OpenAssistants is a collection of open source libraries aimed at developing robust AI assistants rather than autonomous agents. By focusing on specific tasks and incorporating human oversight, OpenAssistants strives to minimize error rates typically found in agentic systems.
openassistantsthe core library responsible for the main function calling / message generating runtimeopenassistants-fastapia set of FastAPI routes used for interacting with the core runtime loop through a REST APIopenassistants-reactan example chat client that supports rich streaming outputs like tables, plots, form inputs and text.
OpenAssistants is built on LangChain and designed to be an open alternative to OpenAI's Assistants API. We also think it's easier to use!
✨ Check out the live demo
Join us in creating AI assistants that are not only useful but dependable for production use today.
- Included Chat UI
- Support function calling with any LLM (open source & proprietary)
- Declarative library of functions
- Built-in SQL functions (DuckDB support)
- Extend with any custom Python function
- Support for 50+ functions in a single chat Assistant
- Native OpenAI Functions integration
OpenAssistants UI
To run the project locally:
-
Navigate to
examples/fast-api-serverto start the backend example:- Run
poetry install - Activate the virtual environment with
poetry shell - Set OPENAI_API_KEY env var with
export OPENAI_API_KEY=sk-my-key - Start the server with
./run.sh
- Run
- From the root, run
yarn install - Navigate to the sample next app
cd examples/next - Launch the development server with
yarn dev - Access the application at
localhost:3000
Join our community and start contributing today!
To develop the UI library locally, youll also want to compile in watch for hot reloading. In a new terminal window
- navigate to the component library
cd packages/openassistants-react - run the compiler in watch mode
yarn start
Now any changes to the library will also get picked and displayed in your process running the next app