xingyaoww / code-act

Official Repo for ICML 2024 paper "Executable Code Actions Elicit Better LLM Agents" by Xingyao Wang, Yangyi Chen, Lifan Yuan, Yizhe Zhang, Yunzhu Li, Hao Peng, Heng Ji.
MIT License
468 stars 38 forks source link

Clarification on Session Context and Memory Retention in CodeAct #12

Open JohannesvdMeer opened 2 months ago

JohannesvdMeer commented 2 months ago

Hello,

I’ve been using CodeAct integrated with GPT-4o for a chatbot interface. I noticed that the model doesn't seem to maintain context across separate API calls within the same session. Could you clarify how CodeAct handles session context and whether there are any recommended strategies for ensuring continuity across conversations? Any insights would be greatly appreciated.

Thanks!

Johannes