AI Engineer | Data Engineer | Python Developer | Application Architect
🌐 Portfolio | 💼 LinkedIn

A Large Language Model (LLM) is an advanced AI system trained to understand and generate human-like text.
It learns from massive public datasets such as Wikipedia, books, and websites.

Key capabilities:

• Answering questions
• Summarizing content
• Translating languages
• Engaging in human-like conversation

Examples: ChatGPT, Google Gemini, Claude, LLaMA

While LLMs are powerful, they come with limitations in business environments:

• No access to internal documents, company policies, or private databases
• Cannot provide accurate, up-to-date answers about internal or proprietary information
• Lack of context about your organization limits their usefulness in real-world applications

Retrieval-Augmented Generation (RAG) bridges the gap between LLMs and enterprise knowledge:

• Retrieves relevant internal data in real-time
• Augments the LLM's responses with accurate, business-specific context
• Delivers precise and trustworthy answers based on your own content

Result: Smarter, enterprise-ready AI that truly understands your business

This project implements a Retrieval-Augmented Generation (RAG) system using local PDF documents as the knowledge base. Built with LangChain, ChromaDB, and a local LLM via Ollama, this setup enables question-answering over your own documents using efficient vector search and contextual responses from LLMs.

• PDF Loader: Extracts text from PDF files in your local data/ directory.
• Text Chunking: Splits text into overlapping chunks using RecursiveCharacterTextSplitter to preserve context.
• Vector Store (ChromaDB): Stores and indexes embeddings for fast similarity search.
• Local LLM via Ollama: Generates contextual answers using lightweight models like Mistral, all on your local machine.
• Testing Suite: Validate system accuracy using test queries.

rag_local_pdfs/
│
├── chroma/                    # chroma directory will be added when we run populate_database.py
├── data/                      # Local PDFs for ingestion
├── media/                     # Concept images, documents
├── get_embedding_function.py  # Defines embedding logic
├── populate_database.py       # Loads, chunks, embeds, and stores PDFs
├── query_data.py              # Queries ChromaDB and returns LLM response
├── requirements.txt           # Python dependencies
└── README.md

cd rag_local_pdfs

python -m venv venv
source venv/bin/activate   # on Windows use venv\Scripts\activate
pip install -r requirements.txt

Put your PDF files in the data/ directory for ingestion.

Install Ollama from https://ollama.com, then run: Below step will download model

ollama run llama 3.2

Make sure Ollama is running in the background. By checking http://127.0.0.1:11434/ you should see "Ollama is running" in the page

python populate_database.py

python query_data.py "What is the summary of the document?"

** Local LLM Support**

Uses Ollama for running models like:
-llama3.2
-gemma

Change model name in query_data.py if you prefer a different LLM.

-GUI or Streamlit interface
-Multi-file PDF summarization
-Real-time chat over documents
-RAG + Agent framework (e.g., LangGraph or CrewAI)