I’m working on a Retrieval-Augmented Generation (RAG) project with a friend, and we’re trying to improve its architecture and performance.

Specifically, I’m struggling with:

• Improving retrieval quality and relevance
• Reducing hallucinations in responses
• Structuring the pipeline more efficiently

I’ve implemented [https://ragenginev1.vercel.app/\].

Could someone guide me on best practices or point out what I might be doing wrong? I can share specific code snippets if needed.