RAGHub

LLM-RAG solutions for various usecases

Tools Used:

• Python (Programming Language)
• LangChain (Framework)
• OpenAI GPT 4o Mini (LLM)
• Gradio (UI)
• Chroma DB (Vector Database)

Error Resolution in RAG:

Evaluate Chunking Strategy is the key

• Send entire documents as a context
• Have larger/smaller chunk sizes
• Increase/Decrease overlap

1. Simple RAG

This is a naive RAG implementation to help customers understand the background of employees and the products offered by an Insurance Company. Some improvements:

• Error: Unable to retrieve info present in the database.
• Solution: Changing the number of retrieved chunks to 25 compared to the default.

Output using Default Config

Output after Refinement