RAG Document Validation Framework

A lightweight AI system for automatic search, validation, and verification of documentary content. Built for compliance workflows (e.g., ISO 27001) using RAG, vector databases, and LangGraph.

Author & Training

Udemy Certificate — RAG Systems with LangChain & Vector Databases
View Certificate

Features

• Ingests PDF, DOCX, PPTX
• RAG-based question answering using local Llama 3 via Ollama
• Vector stores: Chroma (default), FAISS, Pinecone
• Answer verification (supported / partially / unsupported)
• Requirement validation (yes/no + justification)
• LangGraph pipeline: retrieve → generate → verify → END

Requirements

Install Ollama and pull the models:

ollama pull llama3
ollama pull nomic-embed-text

Setup

1. Clone the repository:

   git clone https://github.com/monaii/Rag-system.git
   cd Rag-system

2. Install dependencies:

   pip install -r requirements.txt

3. Environment Setup (.env): Create a .env file with the following configuration:

   LLM_PROVIDER=ollama
   OLLAMA_MODEL=llama3
   EMBED_MODEL=nomic-embed-text
   VECTOR_DB=chroma

Usage (Notebook)

All functionality is inside the main notebook: RAG-PROJECT.ipynb.

Example usage in the notebook:

# Ingest documents from data/ folder
ingest()

# Chat with your documents
chat("What is the CISO responsible for?")

# Verify answers against context
verify(question, answer, context)

# Validate specific requirements
validate(["The policy must define an incident response procedure."])

# Run the LangGraph pipeline
app_rag.invoke({"question": "What controls are defined?"})

Project Structure

• data/ - Input documents
• vector/ - Chroma DB storage (ignored in git)
• RAG-PROJECT.ipynb - Main project notebook
• requirements.txt - Python dependencies