google tools, voice, Heroku deployment) are marked as experimental or currently broken.
  • Free Render hosting pauses after 15 minutes of inactivity, causing a slow resume.

Community – The repo is maintained by an experienced Rails developer and welcomes contributors. It includes a test suite, Docker workflow, and documentation for extending models, adding storage back‑ends, or fixing deployment scripts.

Related

  • Project
  • Project
  • Project
  • Project