HR AI Assistant

A full-stack, serverless HR assistant that answers policy questions and retrieves employee records using Retrieval-Augmented Generation (RAG) and structured data querying.

Run at: https://hrassistant1.netlify.app/

Architecture

• Frontend: Plain HTML5, Tailwind CSS (CDN), deployed on Netlify.
• Backend: FastAPI running serverless on Modal.
• LLM & Embeddings: Groq (llama-3.3-70b-versatile) and Hugging Face sentence-transformers (vector embeddings).
• Vector Store: ChromaDB for policy document retrieval.
• Package Manager: uv

Project Structure

.
├── data/              # HR policy documents and structured employee datasets
├── frontend/          # Web interface
│   └── index.html
├── src/               # Application logic
│   ├── app_modal.py   # Modal serverless entry point
│   └── main.py        # FastAPI routes and RAG/DB logic
├── pyproject.toml     # Project dependencies
└── uv.lock            # Lockfile for reproducible builds

Setup & Local Development

Prerequisites

• Python 3.11+
• uv package manager installed
• Modal CLI configured (modal setup)
• Groq API Key set in Modal secrets (groq-secret)

Running Locally

1. Install dependencies:

uv sync

2. Start the Modal development server:

uv run modal serve src/app_modal.py

3. Open frontend/index.html in your browser or serve it locally.

Deployment

Backend (Modal)

Deploy the backend to Modal serverless hosting:

uv run modal deploy src/app_modal.py

Frontend (Netlify)

1. Update API_URL in frontend/index.html with your Modal production URL.
2. Push changes to GitHub. Netlify will build and host the site automatically.

API Specification

POST /ask

Request Body

{
  "employee_id": "EMP001",
  "question": "How many annual leave days do I have left?"
}

Response

{
  "answer": "You have 18 annual leave days left.",
  "source": "Employee Database (SQL)
"
}