ReasonGraph is an open-source web platform designed to visualize and analyze the reasoning processes of Large Language Models (LLMs). It targets researchers and developers working with LLMs, offering an intuitive UI to transform text-based reasoning outputs into interactive flowcharts, thereby enhancing understanding and debugging of complex inference paths.

How It Works

ReasonGraph converts LLM reasoning outputs, including sequential and tree-based methods like Chain of Thoughts and Tree of Thoughts, into visual flowcharts. Its modular framework allows for easy integration of new reasoning methods and LLM providers. The platform also supports "meta-reasoning," enabling models to self-select optimal reasoning strategies, and handles extended reasoning outputs for deeper analysis.

Quick Start & Requirements

• Install: Clone the repository and set up the environment.
• Prerequisites: Python 3.11.8, requests, openai, together, flask, google, google-genai, google-generativeai. API keys for supported LLM providers (Anthropic, OpenAI, Google, Grok, Deepseek, Qwen, Together.AI) are required for inference.
• Run: python app.py
• Demo: An online demo is available on Hugging Face Spaces: https://huggingface.co/spaces/ZongqianLi/ReasonGraph

Highlighted Details

• Supports visualization of over 50 state-of-the-art LLMs.
• Implements mainstream sequential and tree-based reasoning methods.
• Features a modular framework for easy extension.
• Offers multilingual support and an intuitive UI.

Maintenance & Community

• The project is associated with ACL 2025.
• A Discord group is available for community interaction.

Licensing & Compatibility

• The license is not explicitly stated in the provided README snippet.

Limitations & Caveats

• The project is presented as a demo for a paper, suggesting it may be experimental. API keys are required for full functionality, and their management is handled via a local api_keys.json file.