pg_sorted_heap: PostgreSQL extension for sorted heap storage and vector search

For decades, PostgreSQL has relied on the heap for storage, leaving ordering to the index. pg_sorted_heap challenges this by introducing a table access method that physically organizes data. By leveraging primary keys and zone-map pruning, it effectively reduces I/O overhead for range scans and ordered retrievals, treating the table itself as a clustered structure. This is a pragmatic move for workloads where range-based queries dominate and index bloat is a concern. Beyond storage, the extension pushes Postgres further into the AI stack. The integration of HNSW (Hierarchical Navigable Small World) for built-in vector types provides the high-performance approximate nearest neighbor (ANN) search required for modern LLM applications. The most ambitious addition, however, is the GraphRAG tooling. By allowing graph-based retrieval augmented generation directly within the database, it eliminates the typical 'architectural tax' of syncing data between a relational store and a separate graph database. From a technical standpoint, the strength lies in consolidation. The weakness is the inherent complexity of custom access methods; maintaining physical order carries a write-penalty during inserts and updates (fragmentation) that DBAs must manage. However, for engineers tired of managing a fragmented stack of VectorDBs and GraphDBs, the trade-off is compelling. This is a high-value tool for data engineers building RAG pipelines who want to minimize latency and operational overhead. If your application requires tight integration between relational metadata, vector embeddings, and graph relationships, pg_sorted_heap transforms Postgres from a storage layer into a comprehensive retrieval engine.