HomeTechnologyPostgreSQL

PostgreSQL in use

PostgreSQL as the only database, vector search included: pgvector with an HNSW index instead of a second service alongside it. One backup, one transaction space, one place the data lives.

Full-stack development3 case studies

A long plan table alone in a hangar, one large schema sheet unrolled across it, a person reading it against the open gateAI-generated
One schema, one place the data lives — read before anything is written.

Proven by

3 published projects with PostgreSQL — each with stack, role and result.

What that means in practice

One database instead of two: in the profile generator, text and embeddings sit in the same instance and vector search runs over pgvector with an HNSW index — one backup, one transaction space. Schema changes run as migrations with a way back. Prisma and raw SQL sit side by side, because a query plan can only be read off the second.

  • Schema design and migrations without downtime

  • pgvector with HNSW for semantic search

  • Full text over tsvector, hybrid with vectors

  • Prisma and raw SQL side by side

One server rack standing alone on the hangar floor, doors open, its cable run dropping into a floor ductAI-generated
One database, one backup, one transaction space — and one cable to it
A sheet clipped to a drafting board in the hangar, one line of it underscored by handAI-generated
The setting you learn after the first deploy — and then write down
A creased sheet on the hangar floor inside a single lamp’s pool of light, one block struck throughAI-generated
The query that worked until the table grew — struck through, kept

Other technologies

Contact

Half a minute, and you know more.

Half a minute on what I work on and how. If a question is left, write to me — an answer within one working day.