简介
chatgpt-java是一个OpenAI的Java版SDK,支持开箱即用。目前以支持官网全部Api。支持最新版本GPT-3.5-Turbo模型以及whisper-1模型。增加chat聊天对话以及语音文件转文字,语音翻译。
开源地址:https://github.com/Grt1228/chatgpt-java
快速开始
导入pom依赖
com.unfbx chatgpt-java 1.0.3
package com.unfbx.eventTest.test;import com.unfbx.chatgpt.OpenAiClient;import com.unfbx.chatgpt.entity.completions.CompletionResponse;import java.util.Arrays;public class TestB { public static void main(String[] args) { //代理可以为null Proxy proxy = new Proxy(Proxy.Type.HTTP, new InetSocketAddress("192.168.1.111", 7890)); OpenAiClient openAiClient = OpenAiClient.builder() .apiKey("sk-**************") .proxy(proxy) .build(); //简单模型 //CompletionResponse completions = //openAiClient.completions("我想申请转专业,从计算机专业转到会计学专业,帮我完成一份两百字左右的申请书"); //最新GPT-3.5-Turbo模型 Message message = Message.builder().role(Message.Role.USER).content("你好啊我的伙伴!").build(); ChatCompletion chatCompletion = ChatCompletion.builder().messages(Arrays.asList(message)).build(); ChatCompletionResponse chatCompletionResponse = openAiClient.chatCompletion(chatCompletion); chatCompletionResponse.getChoices().forEach(e -> { System.out.println(e.getMessage()); }); }}
支持流式输出
官方对于解决请求缓慢的情况推荐使用流式输出模式。
主要是基于SSE 实现的(可以百度下这个技术)。也是最近在了解到SSE。OpenAI官网在接受Completions接口的时候,有提到过这个技术。 Completion对象本身有一个stream属性,当stream为true时候Api的Response返回就会变成Http长链接。 具体可以看下文档:https://platform.openai.com/docs/api-reference/completions/create
package com.unfbx.chatgpt;********************/** * @author https:www.unfbx.com * 2023-02-28 */public class OpenAiStreamClientTest { private OpenAiStreamClient client; @Before public void before() { //创建流式输出客户端 Proxy proxy = new Proxy(Proxy.Type.HTTP, new InetSocketAddress("192.168.1.111", 7890)); client = OpenAiStreamClient.builder() .connectTimeout(50) .readTimeout(50) .writeTimeout(50) .apiKey("sk-******************************") .proxy(proxy) .build(); } //GPT-3.5-Turbo模型 @Test public void chatCompletions() { ConsoleEventSourceListener eventSourceListener = new ConsoleEventSourceListener(); Message message = Message.builder().role(Message.Role.USER).content("你好啊我的伙伴!").build(); ChatCompletion chatCompletion = ChatCompletion.builder().messages(Arrays.asList(message)).build(); client.streamChatCompletion(chatCompletion, eventSourceListener); CountDownLatch countDownLatch = new CountDownLatch(1); try { countDownLatch.await(); } catch (InterruptedException e) { e.printStackTrace(); } } //常规对话模型 @Test public void completions() { ConsoleEventSourceListener eventSourceListener = new ConsoleEventSourceListener(); Completion q = Completion.builder() .prompt("我想申请转专业,从计算机专业转到会计学专业,帮我完成一份两百字左右的申请书") .stream(true) .build(); client.streamCompletions(q, eventSourceListener); CountDownLatch countDownLatch = new CountDownLatch(1); try { countDownLatch.await(); } catch (InterruptedException e) { e.printStackTrace(); } }}
输出的是sse流式数据:
22:51:23.620 [OkHttp - OpenAI建立sse连接...22:51:23.623 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"role":"assistant"},"index":0,"finish_reason":null}]}22:51:23.625 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"你"},"index":0,"finish_reason":null}]}22:51:23.636 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"好"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"!"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"有"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"什"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"么"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"我"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"可以"},"index":0,"finish_reason":null}]}22:51:23.911 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"帮"},"index":0,"finish_reason":null}]}22:51:23.912 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"助"},"index":0,"finish_reason":null}]}22:51:23.934 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"你"},"index":0,"finish_reason":null}]}22:51:24.203 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"的"},"index":0,"finish_reason":null}]}22:51:24.203 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"吗"},"index":0,"finish_reason":null}]}22:51:24.203 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{"content":"?"},"index":0,"finish_reason":null}]}22:51:24.276 [OkHttp - OpenAI返回数据:{****省略无效数据******"model":"gpt-3.5-turbo-0301","choices":[{"delta":{},"index":0,"finish_reason":"stop"}]}22:51:24.276 [OkHttp - OpenAI返回数据:[DONE]22:51:24.277 [OkHttp - OpenAI返回数据结束了22:51:24.277 [OkHttp - OpenAI关闭sse连接...
流式输出如何集成Spring Boot实现 api接口?
可以参考项目:https://github.com/Grt1228/chatgpt-steam-output
实现自定义的EventSourceListener,例如:OpenAIEventSourceListener并持有一个SseEmitter,通过SseEmitter进行数据的通信
postman测试
发送请求:Get http://localhost:8080/test/sse?uid=123
看下response (需要新版本postman)
重点关注下header:Content-Type:text/event-stream
如果想结合前端显示自行百度sse前端相关实现说明
支持最新版的语音转文字,语音翻译api请参考测试代码:OpenAiClientTest.java