This repository contains the hands-on labs for the Generative AI with pgvector and Aurora PostgreSQL Workshop. Each lab demonstrates a production-relevant use case for pgvector on Amazon Aurora PostgreSQL, integrated with Amazon Bedrock foundation models.
| Lab | Module | Difficulty | Description |
|---|---|---|---|
| 01 - Semantic Search | Semantic Search and Sentiment Analysis | Beginner | Build a search engine that understands meaning and analyzes customer sentiment using Hugging Face models and Aurora ML |
| 02 - Product Recommendations | Product Recommendations | Beginner | Create a personalized product recommendation engine using Bedrock embeddings and similarity algorithms |
| 03 - RAG | Retrieval Augmented Generation | Intermediate | Implement a Q&A chatbot with accurate, grounded responses using RAG architecture |
| 04 - Movie Recommendations | Aurora ML with Bedrock | Intermediate | Build a movie recommendation system using the aws_ml extension for in-database inference |
| 05 - Blaize Bazaar | E-Commerce Platform | Advanced | Deploy a complete e-commerce platform with AI-powered search and recommendations |
| 06 - Incident Detection | Incident Detection and Remediation | Advanced | Implement intelligent database monitoring with agentic workflows and auto-remediation |
| 07 - Aurora ML Chatbot | Aurora ML Chatbot | Intermediate | Build an AI-powered chatbot that runs inference directly within the database using Aurora ML |
| 08 - Valkey Chatbot | Caching with ElastiCache for Valkey | Intermediate | Build a travel chatbot with semantic caching using Aurora PostgreSQL and ElastiCache for Valkey |
- Getting Started: Labs 01 → 02 → 03 cover the fundamentals of vector search, embeddings, and RAG.
- Advanced Patterns: Labs 05 and 06 explore production-scale architectures with agentic workflows.
- Targeted: Each lab is self-contained — pick the use case most relevant to your needs.
- AWS account with appropriate permissions
- Basic knowledge of PostgreSQL and Python
- 15–30 minutes for environment setup
Follow the guided experience at catalog.workshops.aws/pgvector, which provides a pre-configured AWS environment with all dependencies installed.
git clone https://github.com/aws-samples/aurora-postgresql-pgvector.git
cd aurora-postgresql-pgvectorRefer to each lab's README for specific setup instructions and dependencies.
- Amazon Aurora PostgreSQL with pgvector 0.8.0+
- Amazon Bedrock for foundation models (Titan, Claude)
- Amazon SageMaker for ML model hosting
- AWS MCP Servers for AI-database interactions
- Amazon Bedrock Agents for autonomous workflows
- Vector embeddings (up to 16,000 dimensions)
- HNSW and IVFFlat indexing for approximate nearest neighbor search
- Hybrid search (vector + full-text)
- RAG with response streaming
- In-database ML inference via Aurora ML
- Agentic workflows with auto-remediation
The workshop's Code Editor (VS Code in browser) comes pre-configured with:
- Python 3.11 with ML/AI libraries
- PostgreSQL client tools with pgvector
- AWS CLI and SDKs
- Jupyter notebook support
- AWS Blog: Leverage pgvector and Aurora PostgreSQL for NLP, Chatbots and Sentiment Analysis
- AWS Blog: Supercharging Vector Search with pgvector 0.8.0
- AWS Blog: Database Development with AWS MCP Servers
- PostgreSQL MCP Server Documentation
- pgvector Official Repository
- Amazon Aurora Documentation
- This repository is intended for educational purposes. Sample code should be adapted before production use.
- Running these labs will incur AWS charges. Always clean up resources after completing a lab.
This project is licensed under the MIT-0 License.
