DeepYardDeepYard
A

AsyncFC

Asynchronous function calling framework that parallelizes LLM tool execution for faster agents

Open SourceFree

About

Academic framework from UC Berkeley that decouples LLM decoding from function execution, enabling concurrent tool use without requiring model modifications. Reduces end-to-end latency through parallelism by allowing agents to execute multiple functions simultaneously while the LLM continues processing. Particularly useful for agents that need to call multiple tools or APIs in workflows where execution order isn't strictly sequential.

Details

Language
Patterns

Tags

frameworktool-useopen-sourcepythonautonomous