https://github.com/fx2y/reflexion-go

[WIP] Reflexion Agent: A Go implementation of an advanced AI-powered tool-calling agent using Temporal workflows and instructor-ai. Demonstrates self-reflection and iterative improvement in AI responses.
https://github.com/fx2y/reflexion-go

ai go nlp opentelemetry reflexion search temporal tool-calling workflow

Last synced: about 7 hours ago
JSON representation

[WIP] Reflexion Agent: A Go implementation of an advanced AI-powered tool-calling agent using Temporal workflows and instructor-ai. Demonstrates self-reflection and iterative improvement in AI responses.

• Host: GitHub
• URL: https://github.com/fx2y/reflexion-go
• Owner: fx2y
• Created: 2024-08-16T08:27:07.000Z (about 1 month ago)
• Default Branch: main
• Last Pushed: 2024-08-16T10:23:15.000Z (about 1 month ago)
• Last Synced: 2024-09-24T08:33:42.184Z (4 days ago)
• Topics: ai, go, nlp, opentelemetry, reflexion, search, temporal, tool-calling, workflow
• Language: Go
• Homepage:
• Size: 20.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Reflexion Agent

## Overview

Reflexion Agent is a powerful tool-calling agent implemented using the Temporal Go SDK and instructor-ai/instructor-go. It demonstrates an advanced AI-powered workflow that combines initial response generation, self-reflection, and information retrieval to provide comprehensive and accurate answers to user queries.

## Features

- Implements the Reflexion architecture for improved AI responses

- Utilizes Temporal workflows for robust and scalable execution

- Integrates instructor-ai for AI-powered decision making

- Includes a mock search service for demonstration purposes

- Implements OpenTelemetry tracing for observability

## Implementation Details

The Reflexion Agent is implemented as a Temporal workflow with three main activities:

1. Initial Responder: Generates an initial answer, reflection, and search query.

2. Search: Performs a search based on the generated query.

3. Revisor: Revises the initial answer based on search results.

The workflow orchestrates these activities to produce a final, refined answer to the user's question.

## Configuration

- Update the `instructor.NewClient("your-api-key")` in `main.go` with your actual API key.

- Modify the `MockSearchService` in `mock_search.go` to integrate with a real search service if needed.

## Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

## License

This project is licensed under the MIT License - see the LICENSE file for details.