Python Requirements Overview

This document explains the Python packages used in the project, their specific versions, and their roles in the RAG-CAG pipeline.

streamlit>=1.32.0

• Interactive web framework for Python.

• Provides a local UI for document upload and chat.

• Optional (can be removed if using Open WebUI only).

PyPDF2>=3.0.1

• Parses and extracts text from PDF files.

• Useful for fallback parsing when Docling isn’t used or OCR fails.

requests>=2.31.0

• HTTP client library to send and receive web requests.

• Used for:

  • Querying Ollama API

  • Uploading documents to Open WebUI

  • Sending data to internal services

numpy>=1.26.4

• Numerical array library used to support vector math and embeddings.

• Required by sentence-transformers.

sentence-transformers>=2.2.2

• Provides pre-trained transformer models for embedding text.

• Used to generate vector embeddings of document chunks or queries.

langchain>=0.1.16

• Modular framework to build LLM pipelines.

• Used to:

  • Connect retrievers, prompts, and LLMs

  • Build RetrievalQA and RAG-CAG chains

langchain-docling>=0.1.3

• Extension for langchain to integrate with docling.

• Loads structured document chunks (from PDF, DOCX, etc.).

• Enables more accurate text splitting and chunking compared to PyPDF2 alone.

Final  requirements.txt

Contents

streamlit>=1.32.0
PyPDF2>=3.0.1
requests>=2.31.0
numpy>=1.26.4
sentence-transformers>=2.2.2
langchain>=0.1.16
langchain-docling>=0.1.3