Document Portal is an interactive, AI-powered platform that demonstrates Retrieval-Augmented Generation (RAG) workflows on PDF documents. It enables users to:
- Analyze documents and extract key metadata.
- Ask questions about one or multiple documents using a conversational AI interface.
- Compare different versions of documents to quickly spot changes.
- Operate with robust industry-standard logging, error handling, and session management.
This application is built for research, knowledge management, and enterprise workflows, offering a clean, intuitive interface for document-centric AI operations.
-
Document Analysis & Chat Upload PDFs to extract text and metadata, and interactively query their content using natural language. Supports both single-document and multi-document workflows.
-
Document Comparison Upload two versions of a document and instantly detect differences, making version control and review easier.
-
Efficient & Scalable RAG Pipeline Uses FAISS for vector storage and LangChain for retrieval and answer generation. Existing indexes are reused to reduce processing time.
-
Interactive UI Built with Streamlit, providing a responsive and minimal interface for smooth user interaction.
[Upload PDFs] --> [Data Ingestion & Preprocessing] --> [FAISS Vectorization] --> [RAG Query] --> [Answers / Comparison]
Step-by-step workflow:
- Upload Documents – Users can upload one or multiple PDFs.
- Data Ingestion & Processing – Documents are parsed and structured for embeddings.
- Vectorization – FAISS creates vector representations for fast similarity search.
- RAG Pipeline – LangChain retrieves context and generates natural language responses.
- Output – Answers, analysis, or document differences are displayed interactively.
Document Chat Example:
Document Comparison:
Document Summarizer:
Project Structure Overview:
project-name/
│
├── .github/
│ └── workflows/ # GitHub Actions CI/CD workflows
│ ├── aws.yml # AWS deployment workflow
│ ├── task_definition.json # ECS or container task definition
│ └── template.yml # Template for workflows
│
├── .idea/ # PyCharm IDE config files (can be ignored in Git)
│ └── ...
│
├── api/
│ └── main.py # Main API entrypoint (FastAPI/Flask)
│ └── ... # Other API endpoints, routers
│
├── archive/
│ └── src/ # Old or backup source code for reference
│ └── ...
│
├── config/
│ └── config.yaml # Core configuration settings
│ └── ... # Additional config files (JSON/YAML)
│
├── data/
│ ├── doc_analysis/ # Data for document analysis workflow
│ ├── multidoc_chat/ # Data for multi-document chat workflow
│ └── single_doc/ # Data for single-document chat workflow
│
├── exception/
│ ├── __init__.py # Makes folder a Python package
│ └── Custom_exception.py # Custom exception classes for robust error handling
│
├── faiss_index/
│ ├── index.faiss # FAISS vector index for embeddings
│ ├── index.pkl # Serialized FAISS object for fast loading
│ └── ... # Any session or auxiliary files
│
├── logger/
│ ├── __init__.py # Python package initializer
│ └── custom_logger.py # Logging utilities for debugging & monitoring
│
├── model/
│ ├── __init__.py # Python package initializer
│ └── models.py # Trained ML/DL models or model utilities
│
├── notebook/
│ └── ... # Jupyter notebooks for experiments and testing
│
├── prompts/
│ ├── __init__.py # Python package initializer
│ └── prompts.py # Prompt templates and utilities for RAG pipelines
│
├── src/
│ ├── document_analyzer/
│ │ ├── __init__.py
│ │ └── data_analysis.py # Core logic for document analysis
│ │
│ ├── document_chat/
│ │ ├── __init__.py
│ │ └── retrieval.py # Single & multi-document chat workflows
│ │
│ ├── documents_compare/
│ │ ├── __init__.py
│ │ └── document_comparator.py # Logic to compare different versions of documents
│ │
│ └── document_ingestion/
│ ├── __init__.py
│ └── data_ingestion.py # Parsing & preprocessing documents for RAG
│
├── static/
│ └── style.css # Frontend CSS styling
│
│
├── templates/
│ └── index.html # HTML templates for FastAPI/Flask or frontend rendering
│
├── utils/
│ ├── __init__.py # Python package initializer
│ ├── config_loaders.py # Load and manage configuration files
│ ├── document_ops.py # Helper functions for document processing
│ ├── file_io.py # File reading/writing utilities
│ └── model_loader.py # Load ML/DL models efficiently
│
├── .dockerignore # Docker ignore rules
├── .gitattributes # Git attributes
├── .gitignore # Git ignore rules
├── Dockerfile # Docker container setup
├── README.md # Project README (documentation)
├── app.py # Main app entrypoint (Streamlit/FastAPI/Flask)
├── requirements.txt # Python dependencies
├── setup.py # Package setup script
└── test.py # Test scripts or Streamlit UI for quick prototyping
- Python 3.10+
- Streamlit – Interactive frontend
- LangChain – RAG orchestration
- FAISS – Vector database for embeddings
- Custom Modules – Document ingestion, analysis, retrieval, and comparison
-
Clone the repository:
git clone https://github.com/sayed-ashfaq/document-RAG-app.git cd document-portal -
Install dependencies:
pip install -r requirements.txt
-
Run the application:
streamlit run app.py
-
Open the Streamlit interface and start uploading PDFs, chatting with documents, or comparing versions.
- Existing FAISS indexes are reused for faster querying.
- Designed for both single and multi-document workflows in a single interface.
- Ideal for research, document management, and AI-powered knowledge extraction.
- Easily extendable with new RAG workflows or document types.