Django + MongoDB Vector Search + LlamaIndex + Voyage AI + GPT-4 demo for asking natural-language questions over global nuclear energy data.

• Converts the Kaggle CSV into one JSON file per country-year row.
• Embeds the JSON records with Voyage AI.
• Stores vectors in MongoDB Atlas.
• Retrieves relevant records with MongoDB Vector Search.
• Uses GPT-4 through LlamaIndex to return grounded, quantitative answers.
• Includes a minimal Django UI for screen-recorded demos.

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create .env at the project root and fill in your values:

DJANGO_SECRET_KEY=replace-with-a-local-dev-secret
DEBUG=True
MONGODB_URI=your-mongodb-atlas-connection-string
MONGODB_DB=nuclear_rag
OPENAI_API_KEY=your-openai-key
VOYAGE_API_KEY=your-voyage-key

For a quick install command in demos:

pip install llama-index pymongo llama-index-vector-stores-mongodb llama-index-embeddings-voyageai voyageai openai python-dotenv django-mongodb-backend certifi

Download global_nuclear_energy_intelligence_1965_2025.csv from Kaggle and place it at the project root. The CSV and generated data/ folder are intentionally ignored by git.

Convert the CSV into JSON documents:

python convert.py

The converter currently exports rows from 2015 onward to keep ingestion fast for demos.

Create a Vector Search index named vector_index on database nuclear_rag, collection nuclear_energy:

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1536,
      "similarity": "cosine"
    }
  ]
}

python ingest.py

python query.py --query "Which countries generated the most nuclear electricity in 2025, and how dependent are they on it?"

The prompt is configured to return a ranked markdown table with these columns: Rank, Country, Nuclear Electricity (TWh), Nuclear Share of Grid (%), and Dependency Tier.

python manage.py migrate
python manage.py runserver

Open http://127.0.0.1:8000/.

• ModuleNotFoundError: No module named 'dotenv':

  pip install -r requirements.txt
  python query.py --query "test"

• If python points to the wrong interpreter, always run scripts via .venv/bin/python.