installWithProgress method

  1. @override
Stream<int> installWithProgress(
  1. ModelSource source, {
  2. CancelToken? cancelToken,
})
override

Installs the model with progress tracking

Returns a stream of progress percentages (0-100)

Parameters:

  • source: The model source to install from
  • cancelToken: Optional token for cancelling the installation

Note: Some sources may not support true progress:

  • AssetSource: simulates progress (copy is instant)
  • BundledSource: returns 100 immediately (no download)
  • FileSource: returns 100 immediately (just registration)

Example:

final cancelToken = CancelToken();

try {
  await for (final progress in handler.installWithProgress(
    source,
    cancelToken: cancelToken,
  )) {
    print('Progress: $progress%');
  }
} catch (e) {
  if (CancelToken.isCancel(e)) {
    print('Installation cancelled');
  }
}

Throws:

Implementation

@override
Stream<int> installWithProgress(
  ModelSource source, {
  CancelToken? cancelToken,
}) async* {
  // Same as above - file registration is instant
  if (source is! FileSource) {
    throw ArgumentError('FileSourceHandler only supports FileSource');
  }

  // Verify external file exists
  final exists = await fileSystem.fileExists(source.path);
  if (!exists) {
    throw Exception('External file does not exist: ${source.path}');
  }

  // Generate unique filename for tracking
  final filename = path.basename(source.path);

  // Register external file in file system
  await fileSystem.registerExternalFile(filename, source.path);

  // Protect file from cleanup operations
  await protectedFiles.protect(filename);

  // Register external path mapping
  await protectedFiles.registerExternalPath(filename, source.path);

  // External files are immediately available, report 100% after registration
  yield 100;

  // Get file size for metadata
  final sizeBytes = await fileSystem.getFileSize(source.path);

  // Save metadata to repository
  final modelInfo = ModelInfo(
    id: filename,
    source: source,
    installedAt: DateTime.now(),
    sizeBytes: sizeBytes,
    type: ModelType.inference,
    hasLoraWeights: false,
  );

  await repository.saveModel(modelInfo);
}