An AI-powered FAQ retrieval system that uses semantic search to provide accurate answers to customer questions about CRM software. The system leverages RAG (Retrieval-Augmented Generation) architecture with Google's Gemini 2.0 LLM and advanced embeddings to deliver precise responses from a knowledge base.
- Semantic search using Pinecone vector database
- High-quality text embeddings with HuggingFace's instructor-large model
- Natural language understanding with Google's Gemini 2.0 Flash
- Configurable similarity threshold for accurate retrieval
- Simple web interface built with Streamlit
- LLM: Google Gemini 2.0 Flash
- Embeddings: HuggingFace instructor-large
- Vector Database: Pinecone
- Framework: LangChain
- Frontend: Streamlit
- Data Source: CSV knowledge base
https://www.loom.com/share/94afac7ec63340bfb6fc35404b9d4f9d?sid=2f2d8a19-242d-440d-b0c4-ecdb3f828d0d