gemini_box

A Google Gemini provider based on ai_box.

Usage

import 'package:ai_box/ai_box.dart';
import 'package:gemini_box/gemini_box.dart';

Future<void> main() async {
  final ai = Gemini(apiKey: 'YOUR_API_KEY');

  // One-shot text generation.
  final answer = await ai.generateText(
    model: 'gemini-3.5-flash',
    message: 'Say hello in one short sentence.',
  );
  print(answer);

  // Multi-turn chat.
  final res = await ai.chat(
    model: 'gemini-3.5-flash',
    messages: [
      LLMContent.system('You are concise.'),
      LLMContent.user('What is the capital of Japan?'),
    ],
    maxTokens: 32,
  );
  print(res.text);
  print(res.usage); // token usage
}

Streaming

True SSE streaming — text arrives incrementally and the final chunk carries the finish reason and token usage:

await for (final text in ai.generateTextStream(
  model: 'gemini-3.5-flash',
  message: 'Write a haiku about Dart.',
)) {
  stdout.write(text);
}

Working with images

Pass images as LLMContentParts on any message — ai_box converts them to Gemini inlineData / fileData automatically:

final res = await ai.chat(
  model: 'gemini-3.5-flash',
  messages: [
    LLMContent.user('Please describe this image.', attachments: [
      LLMImagePart.bytes(jpegBytes, mimeType: 'image/jpeg'),
    ]),
  ],
);
print(res.text);

Low-level typed API

For full control over the request body, use generateContent with the typed request/response classes:

final response = await ai.generateContent(
  model: 'gemini-3.5-flash',
  requestBody: RequestBody(
    contents: [
      Content(parts: [Part(text: 'Hello, Gemini!')]),
    ],
    generationConfig: GenerationConfig(
      temperature: 0.7,
      maxOutputTokens: 1000,
    ),
  ),
);
print(response.candidates?.first.content?.parts?.first);

Files API

Upload and manage files via ai.files (the Gemini Files API):

final file = await ai.files.uploadFile(file: File('path/to/image.jpg'));
final files = await ai.files.getFiles();

Models and key validation

final models = await ai.getModels(); // live model list from the API
final ok = await ai.validateKey();

Pass any model ID the API serves (see getModels()), e.g. gemini-3.5-flash or other current Gemini 3.x models.

Multimodal input (images, files), tool calling, structured output and the sealed LLMException error hierarchy are all provided by ai_box — see its README for details.