- PostgreSQL vector extensions (pgvector, pgvectorscale) - architecture and capabilities
- Indexing strategies: HNSW vs IVFFlat for different scale requirements
- Hybrid search patterns: combining vector similarity with traditional SQL filters
- Integration with popular frameworks: LangChain, LlamaIndex, Semantic Kernel
- Schema design for embeddings, metadata, and document chunks
- Performance optimization: query tuning, indexing strategies, and connection pooling
- Production considerations: monitoring, scaling, backup strategies

Through live demos, we'll build a complete RAG pipeline:
1. Document ingestion and chunking strategies
2. Generating embeddings with open models (sentence-transformers, BAAI)
3. Storing and querying vectors in PostgreSQL
4. Implementing semantic search with metadata filtering
5. Integrating with LLM frameworks for context retrieval

We'll compare approaches across different scales - from single-node deployments
to distributed setups, and discuss when PostgreSQL makes sense vs. specialized
vector databases. All code will be shared in an open-source repository with
Docker Compose setup for local development.

Key Takeaways:
- Understand vector search fundamentals and indexing algorithms
- Build production-ready RAG systems using open standards
- Make informed decisions about vector database architecture
- Leverage existing PostgreSQL expertise for AI applications
- Avoid common pitfalls in embedding storage and retrieval

Prerequisites: Basic SQL knowledge, familiarity with LLMs and embeddings (but
we'll cover the essentials)