https://github.com/mirabdullahyaser/reasoning-and-action-react-framework-using-langchain

Implementation of ReAct prompting technique (Reasoning + Acting). The method is designed to enhance the reasoning capabilities of LLMs, allowing them to reason and act intelligently within simulated environments.
https://github.com/mirabdullahyaser/reasoning-and-action-react-framework-using-langchain

actions agents gpt-4 langchain llm llm-agents python react reasoning

Last synced: about 1 month ago
JSON representation

Implementation of ReAct prompting technique (Reasoning + Acting). The method is designed to enhance the reasoning capabilities of LLMs, allowing them to reason and act intelligently within simulated environments.

• Host: GitHub
• URL: https://github.com/mirabdullahyaser/reasoning-and-action-react-framework-using-langchain
• Owner: mirabdullahyaser
• Created: 2024-06-16T10:26:14.000Z (6 months ago)
• Default Branch: main
• Last Pushed: 2024-06-16T11:44:24.000Z (6 months ago)
• Last Synced: 2024-06-16T12:47:17.730Z (6 months ago)
• Topics: actions, agents, gpt-4, langchain, llm, llm-agents, python, react, reasoning
• Language: Jupyter Notebook
• Homepage:
• Size: 10.7 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Reasoning and Action (ReAct)

ReAct, which stands for Reasoning + Acting, is a prompting technique developed by professors from Princeton University in collaboration with Google researchers.

The method is designed to enhance the reasoning capabilities of LLMs, allowing them to reason and act intelligently within simulated environments.

It allows LLMs to interact with external tools, enhancing their decision-making processes. With React, LLMs can understand and generate text, make informed decisions and take actions based on their understanding.

# ReAct Agents in LangChain

The ReAct agent in LangChain is a versatile agent that utilizes the ReAct framework to select the appropriate tool based on its description.

This agent is the most general-purpose action agent available in LangChain and can be highly beneficial in situations where multiple tools are available, and selecting the right tool is time-consuming. The ReAct agent allows you to specify a description for each tool and then automatically selects the appropriate tool based on the description.

# How it works

1. The system receives input from the user.

2. Based on the input, the agent decides whether to use a tool and, if so, what the input to that tool should be.

3. The tool is then called with the appropriate input, and the output of the tool is recorded as an observation.

4. The history of the tool, tool input, and observation is passed back into the agent, which decides the next step.

5. The above steps are repeated until the agent determines that no further tools are needed. At this point, it will respond directly to the user.