IntelliHR - RAG-Enabled AI Assistant with MCP Architecture

An intelligent HR assistant that orchestrates multiple data sources using Model Context Protocol (MCP) and Retrieval-Augmented Generation (RAG) to provide accurate, context-aware responses.

Built as a capstone project to demonstrate practical AI/ML system design and full-stack development.

Table of Contents:

• Highlights
• Architecture
• Demo Screenshots
• Tech Stack
• Features
• Installation
• Project Structure
• Future Roadmap
• Contribute

---

Highlights

• Smart Orchestration - LLM intelligently routes queries to appropriate data sources
• RAG Integration - Semantic search over policy documents with zero hallucination
• Beautiful UI - Modern glassmorphism design with dark theme
• Fast Responses - Sub-2 second query processing with Groq's llama-3.3-70b
• Context-Aware - Actually reads and respects document confidentiality markers
• Real-time Analytics - Track queries, tools used, and system performance

• Back to the top

---

Architecture

┌─────────────────────────────────────────────┐
│          User Interface (Streamlit)         │
│  Glassmorphism UI • Dark Theme • Analytics  │
└────────────────┬────────────────────────────┘
                 │
                 ▼
┌─────────────────────────────────────────────┐
│     Orchestrator (Groq LLM)                 │
│  llama-3.3-70b-versatile                    │
│  • Function Calling                         │
│  • Tool Selection                           │
│  • Response Generation                      │
└────────────────┬────────────────────────────┘
                 │
        ┌────────┴─────────┐
        │  MCP Framework   │
        └────────┬─────────┘
                 │
    ┌────────────┼────────────┐
    │            │            │
    ▼            ▼            ▼
┌─────────┐ ┌──────────┐ ┌─────────┐
│Database │ │Filesystem│ │   RAG   │
│ Server  │ │  Server  │ │ Server  │
├─────────┤ ├──────────┤ ├─────────┤
│ SQLite  │ │Text Files│ │ChromaDB │
│Employee │ │Announce- │ │Policies │
│Records  │ │ments     │ │(PDFs)   │
└─────────┘ └──────────┘ └─────────┘

How it works:

• User Query → Enters question in natural language
• Orchestrator → LLM analyzes and selects appropriate tools
• MCP Servers → Execute queries on respective data sources
• Response → LLM combines results into coherent answer

• Back to the top

---

Demo Screenshots

Main Page:

You may select questions from quick actions or ask your own:

Querying:

More Querying:

Analytics Dashboard:

---

Tech Stack

Core Technologies:

• Python 3.11+ - Primary language
• Streamlit - Web UI framework
• Groq API - LLM inference (llama-3.3-70b-versatile)
• SQLite - Employee database
• ChromaDB - Vector database for RAG
• LangChain - RAG pipeline orchestration

Key Libraries:

• langchain - Document processing and RAG
• langchain-groq - Groq LLM integration
• langchain-huggingface - Embeddings (all-MiniLM-L6-v2)
• chromadb - Vector storage
• pypdf - PDF document loading
• asyncio - Asynchronous operations

Design:

• Custom CSS with glassmorphism
• Responsive design
• Dark theme with purple gradient
• SVG architecture diagrams

• Back to the top

---

Features

Current Features (v1.0)

✅ Multi-Source Querying

• Employee database search (name, department, ID)
• Announcement retrieval (holidays, events, updates)
• Policy document search with RAG

✅ Intelligent Routing

• LLM automatically selects correct data source
• Multi-tool queries supported
• Context-aware responses

✅ Professional UI

• Real-time chat interface
• Session statistics tracking
• Tool usage visualization
• Query history with expandable details

✅ Data Sources

• 10 employee records
• 4 announcement files
• 3 policy documents (Leave, POSH, Salary)

System Capabilities

• 9 Tools across 3 MCP servers

• Sub-2s average response time

• Zero hallucination with RAG

• Conversation memory for follow-up questions

• Back to the top

---

Installation

Prerequisites:

• Python 3.11 or higher
• Groq API key (Get one free)

Step 1 : Clone the repository

git clone https://github.com/yourusername/intellihr-rag-mcp.git
cd intellihr-rag-mcp

Step 2 : Create virtual environment

python -m venv .venv

# Activate
# Windows:
.venv\Scripts\activate
# Linux/Mac:
source .venv/bin/activate

Step 3 : Install Dependencies

pip install -r requirements.txt

Step 4 : Set up environment variables

# Windows PowerShell
$env:GROQ_API_KEY="your-groq-api-key-here"

# Linux/Mac
export GROQ_API_KEY="your-groq-api-key-here"

Step 5 : Initialize the Database

python setup_database.py

Step 6 : Run the mcp servers to initialize them

python mcp_servers/database_server.py
python mcp_servers/rag_server.py
python mcp_servers/file_system_server.py

Step 7 : Run the orchestrator

python orchestrator.py

Step 8 : Run the Application

streamlit run app.py

The app will open in your browser at http://localhost:8501

• Back to the top

---

Project Structure

intellihr-rag-mcp/
├── app.py                      # Main Streamlit application
├── orchestrator.py             # LLM orchestrator for tool routing
├── setup_database.py           # Database initialization script
├── requirements.txt            # Python dependencies
│
├── mcp_servers/               # MCP Server implementations
│   ├── database_server.py     # Employee database server
│   ├── filesystem_server.py   # Announcements server
│   └── rag_server.py          # Policy documents RAG server
│
├── ui/                        # UI components
│   ├── __init__.py
│   └── styles.py              # Custom CSS styles
│
├── data/                      # Data storage
│   ├── announcements/         # Text files for announcements
│   ├── policies/              # PDF policy documents
│   └── chroma_store/          # ChromaDB vector database
│
├── employees.db               # SQLite database (created after setup)
└── README.md                  # This file

• Back to the top

---

Future Roadmap

• Enhanced RAG - Vector search for announcements (eliminate hallucination)
• Voice Interface - Speech-to-text and text-to-speech
• Multi-tenancy - Department-specific access control
• Advanced Analytics - Query insights and usage patterns
• API Endpoints - REST API for programmatic access
• Mobile Responsive - Optimized mobile experience
• Export Features - Download chat history and reports
• Admin Dashboard - Manage documents and users
• Fine-tuned Models - Custom model for domain-specific queries
• Docker Deployment - Containerized deployment

Under Consideration:

• Integration with Slack/Teams
• Email notification system
• Document auto-ingestion pipeline
• Multi-language support
• Real-time collaboration features

• Back to the top

---

Contribution

Contributions are welcome! Here's how you can help:

• Fork the repository
• Create a feature branch (git checkout -b feature/AmazingFeature)
• Commit changes (git commit -m 'Add AmazingFeature')
• Push to branch (git push origin feature/AmazingFeature)
• Open a Pull Request

Areas for Contribution

• UI/UX improvements

• Additional data source integrations

• Performance optimizations

• Bug fixes

• Documentation improvements

• Test coverage

• Back to the top