generateStream method

@override Stream<String> generateStream(String prompt, { Iterable<Attachment> attachments = const [], })

Generates a stream of text based on the given prompt and attachments. This method does not interact with a chat or build on any chat history.

prompt is the input text to generate a response for. attachments is an optional iterable of Attachment objects to include with the prompt. These can be images, files, or links that provide additional context for the LLM.

Returns a Stream of String containing the generated text chunks. This allows for streaming responses as they are generated by the LLM.

Implementation

@override
Stream<String> generateStream(
  String prompt, {
  Iterable<Attachment> attachments = const [],
}) async* {
  if (prompt == 'FAILFAST') throw const LlmFailureException('Failing fast!');

  await Future.delayed(const Duration(milliseconds: 1000));
  yield '# Echo\n';

  switch (prompt) {
    case 'CANCEL':
      throw const LlmCancelException();
    case 'FAIL':
      throw const LlmFailureException('User requested failure');
  }

  await Future.delayed(const Duration(milliseconds: 1000));
  yield prompt;

  yield '\n\n# Attachments\n${attachments.map((a) => a.toString())}';
}