Added encoder

CHANGELOG.md

@@ -1,4 +1,4 @@
-## 0.1.1 (unreleased)
+## 0.2.0 (unreleased)
 
 - Added `Vector`, `HalfVector`, `Bit`, and `SparseVector` classes
 

README.md

@@ -31,6 +31,14 @@ Import the library
 import 'package:pgvector/pgvector.dart';
 ```
 
+Add the encoder
+
+```dart
+var connection = await Connection.open(endpoint,
+    settings: ConnectionSettings(
+        typeRegistry: TypeRegistry(encoders: [pgvectorEncoder])));
+```
+
 Enable the extension
 
 ```dart
@@ -49,9 +57,9 @@ Insert vectors
 await connection.execute(
     Sql.named('INSERT INTO items (embedding) VALUES (@a), (@b), (@c)'),
     parameters: {
-      'a': pgvector.encode([1, 1, 1]),
-      'b': pgvector.encode([2, 2, 2]),
-      'c': pgvector.encode([1, 1, 2])
+      'a': Vector([1, 1, 1]),
+      'b': Vector([2, 2, 2]),
+      'c': Vector([1, 1, 2])
     });
 ```
 
@@ -61,11 +69,11 @@ Get the nearest neighbors
 List<List<dynamic>> results = await connection.execute(
     Sql.named('SELECT id, embedding FROM items ORDER BY embedding <-> @embedding LIMIT 5'),
     parameters: {
-      'embedding': pgvector.encode([1, 1, 1])
+      'embedding': Vector([1, 1, 1])
     });
 for (final row in results) {
   print(row[0]);
-  print(pgvector.decode(row[1].bytes));
+  print(Vector.fromBinary(row[1].bytes));
 }
 ```
 

examples/openai/example.dart

@@ -31,7 +31,9 @@ void main() async {
           port: 5432,
           database: 'pgvector_example',
           username: Platform.environment['USER']),
-      settings: ConnectionSettings(sslMode: SslMode.disable));
+      settings: ConnectionSettings(
+          sslMode: SslMode.disable,
+          typeRegistry: TypeRegistry(encoders: [pgvectorEncoder])));
 
   await connection.execute('CREATE EXTENSION IF NOT EXISTS vector');
 
@@ -52,7 +54,7 @@ void main() async {
             'INSERT INTO documents (content, embedding) VALUES (@content, @embedding)'),
         parameters: {
           'content': input[i],
-          'embedding': pgvector.encode(List<double>.from(embeddings[i]))
+          'embedding': Vector(List<double>.from(embeddings[i]))
         });
   }
 

lib/pgvector.dart

@@ -1,5 +1,6 @@
 export 'src/bit.dart' show Bit;
 export 'src/halfvec.dart' show HalfVector;
 export 'src/pgvector.dart' show pgvector;
+export 'src/postgres.dart' show pgvectorEncoder;
 export 'src/sparsevec.dart' show SparseVector;
 export 'src/vector.dart' show Vector;

lib/src/postgres.dart

@@ -0,0 +1,32 @@
+import 'dart:convert';
+import 'package:postgres/postgres.dart';
+import 'bit.dart';
+import 'halfvec.dart';
+import 'sparsevec.dart';
+import 'vector.dart';
+
+EncodedValue? pgvectorEncoder(TypedValue input, CodecContext context) {
+  final value = input.value;
+
+  if (value is Vector) {
+    final v = value as Vector;
+    return EncodedValue.binary(v.toBinary());
+  }
+
+  if (value is HalfVector) {
+    final v = value as HalfVector;
+    return EncodedValue.text(utf8.encode(v.toString()));
+  }
+
+  if (value is Bit) {
+    final v = value as Bit;
+    return EncodedValue.binary(v.toBinary());
+  }
+
+  if (value is SparseVector) {
+    final v = value as SparseVector;
+    return EncodedValue.binary(v.toBinary());
+  }
+
+  return null;
+}

pubspec.yaml

@@ -7,8 +7,8 @@ environment:
   sdk: ^3.0.0
 
 dependencies:
+  postgres: ^3.4.0
 
 dev_dependencies:
   lints: ^2.0.0
-  postgres: ^3.0.0
   test: ^1.21.0

test/postgres_test.dart

@@ -11,7 +11,9 @@ void main() {
             port: 5432,
             database: 'pgvector_dart_test',
             username: Platform.environment['USER']),
-        settings: ConnectionSettings(sslMode: SslMode.disable));
+        settings: ConnectionSettings(
+            sslMode: SslMode.disable,
+            typeRegistry: TypeRegistry(encoders: [pgvectorEncoder])));
 
     await connection.execute('CREATE EXTENSION IF NOT EXISTS vector');
     await connection.execute('DROP TABLE IF EXISTS items');
@@ -23,25 +25,25 @@ void main() {
         Sql.named(
             'INSERT INTO items (embedding, half_embedding, binary_embedding, sparse_embedding) VALUES (@embedding1, @half_embedding1, @binary_embedding1, @sparse_embedding1), (@embedding2, @half_embedding2, @binary_embedding2, @sparse_embedding2), (@embedding3, @half_embedding3, @binary_embedding3, @sparse_embedding3)'),
         parameters: {
-          'embedding1': Vector([1, 1, 1]).toString(),
-          'embedding2': Vector([2, 2, 2]).toString(),
-          'embedding3': Vector([1, 1, 2]).toString(),
-          'half_embedding1': HalfVector([1, 1, 1]).toString(),
-          'half_embedding2': HalfVector([2, 2, 2]).toString(),
-          'half_embedding3': HalfVector([1, 1, 2]).toString(),
-          'binary_embedding1': '000',
-          'binary_embedding2': '101',
-          'binary_embedding3': '111',
-          'sparse_embedding1': SparseVector([1, 1, 1]).toString(),
-          'sparse_embedding2': SparseVector([2, 2, 2]).toString(),
-          'sparse_embedding3': SparseVector([1, 1, 2]).toString()
+          'embedding1': Vector([1, 1, 1]),
+          'embedding2': Vector([2, 2, 2]),
+          'embedding3': Vector([1, 1, 2]),
+          'half_embedding1': HalfVector([1, 1, 1]),
+          'half_embedding2': HalfVector([2, 2, 2]),
+          'half_embedding3': HalfVector([1, 1, 2]),
+          'binary_embedding1': Bit([false, false, false]),
+          'binary_embedding2': Bit([true, false, true]),
+          'binary_embedding3': Bit([true, true, true]),
+          'sparse_embedding1': SparseVector([1, 1, 1]),
+          'sparse_embedding2': SparseVector([2, 2, 2]),
+          'sparse_embedding3': SparseVector([1, 1, 2])
         });
 
     List<List<dynamic>> results = await connection.execute(
         Sql.named(
             'SELECT id, embedding, binary_embedding, sparse_embedding FROM items ORDER BY embedding <-> @embedding LIMIT 5'),
         parameters: {
-          'embedding': Vector([1, 1, 1]).toString()
+          'embedding': Vector([1, 1, 1])
         });
     expect(results.map((r) => r[0]), equals([1, 3, 2]));
     expect(Vector.fromBinary(results[1][1].bytes), equals(Vector([1, 1, 2])));