@vibeorm/extensions 2.0.0-alpha.8 → 2.0.0

package/README.md

@@ -3,10 +3,10 @@
 > A type-safe TypeScript ORM for Bun and Node. Prisma-schema or TypeScript-DSL input, a canonical schema IR, a generated client that your editor never re-infers, and a dialect-aware SQL layer over PostgreSQL, PGlite, SQLite and MySQL.
 
 ```bash
-bun add @vibeorm/extensions@alpha
+bun add @vibeorm/extensions
 ```
 
-> **Pre-release.** `2.0.0-alpha.x`, published under the `alpha` dist-tag. `npm i vibeorm` still resolves to the stable v1 line, which has a different API — see the migration guide below.
+> **Stable.** `2.0.0`, published under the `latest` dist-tag. The v1 line stays installable as `vibeorm@1`, and is a different API — see the migration guide below.
 
 ## What it provides
 
@@ -99,6 +99,45 @@ The query is split into unicode letter/number tokens (punctuation separates and
 - `engine: "bm25"` refuses `prefix` (`VIBE_UNSUPPORTED_CAPABILITY`) — ParadeDB prefix behavior is decided by the index tokenizer, which this extension does not configure.
 - mysql InnoDB ignores tokens shorter than `innodb_ft_min_token_size` (default 3), so one- and two-letter prefixes match nothing there.
 
+### Filtering hits: `where` composes with the core filter language
+
+`search({ where })` compiles through the same filter compiler as every core query — scalar
+filters, relation filters and `dbNow()` comparisons all AND into the match. The menu
+composition (type-ahead over items that are orderable *right now*):
+
+```prisma
+model MenuItem {
+  id            Int       @id @default(autoincrement())
+  name          String
+  description   String
+  availableFrom DateTime?
+  availableTo   DateTime?
+  category      Category? @relation(fields: [categoryId], references: [id])
+  categoryId    Int?
+
+  @@search([name, description], weights: [2, 1], language: "simple")
+}
+```
+
+```ts
+import { dbNow } from "@vibeorm/runtime";
+
+const hits = await db.menuItem.search({
+  query: "piz",              // mid-keystroke
+  prefix: true,
+  take: 8,
+  where: {
+    category: { is: { active: true } },
+    availableFrom: { lte: dbNow() },   // the DATABASE's clock, not the process's
+    availableTo: { gte: dbNow() },
+  },
+});
+```
+
+For typo tolerance ("piza") declare `@@trgm([name])` too and fall back to
+`findMany({ where: { name: { similarTo: query } } })` when a prefix search returns nothing
+(postgres/pglite only).
+
 ### Hydrating hits with relations
 
 `search()` returns bare rows — it has no `include`/`select` of relations. The intended pattern is the `select: "id"` projection plus a re-read through your own typed query, which carries relations, `omit`, computed fields and everything else the core API does:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@vibeorm/extensions",
-  "version": "2.0.0-alpha.8",
+  "version": "2.0.0",
   "description": "Tier-1 VibeORM extensions — full-text search, pgvector similarity and pg_trgm fuzzy matching",
   "keywords": [
     "orm",
@@ -54,9 +54,9 @@
     "bun": ">=1.2.0"
   },
   "dependencies": {
-    "@vibeorm/runtime": "2.0.0-alpha.7",
-    "@vibeorm/schema": "2.0.0-alpha.5",
-    "@vibeorm/sql": "2.0.0-alpha.5"
+    "@vibeorm/runtime": "2.0.0",
+    "@vibeorm/schema": "2.0.0",
+    "@vibeorm/sql": "2.0.0"
   },
   "publishConfig": {
     "access": "public"