postvec-server

Managed PostgreSQL

postvec on managed PostgreSQL

The database needs pgvector 0.8 or newer and a role with CREATE on the application database. The server holds the models and provider keys, runs the worker and serves search. Use a direct database endpoint and give the worker ownership of the source tables or membership in their owning roles.

• Amazon RDS
• Aurora
• Cloud SQL
• Azure Flexible Server
• Supabase
• Neon

1. Install the schema

   Creates the postvec schema as a plain SQL install. The install is transactional and it runs again after a server upgrade to update the schema in place.

   postvec-server managed install \
     --dsn 'postgresql://postvec_worker@db.example/app?sslmode=verify-full' \
     --password-file /etc/postvec-server/database.pw

2. Add the database to the server

   A managed entry starts the sync worker. Several nodes can list the same database; they elect one leader. proxy_port opens the search proxy.

   "managed": [{
     "name": "prod",
     "dsn": "postgresql://postvec_worker@db.example/app?sslmode=verify-full",
     "password_file": "/etc/postvec-server/database.pw",
     "sync": true,
     "proxy_port": 5433
   }]

3. Enable a column and search

   After enabling a column and waiting for its initial embeddings, connect through the proxy to search with text. The proxy embeds the query and forwards vector search to PostgreSQL. The managed guide covers supported query forms and connection settings.

   -- through the proxy port (5433)
   SELECT * FROM postvec.search(
     'public.docs', 'body', 'reset password',
     limit_n => 5
   );