EmbeddingService class

Dart-based embedding service backed by a dedicated background isolate.

All ONNX inference runs on a long-lived worker isolate so that the main isolate (UI thread) is never blocked. The public API is unchanged:

await EmbeddingService.init(modelPath: '/path/to/model.onnx');
final embedding = await EmbeddingService.embed('Hello world');
await EmbeddingService.disposeAsync();

Constructors

EmbeddingService()

Properties

hashCode → int
The hash code for this object.
no setterinherited
runtimeType → Type
A representation of the runtime type of the object.
no setterinherited

Methods

noSuchMethod(Invocation invocation) → dynamic
Invoked when a nonexistent method or property is accessed.
inherited
toString() → String
A string representation of this object.
inherited

Operators

operator ==(Object other) → bool
The equality operator.
inherited

Static Properties

debugMode ↔ bool
Debug mode flag — forwarded to the worker on each embed call.
getter/setter pair

Static Methods

dispose() → void
Backward compatible, non-blocking dispose.
disposeAsync() → Future<void>
Release resources deterministically.
embed(String text) → Future<Float32List>
Convert text to an embedding vector.
embedBatch(List<String> texts, {int concurrency = 1, void onProgress(int completed, int total)?}) → Future<List<Float32List>>
Batch embed multiple texts (sequential processing).
init({Uint8List? modelBytes, String? modelPath, OrtSessionOptions? options, int? intraOpNumThreads}) → Future<void>
Initialize ONNX model on a background worker isolate.
startWorkerForTest(void entryPoint(SendPort mainSendPort)) → Future<void>
(FOR TESTING ONLY) Runs entryPoint as the worker instead of the ONNX worker, without an init command. It receives the same messages (embed, dispose) and must answer them the same way.