Incorporate Reactome User Guidelines into the React-to-me RAG Pipeline #105
Description
Background
React-to-me is a conversational assistant that enables users to interact with the Reactome database using natural language. While the chatbot currently supports querying Reactome data, it does not yet support questions about how to use Reactome itself (e.g. user workflows, best practices, and documentation).
Reactome user guideline documentation is publicly available and updated with each Reactome release. The goal of this task is to integrate these user guidelines into the existing RAG-based React-to-me pipeline, so the chatbot can accurately answer "how-to" questions about Reactome usage.
Objective
Extend the existing React-to-me pipeline to support Reactome User Guidelines as a first-class knowledge source using Retrieval-Augmented Generation (RAG) with LangChain and LangGraph.
Key Resources
- Reactome user guideline PDFs (updated per release):
https://zenodo.org/records/17885158
Key Requirements & Considerations
- The existing project makes extensive use of:
- Retrieval-Augmented Generation (RAG)
- LangChain
- LangGraph
- Reactome user guidelines are currently provided as PDFs via Zenodo.
- These PDFs are not directly linked to Reactome website URLs.
- Responses generated from user guideline content should reference relevant Reactome web pages when possible.
Success Criteria
- The chatbot can answer "how do I use Reactome?"-style questions
- Responses are grounded in official Reactome user documentation
- Responses reference relevant Reactome website pages where applicable
- The solution aligns with the existing LangChain / LangGraph architecture