AIAnytime / AIAnytime/Intro-to-LangChain-Expression-Language

Modeling a complex process with LCEL

未关闭
#1 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Jupyter Notebook
星标
4
派生
1
PR 合并指标
30 天内没有已合并 PR

描述

Thank you for your tremendous work! Maybe you could give me a hint how to achieve such process usinc LCEL:
```
1. User message as an input.
2. Message type classification [General, Action request].
2.1. If General -> forward the query to a React agent with basic tools and chat memory.
2.2. If Action request -> summarize the request with a parsed output chain -> forward the output to another chain along with a retrieved context for action execution (e.g. execute some arbitrary python code, clean up a desktop, call some API etc.).
```

I can't use a simple react agent with plenty of tools for that, because without the classification and summarization step, agent fails to understand what to do. I must distinguish some basic stuff from the complex, dedicated stuff which may involve many steps to execute properly. Ofc, I could have two or more agents and determine which to ask in the beginning, but it's not what I want to achieve.

Thank you!

贡献指南

这个仓库没有索引到贡献指南

评估

这个 Issue 还没有评估数据。

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。