genai #

A unified Dart/Flutter package for working with multiple Generative AI providers (Google Gemini, OpenAI, Anthropic, Azure OpenAI, Ollama, etc.) using a single request model.

• ✅ Supports normal & streaming responses
• ✅ Unified AIRequestModel across providers
• ✅ Configurable parameters (temperature, top-p, max tokens, etc.)
• ✅ Simple request utilities (executeGenAIRequest, streamGenAIRequest)
• ✅ Extensible — add your own provider easily

🚀 Installation #

Add genai to your pubspec.yaml:

dependencies:
  genai: ^0.1.0

Then run:

flutter pub get

⚡ Quick Start #

1. Import the package #

import 'package:genai/genai.dart';

2. Create a request #

final request = AIRequestModel(
  modelApiProvider: ModelAPIProvider.gemini, // or openai, anthropic, etc.
  model: "gemini-2.0-flash",
  apiKey: "<YOUR_API_KEY>",
  url: kGeminiUrl,
  systemPrompt: "You are a helpful assistant.",
  userPrompt: "Explain quantum entanglement simply.",
  stream: false, // set true for streaming
);

3. Run a non-streaming request #

final answer = await executeGenAIRequest(request);
print("AI Answer: $answer");

4. Run a streaming request (SSE) #

final stream = await streamGenAIRequest(request.copyWith(stream: true));
stream.listen((chunk) {
  print("AI Stream Chunk: $chunk");
}, onError: (err) {
  print("Stream Error: $err");
});

5. Auto-handle both (recommended) #

await callGenerativeModel(
  request,
  onAnswer: (ans) => print("AI Output: $ans"),
  onError: (err) => print("Error: $err"),
);

⚙️ Configuration #

Each request accepts modelConfigs to fine-tune output.

Available configs (defaults provided):

• temperature → controls randomness
• top_p / topP → nucleus sampling probability
• max_tokens / maxOutputTokens → maximum length of output
• stream → enables streaming

Example:

final request = request.copyWith(
  modelConfigs: [
    kDefaultModelConfigTemperature.copyWith(
      value: ConfigSliderValue(value: (0, 0.8, 1)),
    ),
    kDefaultGeminiModelConfigMaxTokens.copyWith(
      value: ConfigNumericValue(value: 2048),
    ),
  ],
);

📡 Supported Providers #

🛠️ Advanced Streaming (Word-by-Word) #

final stream = await streamGenAIRequest(request.copyWith(stream: true));

processGenAIStreamOutput(
  stream,
  (word) => print("Word: $word"), // called for each word
  (err) => print("Error: $err"),
);

🔒 Authentication #

• OpenAI / Anthropic / Azure OpenAI → API key passed as HTTP header.
• Gemini → API key passed as query param ?key=YOUR_API_KEY.
• Ollama → local server, no key required.

Just set apiKey in your AIRequestModel.

📦 Extending with New Providers #

Want to add a new AI provider?

1. Extend ModelProvider

2. Implement:

   • defaultAIRequestModel
   • createRequest()
   • outputFormatter()
   • streamOutputFormatter()

3. Register in kModelProvidersMap

That’s it — it plugs into the same unified request flow.

✅ Example: Gemini #

final request = GeminiModel.instance.defaultAIRequestModel.copyWith(
  model: "gemini-pro",
  apiKey: "<YOUR_KEY>",
  userPrompt: "Write me a haiku about Flutter.",
);

final answer = await executeGenAIRequest(request);
print(answer);

🤝 Contributing #

We welcome contributions to the genai package! If you'd like to contribute, please fork the repository and submit a pull request. For major changes or new features, it's a good idea to open an issue first to discuss your ideas.

Maintainer(s) #

• Ankit Mahato (GitHub, LinkedIn, X)
• Ashita Prasad (GitHub, LinkedIn, X)
• Manas Hejmadi (contributor) (GitHub)

License #

This project is licensed under the Apache License 2.0.