ChainStream
A Stream-based LLM Agent Framework for Continuous Context Sensing and Sharing
一个流式的LLM智能体框架，用于连续上下文感知和共享。
Welcome to Everyone! 
We welcome everyone to participate in the development of ChainStream, whether it is a developer, user, contributor, or any other role.
Join Us! 
ChainStream is still work in progress, welcome to every user, suggester and contributor. Feel free to connect with us through Github Repo or Zulip Group :). Also, this documentation is still under development, please check back frequently for updates.
Documentation Architecture
For everyone, we suggest you read:
- ChainStream Overview to get an overview of ChainStream's basic concepts and features.
This documentation is written for three main audiences:
- ChainStream User This audience is for those who want to use ChainStream to build their own agent or integrate it into their existing system.
- Agent Developer This audience is for those who want to understand the inner workings of ChainStream and build their own agent.
- ChainStream Sys Contributor This audience is for those who want to contribute to the development of ChainStream.
Updates
- 2024.06: ChainStream website is launched.
- 2024.06: ChainStream project team participated in Mobisys2024 in Tokyo and presented a report at EdgeFM Workshop. The paper will be published soon.
- 2024.06: ChainStream Github Repo is launched.
- 2024.05: The introduction paper of ChainStream was accepted by MobiSys 2024, Workshop on Edge and Mobile Foundation Models.
- 2023.12: The project of ChainStream started preliminary development.
- 2023.10: The project of ChainStream was initiated.
Copyright
- ChainStream is designed and developed by the AIoT Team from Institute for AI Industry Research (AIR), Tsinghua University.
- Main contributors: Yuanchun Li, Jiacheng Liu, Wenxing Xu, Liangyan Li, Wenjie Du and others. We will release more detailed references in the future.