fire_api_flutter #

fire_api_flutter is the Flutter implementation for fire_api.

It uses cloud_firestore and firebase_storage for the normal shared API surface, and uses Firestore REST for vector search while still authenticating as the current Firebase user.

Setup #

Initialize after Firebase:

import 'package:fire_api_flutter/fire_api_flutter.dart';

void main() {
  FirebaseFirestoreDatabase.create();
  FirebaseFireStorage.create();
}

Notes #

Vector values #

VectorValue is translated automatically in both directions, including nested maps and lists.

await FirestoreDatabase.instance.collection("items").doc("one").set({
  "embedding": const VectorValue([0.1, 0.2, 0.3]),
});

Vector search #

Vector queries are supported through the shared API:

final docs = await FirestoreDatabase.instance
    .collection("items")
    .findNearest(
      vectorField: "embedding_field",
      queryVector: const VectorValue([3, 1, 2]),
      limit: 10,
      distanceMeasure: VectorDistanceMeasure.euclidean,
      distanceResultField: "vector_distance",
    )
    .get();

Implementation details:

• vector queries use Firestore REST runQuery
• authentication is done with the current Firebase Auth ID token
• this keeps the request in the normal client-auth / Firestore-rules model
• vector queries are get() only and do not support realtime listeners

If you need custom auth behavior, FirebaseFirestoreDatabase.create(...) also accepts a custom idTokenProvider.

Shared collection deletes #

CollectionReference.deleteAll(...) is implemented in the shared fire_api package, so it works here automatically:

await FirestoreDatabase.instance.collection("sessions").deleteAll();

await FirestoreDatabase.instance.collection("sessions").deleteAll(
  only: {"a", "b", "c"},
  batchSize: 50,
);