https://github.com/asvch/ask_docs

LLM AI chatbot using Advanced Retrieval Augmented Generation (RAG), Langchain, and Streamlit to answer questions about information contained in numerous files.
https://github.com/asvch/ask_docs

chromadb langchain rag ragas streamlit

Last synced: about 2 months ago
JSON representation

LLM AI chatbot using Advanced Retrieval Augmented Generation (RAG), Langchain, and Streamlit to answer questions about information contained in numerous files.

• Host: GitHub
• URL: https://github.com/asvch/ask_docs
• Owner: asvch
• License: mit
• Created: 2024-06-17T01:12:54.000Z (7 months ago)
• Default Branch: main
• Last Pushed: 2024-07-15T18:59:38.000Z (6 months ago)
• Last Synced: 2024-07-15T23:11:45.797Z (6 months ago)
• Topics: chromadb, langchain, rag, ragas, streamlit
• Language: Python
• Homepage:
• Size: 4.43 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
## Advanced Retrieval Augmented Generation (RAG) chatbot to ask questions about the rules of Cricket!

  - LangChain, Streamlit, and OpenAI API

  - Query rewriting (pre-retrieval) and prompt compression (post-retrieval)

  - Chroma database as external vectorstore

  - Ragas metrics to evaluate system

     - Answer Relevancy score of 97%, Context Precision score of 97%, and Context Recall score of 100%

   

  Note: Can be easily modified to Q&A chatbot for any set of documents.

Preview:

  - Only answers questions relating to documents, and does not hallucinate answers to unrelated questions: 

![Advanced_RAG_demo](https://github.com/asvch/ask_cricket/assets/66492476/5118cfc5-60a2-415b-98bf-6053d7ac702d)

All PDF documents in data folder containing the rules of Cricket are from https://www.mumbaicricket.com/mca/Laws_of_Cricket.php, and more documents can be added to folder to expand knowledge of application.