Neon Docs search is now built on Neon

Neon has replaced its hosted documentation search vendor with an in-house system built on Lakebase Postgres, Lakebase Search, Neon Functions, and Neon AI Gateway. Lakebase Search combines pgvector-compatible semantic search through the lakebase_vector extension with BM25 keyword ranking through lakebase_text, enabling hybrid queries that merge vector and text results using reciprocal rank fusion. Content changes trigger a GitHub Action that sends pages to a Neon Function, which retrieves markdown versions of pages, chunks them by heading, generates embeddings, and stores text, vectors, and search indexes in Postgres. To support responsive search-as-you-type behavior, Neon serves cached hybrid results for common queries while using fast BM25-only search for uncached queries, avoiding slower model calls during live requests. The company reports server-side median latencies of 1.9 milliseconds for cache hits and 8.6 milliseconds for BM25 searches, while its new system returned the expected page within the top five results for 94% of 171 real queries, compared with 80% for the prior search service.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.