BabyAGI
Explore the self-building autonomous-agent direction via the functionz function framework — a database-backed store of named functions with dependency tracking that experimental agents like process_user_input and self_build can extend at runtime. The original March-2023 task-list agent has been archived.
Description
BabyAGI is Yohei Nakajima's experimental autonomous-agent project. The original March 2023 BabyAGI — the viral task-creation / prioritisation / execution loop that defined the early 'autonomous agent' category — has been archived to babyagi_archive (September 2024 snapshot). The current repository is a complete rewrite around functionz: a Python function framework that stores, manages, and executes functions from a database with graph-based dependency tracking, automatic loading, logging, and a web dashboard. Experimental agents (process_user_input, self_build) generate new functions from user requests, decomposing them into 'smaller reusable components'. Yohei is the sole maintainer ('nights and weekends'); PR throughput is slow.
Solution
Self-building function-graph loop. The functionz store holds named Python functions with declared dependencies. Self-building agents (process_user_input, self_build) receive a user request, break it into 'smaller reusable components', generate the required functions, register them in the database, and execute them — logging every step and surfacing the graph in a web dashboard. The original task-list shape (task creation → prioritisation → execution → enqueue) lives in the archived repo, not the current code.
Primary use cases
- self-building agent research prototype
- function-store / function-graph experimentation
- historical reference for the 2023 task-list agent shape (via babyagi_archive)
Open the full interactive page →
Diagram, neighbourhood map, code examples, related patterns and full provenance.