OllaHub/README.md

Quick start · Docs · Issues

What it is

OllaHub is a desktop chat app for the models you run with Ollama. On first run it scans your hardware, recommends a model and downloads it. After that it chats with streaming, keeps history in a local SQLite database, and can add context from web search, MCP tools and scheduled tasks. The UI is Next.js inside a Tauri 2 window, and a Rust backend does the work.

OllaHub is a client, not a model runtime. It talks to the Ollama API (default http://localhost:11434, configurable in Settings), tries to start Ollama when it is installed but stopped, and can download the Ollama installer when it is missing.

Why use it

• Local by default. Prompts go to your own Ollama, and chats live in ollahub.db in the app data directory, with FTS5 full-text search. Besides downloads (models, the embedding model, updates), web search is what reaches the internet.
• Web answers ranked on your machine. Headless Chrome scrapes Google, Bing, Yahoo, DuckDuckGo and Startpage (order configurable), then local embeddings (all-MiniLM-L6-v2 on ONNX Runtime) rank and prune the pages, with a BM25-like fallback when the embedding model is unavailable. A deep-research mode runs planning → searching → aggregating → validating → formulating.
• More than a chat box. MCP servers run as child processes configured in mcp_config.json, their tools can be called from the chat, and cron tasks (search and summarize, notification, custom prompt) keep running in the background.

When not to use it: if you want a finished, tested product (version 0.1.0, no test suite), a model runtime of its own, or a server you can run without a desktop.

How it works

• Frontend (app/, components/, hooks/, lib/, store/): Next.js 16 with output: 'export', React 19, Tailwind 4, Radix UI and Zustand. It calls the backend through Tauri invoke and events.
• Backend (src-tauri/src/): lib.rs holds the commands for chat (chat_stream), models, the Ollama lifecycle, sessions, MCP, search, tasks, export and the installer. db.rs handles SQLite, web_scraper.rs the search engines, embeddings.rs the ranking, and scheduler.rs the cron tasks stored in tasks.json.
• Models: list, delete and pull with progress, install a local .gguf file, and get a recommendation for your hardware.
• System monitor: CPU and RAM stats plus GPU VRAM, usage and temperature. GPU detection uses wmic and nvidia-smi on Windows and lspci on Linux.
• Updates: the Tauri updater checks automatically. Its endpoint in src-tauri/tauri.conf.json points to a GitHub Releases latest.json.

More notes: docs/updater-setup.md (signing keys and updater) and docs/style-guide.md (UI spacing, radii, accessibility).

Quick start

No release is published on this repository yet, so build it locally. You need Node.js 20, pnpm 8, a stable Rust toolchain and the Tauri 2 prerequisites (the versions come from the CI workflow). On Ubuntu the workflow also installs libwebkit2gtk-4.1-dev libssl-dev libayatana-appindicator3-dev librsvg2-dev patchelf. Ollama is needed at runtime.

git clone https://git.evandro.dev.br/evandro/OllaHub.git
cd OllaHub
pnpm install

pnpm tauri dev          # starts `pnpm dev` (Next.js on http://localhost:3000) inside the Tauri window
pnpm tauri build        # runs `pnpm generate-changelog && pnpm build`, then bundles the installer
pnpm lint               # ESLint
pnpm bump-version       # patch | minor | major, optional --commit

The workflow .github/workflows/release.yml builds msi (Windows), dmg (macOS, Apple Silicon and Intel) and AppImage (Linux) on pushes to main, master and develop, and publishes release v<version> on pushes to main or master.

Warning

Two things the app does on its own.

1. Opening the chat page pulls the small helper model qwen2.5:0.5b if it is not installed yet, used for automatic chat titles. Expect an extra download on first use.
2. Closing the window only hides it. The close request is intercepted (prevent_close) so scheduled tasks keep running, but there is no tray icon to bring the window back yet; the tray step is still pending in IMPLEMENTATION_GUIDE.md.

Project status

Early: version 0.1.0, 47 commits between 2025-10-27 and 2025-11-27, no automated tests (package.json defines dev, build, start, lint, generate-changelog and bump-version). Known limitations:

• No tray icon yet: closing the window hides it (see the warning above).
• "Start with the system" does nothing yet: the setting only toggles a stored flag; the code has a TODO until tauri-plugin-autostart is configured.
• RAG inside chat_stream is not wired: the enable_rag branch is commented out as TODO; the frontend assembles web context before the request.
• Embeddings outside Windows: the ONNX Runtime library is downloaded automatically only on Windows. On macOS and Linux, provide it in the app data folder or on the system path, otherwise ranking falls back to BM25-like scoring.
• Unused leftovers: src-tauri/scripts/scraper.py is not called by the Rust backend, and the 188 model files in data/models/ are not referenced by the app code.

License

No published license: all rights reserved.

Built with Tauri, Next.js, Radix UI, Tailwind CSS, Zustand, Lucide, headless_chrome, ONNX Runtime and the all-MiniLM-L6-v2 model.

_{Made by Evandro Fonseca Junior · git.evandro.dev.br}