durcno@1.0.0-alpha.14

Minor Changes

• d2a23ad - Thanks to @almahdi404 !

  feat: add pgvector extension support

  Adds complete pgvector extension support to Durcno, enabling vector similarity search and distance calculations.

  New Features

  • pgvector column type: New pgvector columns for storing and querying vector embeddings
  • Distance functions: L2, L1, Hamming, and cosine distance operators for vector comparisons
  • Vector operations: Support for pgvector distance, similarity search, and filtering
  • Index opclass support: Ability to specify custom operator classes for indexes to support pgvector and other specialized index types

  Usage

  Define vectors in your schema and perform similarity searches:

  import { pgvector, pk, table } from "durcno";
  
  export const Records = table("public", "records", {
    id: pk(),
    embedding: pgvector.vector({ dimensions: 1536 }),
  });
  
  // Query with distance operations
  const results = await db
    .from(Records)
    .select()
    .where(lt(l2Distance(Records.embedding, queryVector), 1.0)) // threshold
    .orderBy(asc(l2Distance(Records.embedding, queryVector)))
    .limit(10);

• 58e60e0 - Thanks to @almahdi404 !

  change(connectors): reduce default connection pool size from 10 to 5

  The default maximum connection pool size (DEFAULT_POOL_MAX) has been lowered from 10 to 5.

  This change affects all connectors (pg, postgres, pglite, bun) when no explicit pool.max is configured. The previous default of 10 was unnecessarily high for most use cases and could exhaust database connection limits in environments with many application instances.

  Users who need more than 5 concurrent connections should set pool.max explicitly:

  connector: pg({
    dbCredentials: { url: process.env.DATABASE_URL! },
    pool: { max: 10 },
  });