Open-source machine learning framework for automating text and voice-based conversations, Rasa Open Source enables developers to build contextual assistants. While historically focused on NLU and dialogue management, the project is now in maintenance mode, directing users towards its newer platform, Hello Rasa, which leverages LLMs and a CALM engine for conversational AI.

How It Works

Rasa Open Source traditionally employed machine learning models for Natural Language Understanding (NLU) to interpret user intents and entities, and dialogue management to track conversation state and decide on responses. The newer CALM (Conversational AI with Language Models) approach, featured in Hello Rasa, shifts this paradigm by integrating LLM flexibility for understanding with strict business logic enforcement via code, using a concept called "Flows" instead of rigid dialogue trees.

Quick Start & Requirements

Installation for development requires Poetry for dependency management. After installing Poetry, the primary build command is make install. A note for macOS users suggests setting export SYSTEM_VERSION_COMPAT=1 if compiler issues arise. Development requirements for running tests can be installed via make prepare-tests-ubuntu (Ubuntu/Debian) or make prepare-tests-macos (macOS).

Highlighted Details

• Supports integration with numerous messaging channels including Slack, Facebook Messenger, Telegram, Twilio, and Microsoft Bot Framework.
• Designed to handle complex, layered conversations with significant back-and-forth interactions.
• Follows Semantic Versioning for releases (Major, Minor, Patch).

Maintenance & Community

Rasa Open Source is currently in maintenance mode. The project's future development is focused on "Hello Rasa" and the CALM engine. Users are encouraged to join the "Agent Engineering Community" for discussions on AI agent architectures, memory, orchestration, and safety. Links to community channels and the Hello Rasa playground are provided.

Licensing & Compatibility

Licensed under the Apache License, Version 2.0. This license generally permits commercial use and integration into closed-source projects, with standard attribution and copyleft requirements for derivative works under the Apache license itself.

Limitations & Caveats

Rasa Open Source is in maintenance mode, indicating that active feature development has ceased for this version. New projects should consider the Hello Rasa platform and CALM engine, as Rasa Open Source is effectively legacy. Documentation and installation instructions provided apply specifically to this older framework.