This project is a practical application of the concepts discussed in Enhancing RAG-based application accuracy by constructing and leveraging knowledge graphs*. It aims to advance the original implementation by integrating its logic into a basic application framework.
- Incorporation of Knowledge Graphs into Agent RAG for enhanced performance.
- Use of the Tavily search engine as an auxiliary tool.
- Simple deployment via LangServe.
- Graph Database: We employ Neo4J for our graph database needs.
- Model: We utilize two models,
gpt-4-0125-previewfor generative tasks, andtext-embedding-3-smallfor embeddings. - App Deployment: The application is hosted on a
t3.xlargeEC2 instance, ensuring robust performance. - App Construction: Development was carried out using LangServe, facilitating a streamlined build process.
Below are the main components of our application, along with links to their implementation:
| Component | Link |
|---|---|
| Application | serve.py |
| Step by Step Implementation | graph_retrieval_playground.ipynb |
Interact with the application via the following URL: http://ec2-54-80-98-106.compute-1.amazonaws.com:8001/agent/playground/. To get a feel for its capabilities, try out these queries:
- "What are the Generative AI Practice Requirements?" - This query demonstrates the use of the Enhanced RAG with Knowledge Graphs.
- "Who won the 2001 Copa America?" - This query showcases the integration with TavilySearchResults.