Pinecone is the leading purpose-built vector database providing managed similarity search for AI applications including text-to-SQL implementations. Free Starter tier includes 100K vectors; Standard and Enterprise tiers with usage-based pricing. The platform provides sub-100ms latency for high-dimensional vector similarity search at scale. Key features include metadata filtering (combine vector search with attribute filters), namespace isolation, serverless scaling, and hybrid search (vector + keyword). Supports dense vectors, sparse vectors (for keyword-like search), and hybrid approaches. Index types include s1 (optimized for cost), p1 (optimized for throughput), and p2 (optimized for latency). Critical for text-to-SQL applications: store schema embeddings, example queries, and documentation for RAG-powered query generation. Integrations with LangChain, LlamaIndex, and major ML frameworks. Page should cover: index type selection guide, metadata filtering patterns, hybrid search configuration, text-to-SQL RAG architecture, pricing model analysis, comparison with Weaviate and Qdrant, and implementation best practices.