Evokoa

Teams building AI applications need their data to be searchable and understandable by AI, but connecting intelligence to the right context is hard. Semantic search often requires costly upfront ingestion and embedding work before anything can be queried. Separately, applications that need graph, vector, and full-text retrieval over existing data typically have to stitch together multiple systems rather than working from a single database foundation like PostgreSQL. This leaves developers building bespoke, fragmented infrastructure just to make their data usable by AI.

Evokoa builds two products addressing this. WonderSearch is a serverless sparse retrieval solution for semantic search that lets users find what people mean, pay per query, and get started without upfront ingestion or embedding fees. Polygres brings graph, vector, and full-text search together as a virtual layer over PostgreSQL, helping applications find relevant context directly from their existing database. Underlying Polygres are two open-source PostgreSQL extensions developed in the open: pgGraph, which brings graph queries and relationship traversal to PostgreSQL, and pgContext, for building AI search and retrieval on a PostgreSQL foundation. The source code and technical documentation for these extensions are available through Evokoa's GitHub organization.