What I have built: Inspired by the awesome work at Langgraph, PyAutogen, and CrewAI, I made an LLM agentic framework called Beehive.
Beehive shares many features with other popular agentic frameworks, including:
- Loops and conditionals between agents - State management - Streaming support - Memory / feedback
In addition, Beehive supports:
- Rapid customization: it's trivial to create your own agent with its unique invokation logic - Nested patterns: think, a Beehive of Beehives - Complex conversational patterns: ensembles and multi-agent debates are supported out-of-the-box
Why I built it: For learning purposes! I wanted to get my feet wet with LLMs and agents. Having tinkered with and explored the frameworks I mentioned above, I wanted to build my own to deepen my understanding.
Any and all feedback is welcome!