How komalvardhanl-hexaware/docuchat-pro works
Project type and purpose
• This is a full-stack Retrieval-Augmented Generation (RAG) document chat application written in Python.
• Users upload PDF, text, Excel, CSV or SQL files; the app ingests them into a FAISS vector store; at query time, it retrieves relevant chunks and uses Azure OpenAI to generate answers; it also sends traces to Arize Phoenix.
File structure
• Top-level files:
• docs_reader.py: document parsers and vector-store loader
• feature1.py: experimental feature or pipeline extension
• test.py: functional or integration tests
• test.pbix: Power BI file for reporting or demo
• requirements.txt: dependency list (streamlit, faiss-cpu, azure-ai, arize, etc.)
• .gitignore, README.md
• README references a ui/streamlit_app.py, so assume a ui folder contains the Streamlit front end.
• Missing config placeholders: .env for Azure keys and Arize credentials.
README for additional insights
• The architecture involves:
• Streamlit front-end
• Document ingestion pipelines (docs_reader)
• FAISS vector store for retrieval
• Azure OpenAI service for generation
• Arize Phoenix for tracing/evaluation
Final advice
• Essential elements—don’t omit the ingestion pipeline, FAISS store, or tracing agent.
• Keep it detailed but avoid unnecessary sub-subcomponents; focus on the seven or eight major blocks above.