Flutter ChatGPT完成请求插件 chatgpt_completions 的使用

特性

• 文本补全（不带流响应）
• 文本补全（带流响应）

开始使用

安装包

在您的项目中添加 chatgpt_completions 包：

flutter pub add chatgpt_completions

使用方法

初始化实例

首先，您需要初始化插件，并提供从 OpenAI 控制台获取的 API 密钥：

/// Generate api key from openai console: https://platform.openai.com/account/api-keys
ChatGPTCompletions.instance.initialize(apiKey: "api_key_here");

文本补全（不带流响应）

以下是如何执行不带流响应的文本补全请求：

String? responseWithoutStream = await ChatGPTCompletions.instance.textCompletions(
  TextCompletionsParams(
    prompt: "What's Flutter?",
    model: GPTModel.davinci,
    temperature: 0.2,
    topP: 1,
    n: 1,
    stream: false,
    maxTokens: 2048,
  ),
);

print("OpenAI: $responseWithoutStream");

文本补全（带流响应）

如果需要实时接收响应，可以设置 stream 参数为 true：

String responseWithStream = "";
StreamSubscription? responseSubscription;

await ChatGPTCompletions.instance.textCompletions(
  TextCompletionsParams(
    prompt: "What's Flutter?",
    model: GPTModel.davinci,
    temperature: 0.2,
    topP: 1,
    n: 1,
    stream: true, // 设置为 true 以启用流响应
    maxTokens: 2048,
  ),
  onStreamValue: (characters) {
    responseWithStream += characters;
    print(responseWithStream);
  },
  onStreamCreated: (subscription) {
    responseSubscription = subscription;
  },
);

停止生成响应：

responseSubscription?.cancel();

使用 GPT-3.5-Turbo 或 GPT-4 模型

对于更高级的模型，如 gpt-3.5-turbo 或 gpt-4，您可以使用 messagesTurbo 参数：

await ChatGPTCompletions.instance.textCompletions(
  TextCompletionsParams(
    messagesTurbo: [
      MessageTurbo(
        role: TurboRole.user,
        content: "What's Flutter?",
      ),
    ],
    model: GPTModel.gpt3p5turbo, // 切换到 gpt-3.5-turbo 模型
  ),
  onStreamValue: (characters) {
    responseWithStream += characters;
    print(responseWithStream);
  },
  onStreamCreated: (subscription) {
    responseSubscription = subscription;
  },
  // 延迟 100ms 接收下一个值
  debounce: const Duration(milliseconds: 100),
);

示例 Demo

以下是一个完整的示例代码，展示了如何使用 chatgpt_completions 插件进行文本补全：

// ignore_for_file: avoid_print

import 'dart:async';
import 'package:chatgpt_completions/chatgpt_completions.dart';

void main() async {
  /// Generate api key from openai console: https://platform.openai.com/account/api-keys
  ChatGPTCompletions.instance.initialize(apiKey: '<openai_api_key>');

  print("Generating answer without stream...");

  // Text completions without stream response (stream: false)
  String? responseWithoutStream = await ChatGPTCompletions.instance.textCompletions(
    TextCompletionsParams(
      prompt: "What's Flutter?",
      model: GPTModel.davinci,
      stream: false,
    ),
  );

  print("OpenAI: $responseWithoutStream");

  print("\n\n-> Generating answer with stream...");
  await Future.delayed(const Duration(seconds: 2));

  // Text completions with stream response (stream: true)
  String responseWithStream = "";

  StreamSubscription? responseSubscription;

  await ChatGPTCompletions.instance.textCompletions(
    TextCompletionsParams(
      prompt: "What's Flutter?",
      model: GPTModel.davinci,
    ),
    onStreamValue: (characters) {
      responseWithStream += characters;
      print(responseWithStream);
    },
    onStreamCreated: (subscription) {
      responseSubscription = subscription;
    },
    // Debounce 200ms for receive next value
    debounce: const Duration(milliseconds: 200),
  );

  // Stop generating
  responseSubscription?.cancel();

  // Using Chat Completion API: GPT-3.5-Turbo, GPT-4,...
  await ChatGPTCompletions.instance.textCompletions(
    TextCompletionsParams(
      messagesTurbo: [
        MessageTurbo(
          role: TurboRole.user,
          content: "Where is the tallest building in the world?",
        ),
      ],
      model: GPTModel.gpt3p5turbo,
    ),
    onStreamValue: (characters) {
      responseWithStream += characters;
      print(characters);
    },
    onStreamCreated: (subscription) {
      responseSubscription = subscription;
    },
    // Debounce 100ms for receive next value
    debounce: const Duration(milliseconds: 100),
  );
}