BabyAGI
Lightweight Python script that creates a self-directing AI task loop, breaking a goal into subtasks and iterating until complete.
About
BabyAGI is a minimal proof-of-concept autonomous agent written in Python by Yohei Nakajima. It uses an LLM to generate tasks, a vector store to keep track of context, and a simple execution loop that continuously creates, prioritizes, and completes subtasks until a top-level objective is reached.
The codebase is intentionally small and readable, making it a popular starting point for learning about agent architectures. It is not a finished product but a research reference and experiment. Running it requires an OpenAI API key and a compatible vector database such as Pinecone or Chroma.
The codebase is intentionally small and readable, making it a popular starting point for learning about agent architectures. It is not a finished product but a research reference and experiment. Running it requires an OpenAI API key and a compatible vector database such as Pinecone or Chroma.