Build a retrieval-augmented generation (RAG) application in a 90-minute, hands-on workshop. The notebook uses MongoDB Atlas Vector Search to retrieve relevant MongoDB documentation, Voyage AI's voyage-4-lite model to create embeddings, and the Gemini API to generate grounded answers.
| Time | Activity |
|---|---|
| 0–10 min | Review the RAG pipeline and verify cloud credentials |
| 10–25 min | Load and chunk the MongoDB documentation dataset |
| 25–45 min | Generate Voyage embeddings and store them in Atlas |
| 45–60 min | Create a Vector Search index and retrieve relevant chunks |
| 60–78 min | Send retrieved context to Gemini and evaluate answers |
| 78–85 min | Explore conversation history in MongoDB |
| 85–90 min | Review the pipeline and next steps |
Complete the prerequisites guide before the event. You will create a Gemini API key, an Atlas M0 cluster, a database user, and an Atlas Model API Key for Voyage AI.
Install Python and uv using the Python + uv guide. Prepare VS Code with the Python and Jupyter extensions.
| File | Purpose |
|---|---|
rag_pipeline.ipynb |
End-to-end RAG lab |
prerequisites.md |
Cloud accounts, API keys, and notebook preparation |
python-uv-setup.md |
Python and uv installation |
pyproject.toml |
Python dependencies |
.env.example |
Environment variable template |
data/mongodb_docs.json |
MongoDB documentation sample dataset |