介绍

OpenaiService 是 Hutool AI 模块中专门对接 OpenAI 服务的接口，提供了对 OpenAI 各种功能的完整支持，包括聊天、图像生成与处理、语音转换、文本向量化等功能。

特有功能

• 视觉相关
  • 图像理解 (chatVision)
  • 文生图 (imagesGenerations)
  • 图片编辑 (imagesEdits)
  • 图片变形 (imagesVariations)
• 语音相关
  • 文本转语音 (textToSpeech)
  • 语音转文本 (speechToText)
• 其他
  • 文本向量化 (embeddingText)
  • 内容审核 (moderations)
  • 推理聊天 (chatReasoning)

使用

视觉相关

• 视觉理解

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.GPT_4O_MINI.getModel()).build(), OpenaiService.class); String chatVision = openaiService.chatVision("图片上有些什么？", Arrays.asList("https://img2.baidu.com/it/u=862000265,4064861820&fm=253&fmt=auto&app=138&f=JPEG?w=800&h=1544", "https://img2.baidu.com/it/u=1682510685,1244554634&fm=253&fmt=auto&app=138&f=JPEG?w=803&h=800")); 

• 文生图

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.DALL_E_3.getModel()).build(), OpenaiService.class); String imagesGenerations = openaiService.imagesGenerations("一位年轻的宇航员站在未来感十足的太空站内，透过巨大的弧形落地窗凝望浩瀚宇宙。 窗外，璀璨的星河与五彩斑斓的星云交织，远处隐约可见未知星球的轮廓，仿佛在召唤着探索的脚步。宇航服上的呼吸灯与透明显示屏上的星图交相辉映，象征着人类科技与宇宙奥秘的碰撞。画面深邃而神秘，充满对未知的渴望与无限可能的想象。"); 

• 图片编辑

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.DALL_E_2.getModel()).build(), OpenaiService.class); File file = FileUtil.file("your imgUrl"); String imagesEdits = openaiService.imagesEdits("茂密的森林中，有一只九色鹿若隐若现",file); 

• 图片变形

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.DALL_E_2.getModel()).build(), OpenaiService.class); File file = FileUtil.file("your imgUrl"); String imagesVariations = openaiService.imagesVariations(file); 

语音相关

• TTS-文字转语音

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.TTS_1_HD.getModel()).build(), OpenaiService.class); InputStream inputStream = openaiService.textToSpeech("万里山河一夜白，\n" +"千峰尽染玉龙哀。\n" + "长风卷起琼花碎，\n" +"直上九霄揽月来。", OpenaiCommon.OpenaiSpeech.NOVA); String filePath = "your filePath"; Path path = Paths.get(filePath); try (final FileOutputStream outputStream = new FileOutputStream(filePath)) { Files.createDirectories(path.getParent()); byte[] buffer = new byte[1024]; int bytesRead; while ((bytesRead = inputStream.read(buffer)) != -1) {	outputStream.write(buffer, 0, bytesRead); } } catch (final IOException e) { throw new RuntimeException(e); } 

• STT-语音转文字

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.WHISPER_1.getModel()).build(), OpenaiService.class); File file = FileUtil.file("your filePath"); String speechToText = openaiService.speechToText(file); 

其他

• 文本向量化

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.TEXT_EMBEDDING_3_SMALL.getModel()).build(), OpenaiService.class); String embeddingText = openaiService.embeddingText("万里山河一夜白，千峰尽染玉龙哀。长风卷起琼花碎，直上九霄揽月来。"); 

• 内容审核

OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.OMNI_MODERATION_LATEST.getModel()).build(), OpenaiService.class); String moderations = openaiService.moderations("检测文本", "https://img2.baidu.com/it/u=862000265,4064861820&fm=253&fmt=auto&app=138&f=JPEG?w=800&h=1544"); 

• 推理聊天

final OpenaiService openaiService = AIServiceFactory.getAIService(new AIConfigBuilder(ModelName.OPENAI.getValue()).setApiKey(key).setModel(Models.Openai.O3_MINI.getModel()).build(), OpenaiService.class); List<Message> messages = new ArrayList<>(); messages.add(new Message("system","你是现代抽象家")); messages.add(new Message("user","给我一个KFC疯狂星期四的文案")); String chatReasoning = openaiService.chatReasoning(messages, OpenaiCommon.OpenaiReasoning.HIGH.getEffort());