Easy GraphRAG Queries 💡

This app came about to make a research project a little tidier and for people more comfortable with their browser than a Python IDE 🐍.

It wraps the Global and Local search methods of GraphRAG v0.3.6 in a Streamlit UI. It was made for exploring a GraphRAG knowledge graph in a local (private) directory, from the comfort of a browser. It works with the OpenAI API 🤖.

You can select and configure the Global and Local search engines, submit queries, and view the results in your browser. You can also view and download supporting AI-generated reports, source document references, as well as the results object.

Why bother?

• User-friendliness: Streamline querying and saving the outputs of research projects from a browser.
• Referencing: Approximate conventional referencing for academic or professional research projects.

Key features:

• Search configuration: Select a search method, the preferred OpenAI model, and how the response should be presented.
• References: The generated response references the most-used source documents for exploration and corroboration.
• Supporting analyses: Get AI-generated reports to support the response.
• Download results: Download query results, analyses, and sources.

How to use it:

1. Clone the repo: Clone this repo to your local GraphRAG project directory.
2. GraphRAG pipeline: If you haven't already, follow the GraphRAG docs to initiate your project environment and create your knowledge graph.
3. Install requirements: Install any requirements you might need in a virtual environment.
4. Run the app: Run the app in your GraphRAG project directory:

streamlit run Graph_query.py

5. Configure search: Enter your API key, choose a model, and specify the data source (a GraphRAG pipeline output folder).
6. Submit query: Type your query and submit it.
7. View results: Click on the expanders to see the response, supporting analyses, and sources.
8. Download: Download the results for your research records or to explore later.

Requirements

Install the required dependencies in a virtual environment if needed:

pip install -r requirements.txt

Acknowledgements

This app was inspired by the Microsoft GraphRAG project.

License

This app is open source and published under the MIT License. There will be no maintenance but look forward to people improving it!

Happy searching! ✨