PDF Insight

Experiment project with LLM RAG using llama3 and langchain

_{⚠️ This project not supported upload file from UI yet, in order to load PDF, place PDF documents into directory rag/tmp/docs and load from UI}

What is RAG?

• Retrieval-Augmented Generation (RAG)
• Enhance the accuracy and reliability of GenAI models with data from external sources

Why RAG ?

• LLM (ChatGPT, Gemini...) does not know your data
• LLM might not know or give out-of-date answer about knowledege beyond the cut-off point
• LLM does not know about specific knowledge (eg: your company data for customer support, your bespoke software user manual...)
• RAG solves above problems

How RAG works ?

• //TODO

Roadmap

• Load PDFs from directory
• Q&A with context from loaded PDFs
• Stream text to UI
• Add message history (memory)
• Format chat response
• Upload PDF from UI
• Measure performance & reliability
• Support images, tables in PDF

Screenshots

Reference

• https://ollama.com/library/llama3
• https://python.langchain.com/docs/use_cases/question_answering
• https://www.databricks.com/glossary/retrieval-augmented-generation-rag

_{Logo generated with Dream Studio (https://beta.dreamstudio.ai/)}