Company Policy Assistant: Local RAG Chatbot with Ollama

A Retrieval-Augmented Generation (RAG) chatbot designed to answer questions about company policies using uploaded PDF or TXT documents. This project demonstrates the use of LangChain, Ollama LLM & embeddings, and Chroma vector database to provide contextual answers with source citations.

🏆 Features

✅ Upload PDF or TXT files containing company policies.

✅ Semantic search with embeddings + vector database (Chroma).

✅ LLM-based answer generation using Ollama.

✅ Retrieves relevant document chunks and cites sources.

✅ Maintains conversation history within the session.

🌐 Interactive web UI built with Streamlit.

🎯 Objectives

This project is designed to showcase skills in:

1. Document ingestion and chunking.
2. Embedding & semantic search for relevant content.
3. Retrieval-Augmented Generation (RAG).
4. Session-based conversation memory.
5. Interactive, user-friendly UI for querying documents.

🛠 Tech Stack

Component	Tool / Library
LLM & Embeddings	Ollama
Vector Store	Chroma
Python Framework	LangChain
Web Interface	Streamlit
Document Loaders	PyPDFLoader (PDF), TextLoader (TXT)

⚙️ Installation & Libraries

1. Install Ollama

Download here

2. Pull Models

ollama serve
ollama pull nomic-embed-text
ollama pull llama3.2

3. Clone the Repository

git clone https://github.com/MirazulHasan/COMPANY_POLICY_ASSISTANT_CHATBOT.git
cd company-policy-chatbot

4. Create a Virtual Environment

python -m venv venv

Activate the environment

Linux / macOS:

source venv/bin/activate

Windows:

venv\Scripts\activate

5. Install Required Python Libraries

You can install all dependencies using pip:

pip install -r requirements.txt

Or manually:

pip install streamlit
pip install langchain
pip install langchain_community
pip install langchain_text_splitters
pip install langchain_chroma
pip install langchain_ollama
pip install chromadb
pip install pypdf

Python 3.11+ recommended.

6. Run Ollama Locally

ollama serve

Required for embeddings and chat.

7. Start the Streamlit App

streamlit run policy_chatbot.py

Open your browser: http://localhost:8501

📝 Usage

1. Upload Policy Documents

• Click "Upload Policy Docs" in the sidebar.

• Accepts PDF or TXT files.

2. Ask Questions

Use the chat input to ask anything like:

What is the company’s leave policy?

How can employees request remote work?

3. View Sources

• Answers come only from uploaded docs.

• Click "View Sources" to see file + page + snippet.

4. Conversation History

• Maintained within the session.

• Each message shows in the chat interface.

🧩 Sample Policy Documents (for testing)

• Local Initiatives Support Corporation

• Template.net

• The People in Dairy

• MD Accountants & Auditors Inc.

💡 How it Works

1. Load Documents → Read PDFs/TXTs
2. Split → Chunk into 1000-char segments with overlap
3. Embed → Convert to vectors using nomic-embed-text
4. Store → In Chroma vector DB
5. Query → Retrieve top-4 relevant chunks
6. Generate → LLM answers using only retrieved context
7. Cite → Show source file + page + snippet
8. Chat → Session memory preserved

📁 Files

• policy_chatbot.py – Main Streamlit + RAG app
• requirements.txt – Python dependencies

📌 Notes

• Ensure nomic-embed-text and llama3.2 are pulled via Ollama.
• Session memory only (in-browser). For persistence: add SQLite or JSON logging.
• Fully local & private — no data leaves your machine.

📚 References

• LangChain Documentation

• Chroma Vector Database

• Ollama

• Streamlit Docs

💻 License

© 2025 Md. Mirazul Hasan

All Rights Reserved.

For educational and internal use.