From f6f6bb13be26bbbcb9a2c8fc39db22f516e8d48d Mon Sep 17 00:00:00 2001 From: jack ning Date: Thu, 21 Aug 2025 13:08:37 +0800 Subject: [PATCH] update --- .../providers/baidu/SpringAIBaiduService.java | 41 +- .../dashscope/SpringAIDashscopeService.java | 21 +- .../deepseek/SpringAIDeepseekService.java | 21 +- .../providers/gitee/SpringAIGiteeService.java | 9 +- .../minimax/SpringAIMinimaxService.java | 21 +- .../ollama/SpringAIOllamaChatConfig.java | 114 +-- .../ollama/SpringAIOllamaChatService.java | 324 +++++++ .../ollama/SpringAIOllamaController.java | 5 +- .../ollama/SpringAIOllamaService.java | 12 +- .../openai/SpringAIOpenaiService.java | 9 +- .../openrouter/SpringAIOpenrouterService.java | 9 +- .../SpringAISiliconFlowService.java | 9 +- .../tencent/SpringAITencentService.java | 9 +- .../volcengine/SpringAIVolcengineService.java | 9 +- .../zhipuai/SpringAIZhipuaiService.java | 41 +- .../springai/service/BaseSpringAIService.java | 37 +- .../ai/springai/service/ChatTokenUsage.java | 29 + .../ai/zhipuai/ZhipuaiChatConfig.java | 4 +- .../ai/zhipuai/ZhipuaiChatService.java | 624 +++++++++++++ .../ai/zhipuai/ZhipuaiController.java | 91 +- .../bytedesk/ai/zhipuai/ZhipuaiService.java | 860 ++++-------------- 21 files changed, 1381 insertions(+), 918 deletions(-) create mode 100644 modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatService.java create mode 100644 modules/ai/src/main/java/com/bytedesk/ai/springai/service/ChatTokenUsage.java create mode 100644 modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatService.java diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/baidu/SpringAIBaiduService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/baidu/SpringAIBaiduService.java index 750038c063..61f20b9134 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/baidu/SpringAIBaiduService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/baidu/SpringAIBaiduService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 15:10:31 + * @LastEditTime: 2025-08-21 12:46:15 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -31,6 +31,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -82,7 +83,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 记录开始时间和初始化token使用统计 long startTime = System.currentTimeMillis(); final boolean[] success = { false }; - final TokenUsage[] tokenUsage = { new TokenUsage(0, 0, 0) }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; // 用于累积所有响应文本 final StringBuilder[] fullResponseText = { new StringBuilder() }; @@ -118,7 +119,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 如果token提取失败，使用累积的完整响应文本来估算token if (tokenUsage[0].getTotalTokens() == 0 && fullResponseText[0].length() > 0) { log.info("Baidu API using accumulated response text for token estimation: {}", fullResponseText[0].toString()); - TokenUsage estimatedUsage = estimateBaiduTokenUsageFromText(fullResponseText[0].toString()); + ChatTokenUsage estimatedUsage = estimateBaiduTokenUsageFromText(fullResponseText[0].toString()); tokenUsage[0] = estimatedUsage; log.info("Baidu API final estimated token usage: {}", estimatedUsage); } @@ -138,7 +139,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { log.info("SpringAIBaiduService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (baiduChatModel == null) { @@ -212,7 +213,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 记录开始时间和初始化token使用统计 long startTime = System.currentTimeMillis(); final boolean[] success = { false }; - final TokenUsage[] tokenUsage = { new TokenUsage(0, 0, 0) }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; // 用于累积所有响应文本 final StringBuilder[] fullResponseText = { new StringBuilder() }; @@ -255,7 +256,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 如果token提取失败，使用累积的完整响应文本来估算token if (tokenUsage[0].getTotalTokens() == 0 && fullResponseText[0].length() > 0) { log.info("Baidu API using accumulated response text for token estimation: {}", fullResponseText[0].toString()); - TokenUsage estimatedUsage = estimateBaiduTokenUsageFromText(fullResponseText[0].toString()); + ChatTokenUsage estimatedUsage = estimateBaiduTokenUsageFromText(fullResponseText[0].toString()); tokenUsage[0] = estimatedUsage; log.info("Baidu API final estimated token usage: {}", estimatedUsage); } @@ -283,17 +284,17 @@ public class SpringAIBaiduService extends BaseSpringAIService { * @param response ChatResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractBaiduTokenUsage(org.springframework.ai.chat.model.ChatResponse response) { + private ChatTokenUsage extractBaiduTokenUsage(org.springframework.ai.chat.model.ChatResponse response) { try { if (response == null) { log.warn("Baidu API response is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var metadata = response.getMetadata(); if (metadata == null) { log.warn("Baidu API response metadata is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Baidu API manual token extraction - metadata: {}", metadata); @@ -390,7 +391,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { log.info("Baidu API manual token extraction result - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } @@ -404,7 +405,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 方法3: 如果手动提取失败，尝试使用原始的extractTokenUsage方法作为后备 log.info("Baidu API manual extraction failed, trying original extractTokenUsage method"); - TokenUsage fallbackUsage = extractTokenUsage(response); + ChatTokenUsage fallbackUsage = extractTokenUsage(response); if (fallbackUsage.getTotalTokens() > 0) { log.info("Baidu API fallback extraction successful: {}", fallbackUsage); return fallbackUsage; @@ -412,13 +413,13 @@ public class SpringAIBaiduService extends BaseSpringAIService { // 方法4: 如果所有方法都失败，尝试估算token使用量 log.info("Baidu API all extraction methods failed, attempting to estimate token usage"); - TokenUsage estimatedUsage = estimateBaiduTokenUsageFromResponse(response); + ChatTokenUsage estimatedUsage = estimateBaiduTokenUsageFromResponse(response); log.info("Baidu API estimated token usage: {}", estimatedUsage); return estimatedUsage; } catch (Exception e) { log.error("Error in manual Baidu token extraction", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -428,10 +429,10 @@ public class SpringAIBaiduService extends BaseSpringAIService { * @param response ChatResponse对象 * @return 估算的TokenUsage对象 */ - private TokenUsage estimateBaiduTokenUsageFromResponse(org.springframework.ai.chat.model.ChatResponse response) { + private ChatTokenUsage estimateBaiduTokenUsageFromResponse(org.springframework.ai.chat.model.ChatResponse response) { try { if (response == null || response.getResults() == null || response.getResults().isEmpty()) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 获取输出文本 @@ -446,7 +447,7 @@ public class SpringAIBaiduService extends BaseSpringAIService { } catch (Exception e) { log.error("Error estimating Baidu token usage", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -456,10 +457,10 @@ public class SpringAIBaiduService extends BaseSpringAIService { * @param outputText 完整的输出文本 * @return 估算的TokenUsage对象 */ - private TokenUsage estimateBaiduTokenUsageFromText(String outputText) { + private ChatTokenUsage estimateBaiduTokenUsageFromText(String outputText) { try { if (outputText == null || outputText.isEmpty()) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 估算token使用量 @@ -474,11 +475,11 @@ public class SpringAIBaiduService extends BaseSpringAIService { "Baidu API estimated tokens - output: {} chars -> {} tokens, estimated prompt: {} tokens, total: {} tokens", outputText.length(), completionTokens, promptTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } catch (Exception e) { log.error("Error estimating Baidu token usage from text", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/dashscope/SpringAIDashscopeService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/dashscope/SpringAIDashscopeService.java index 1ae02476b6..1962328303 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/dashscope/SpringAIDashscopeService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/dashscope/SpringAIDashscopeService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-18 17:04:41 + * @LastEditTime: 2025-08-21 12:46:21 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -33,6 +33,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -87,7 +88,7 @@ public class SpringAIDashscopeService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; final StringBuilder[] fullResponseText = {new StringBuilder()}; // 使用同一个ChatModel实例，但传入不同的选项 @@ -132,7 +133,7 @@ public class SpringAIDashscopeService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("Dashscope API sync fullPromptContent: {}", fullPromptContent); @@ -202,7 +203,7 @@ public class SpringAIDashscopeService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; final StringBuilder[] fullResponseText = {new StringBuilder()}; bytedeskDashscopeChatModel.stream(requestPrompt).subscribe( @@ -259,17 +260,17 @@ public class SpringAIDashscopeService extends BaseSpringAIService { * @param response ChatResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractDashscopeTokenUsage(ChatResponse response) { + private ChatTokenUsage extractDashscopeTokenUsage(ChatResponse response) { try { if (response == null) { log.warn("Dashscope API response is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var metadata = response.getMetadata(); if (metadata == null) { log.warn("Dashscope API response metadata is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Dashscope API token extraction - metadata: {}", metadata); @@ -286,18 +287,18 @@ public class SpringAIDashscopeService extends BaseSpringAIService { promptTokens, completionTokens, totalTokens); if (totalTokens > 0) { - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } } catch (Exception e) { log.debug("Could not get usage via getUsage() method: {}", e.getMessage()); } - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } catch (Exception e) { log.error("Error in Dashscope token extraction", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/deepseek/SpringAIDeepseekService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/deepseek/SpringAIDeepseekService.java index 717c21b577..ad91a9fdc3 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/deepseek/SpringAIDeepseekService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/deepseek/SpringAIDeepseekService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-18 17:04:22 + * @LastEditTime: 2025-08-21 12:46:25 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -34,6 +34,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -87,7 +88,7 @@ public class SpringAIDeepseekService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 deepseekChatModel.stream(requestPrompt).subscribe( @@ -126,7 +127,7 @@ public class SpringAIDeepseekService extends BaseSpringAIService { log.info("SpringAIDeepseekService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (deepseekChatModel == null) { @@ -195,7 +196,7 @@ public class SpringAIDeepseekService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; deepseekChatModel.stream(requestPrompt).subscribe( response -> { @@ -245,17 +246,17 @@ public class SpringAIDeepseekService extends BaseSpringAIService { * @param response ChatResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractDeepSeekTokenUsage(ChatResponse response) { + private ChatTokenUsage extractDeepSeekTokenUsage(ChatResponse response) { try { if (response == null) { log.warn("Deepseek API response is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var metadata = response.getMetadata(); if (metadata == null) { log.warn("Deepseek API response metadata is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Deepseek API token extraction - metadata: {}", metadata); @@ -272,18 +273,18 @@ public class SpringAIDeepseekService extends BaseSpringAIService { promptTokens, completionTokens, totalTokens); if (totalTokens > 0) { - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } } catch (Exception e) { log.debug("Could not get usage via getUsage() method: {}", e.getMessage()); } - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } catch (Exception e) { log.error("Error in Deepseek token extraction", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/gitee/SpringAIGiteeService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/gitee/SpringAIGiteeService.java index 36891f9984..7eec56b496 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/gitee/SpringAIGiteeService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/gitee/SpringAIGiteeService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 13:23:44 + * @LastEditTime: 2025-08-21 12:46:30 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -32,6 +32,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -84,7 +85,7 @@ public class SpringAIGiteeService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 giteeChatModel.stream(requestPrompt).subscribe( @@ -123,7 +124,7 @@ public class SpringAIGiteeService extends BaseSpringAIService { log.info("SpringAIGiteeService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (giteeChatModel == null) { @@ -193,7 +194,7 @@ public class SpringAIGiteeService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; giteeChatModel.stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/minimax/SpringAIMinimaxService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/minimax/SpringAIMinimaxService.java index 5043a9beb7..9a6fecd3ea 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/minimax/SpringAIMinimaxService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/minimax/SpringAIMinimaxService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-22 08:34:28 + * @LastEditTime: 2025-08-21 12:46:36 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -34,6 +34,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -87,7 +88,7 @@ public class SpringAIMinimaxService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 minimaxChatModel.stream(requestPrompt).subscribe( @@ -126,7 +127,7 @@ public class SpringAIMinimaxService extends BaseSpringAIService { log.info("SpringAIMinimaxService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (minimaxChatModel == null) { @@ -195,7 +196,7 @@ public class SpringAIMinimaxService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; minimaxChatModel.stream(requestPrompt).subscribe( response -> { @@ -245,17 +246,17 @@ public class SpringAIMinimaxService extends BaseSpringAIService { * @param response ChatResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractDeepSeekTokenUsage(ChatResponse response) { + private ChatTokenUsage extractDeepSeekTokenUsage(ChatResponse response) { try { if (response == null) { log.warn("Minimax API response is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var metadata = response.getMetadata(); if (metadata == null) { log.warn("Minimax API response metadata is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Minimax API token extraction - metadata: {}", metadata); @@ -272,18 +273,18 @@ public class SpringAIMinimaxService extends BaseSpringAIService { promptTokens, completionTokens, totalTokens); if (totalTokens > 0) { - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } } catch (Exception e) { log.debug("Could not get usage via getUsage() method: {}", e.getMessage()); } - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } catch (Exception e) { log.error("Error in Minimax token extraction", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatConfig.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatConfig.java index b992ce287f..e8738b4085 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatConfig.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatConfig.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2024-05-31 10:24:39 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-08-20 21:25:13 + * @LastEditTime: 2025-08-21 12:39:58 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -13,67 +13,67 @@ */ package com.bytedesk.ai.springai.providers.ollama; -// import org.springframework.ai.chat.client.ChatClient; -// import org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor; -// import org.springframework.ai.ollama.OllamaChatModel; -// import org.springframework.ai.ollama.api.OllamaApi; -// import org.springframework.ai.ollama.api.OllamaOptions; -// import org.springframework.beans.factory.annotation.Value; -// import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; -// import org.springframework.context.annotation.Bean; -// import org.springframework.context.annotation.Configuration; -// import lombok.Data; -// import lombok.extern.slf4j.Slf4j; +import org.springframework.ai.chat.client.ChatClient; +import org.springframework.ai.chat.client.advisor.SimpleLoggerAdvisor; +import org.springframework.ai.ollama.OllamaChatModel; +import org.springframework.ai.ollama.api.OllamaApi; +import org.springframework.ai.ollama.api.OllamaOptions; +import org.springframework.beans.factory.annotation.Value; +import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; +import org.springframework.context.annotation.Bean; +import org.springframework.context.annotation.Configuration; +import lombok.Data; +import lombok.extern.slf4j.Slf4j; -// /** -// * https://ollama.com/ -// * https://www.promptingguide.ai/ -// * Ollama Chat Configuration -// */ -// @Slf4j -// @Data -// @Configuration -// @ConditionalOnProperty(prefix = "spring.ai.ollama.chat", name = "enabled", havingValue = "true", matchIfMissing = false) -// public class SpringAIOllamaChatConfig { +/** + * https://ollama.com/ + * https://www.promptingguide.ai/ + * Ollama Chat Configuration + */ +@Slf4j +@Data +@Configuration +@ConditionalOnProperty(prefix = "spring.ai.ollama.chat", name = "enabled", havingValue = "true", matchIfMissing = false) +public class SpringAIOllamaChatConfig { -// @Value("${spring.ai.ollama.base-url:http://host.docker.internal:11434}") -// private String ollamaBaseUrl; + @Value("${spring.ai.ollama.base-url:http://host.docker.internal:11434}") + private String ollamaBaseUrl; -// @Value("${spring.ai.ollama.chat.options.model:qwen3:0.6b}") -// private String ollamaChatOptionsModel; + @Value("${spring.ai.ollama.chat.options.model:qwen3:0.6b}") + private String ollamaChatOptionsModel; -// @Bean("bytedeskOllamaApi") -// OllamaApi bytedeskOllamaApi() { -// return OllamaApi.builder() -// .baseUrl(ollamaBaseUrl) -// .build(); -// } + @Bean("bytedeskOllamaApi") + OllamaApi bytedeskOllamaApi() { + return OllamaApi.builder() + .baseUrl(ollamaBaseUrl) + .build(); + } -// @Bean("bytedeskOllamaChatOptions") -// OllamaOptions bytedeskOllamaChatOptions() { -// return OllamaOptions.builder() -// .model(ollamaChatOptionsModel) -// // 使用keepAlive而不是timeout来设置超时 -// // .keepAlive("30s") // 使用30秒的超时设置 -// // .numKeep(0) // 不保留历史对话上下文 -// // .useNUMA(ollamaChatOptionsNuma) // 使用正确的方法名useNUMA而不是numa -// .build(); -// } + @Bean("bytedeskOllamaChatOptions") + OllamaOptions bytedeskOllamaChatOptions() { + return OllamaOptions.builder() + .model(ollamaChatOptionsModel) + // 使用keepAlive而不是timeout来设置超时 + // .keepAlive("30s") // 使用30秒的超时设置 + // .numKeep(0) // 不保留历史对话上下文 + // .useNUMA(ollamaChatOptionsNuma) // 使用正确的方法名useNUMA而不是numa + .build(); + } -// @Bean("bytedeskOllamaChatModel") -// OllamaChatModel bytedeskOllamaChatModel() { -// return OllamaChatModel.builder() -// .ollamaApi(bytedeskOllamaApi()) -// .defaultOptions(bytedeskOllamaChatOptions()) -// .build(); -// } + @Bean("bytedeskOllamaChatModel") + OllamaChatModel bytedeskOllamaChatModel() { + return OllamaChatModel.builder() + .ollamaApi(bytedeskOllamaApi()) + .defaultOptions(bytedeskOllamaChatOptions()) + .build(); + } -// @Bean("bytedeskOllamaChatClient") -// ChatClient bytedeskOllamaChatClient() { -// return ChatClient.builder(bytedeskOllamaChatModel()) -// .defaultOptions(bytedeskOllamaChatOptions()) -// .defaultAdvisors(new SimpleLoggerAdvisor()) -// .build(); -// } + @Bean("bytedeskOllamaChatClient") + ChatClient bytedeskOllamaChatClient() { + return ChatClient.builder(bytedeskOllamaChatModel()) + .defaultOptions(bytedeskOllamaChatOptions()) + .defaultAdvisors(new SimpleLoggerAdvisor()) + .build(); + } -// } \ No newline at end of file +} \ No newline at end of file diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatService.java new file mode 100644 index 0000000000..a9ccc9a923 --- /dev/null +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaChatService.java @@ -0,0 +1,324 @@ +/* + * @Author: jackning 270580156@qq.com + * @Date: 2025-02-26 16:59:14 + * @LastEditors: jackning 270580156@qq.com + * @LastEditTime: 2025-08-21 12:45:00 + * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk + * Please be aware of the BSL license restrictions before installing Bytedesk IM – + * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. + * Business Source License 1.1: https://github.com/Bytedesk/bytedesk/blob/main/LICENSE + * contact: 270580156@qq.com + * + * Copyright (c) 2025 by bytedesk.com, All Rights Reserved. + */ +package com.bytedesk.ai.springai.providers.ollama; + +import java.util.List; +import org.springframework.ai.chat.messages.AssistantMessage; +import org.springframework.ai.chat.model.Generation; +import org.springframework.ai.chat.prompt.Prompt; +import org.springframework.ai.ollama.OllamaChatModel; +import org.springframework.ai.ollama.api.OllamaApi; +import org.springframework.ai.ollama.api.OllamaOptions; +import org.springframework.beans.factory.annotation.Autowired; +import org.springframework.beans.factory.annotation.Qualifier; +import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; +import org.springframework.http.HttpStatus; +import org.springframework.stereotype.Service; +import org.springframework.util.Assert; +import org.springframework.util.StringUtils; +import org.springframework.web.client.HttpClientErrorException; +import org.springframework.web.servlet.mvc.method.annotation.SseEmitter; + +import com.bytedesk.ai.robot.RobotLlm; +import com.bytedesk.ai.robot.RobotProtobuf; +import com.bytedesk.ai.springai.service.BaseSpringAIService; +import com.bytedesk.ai.springai.service.ChatTokenUsage; +import com.bytedesk.core.constant.LlmConsts; +import com.bytedesk.core.message.MessageProtobuf; +import com.bytedesk.core.message.MessageTypeEnum; + +import lombok.extern.slf4j.Slf4j; + +@Slf4j +@Service +@ConditionalOnProperty(prefix = "spring.ai.ollama.chat", name = "enabled", havingValue = "true", matchIfMissing = false) +public class SpringAIOllamaChatService extends BaseSpringAIService { + + @Autowired + @Qualifier("bytedeskOllamaApi") + private OllamaApi bytedeskOllamaApi; + + @Autowired(required = false) + @Qualifier("bytedeskOllamaChatModel") + private OllamaChatModel bytedeskOllamaChatModel; + + public SpringAIOllamaChatService() { + super(); // 调用基类的无参构造函数 + } + + /** + * 根据机器人配置创建动态的OllamaOptions + * + * @param llm 机器人LLM配置 + * @return 根据机器人配置创建的选项 + */ + private OllamaOptions createOllamaOptions(RobotLlm llm) { + return super.createDynamicOptions(llm, robotLlm -> OllamaOptions.builder() + .model(robotLlm.getTextModel()) + .temperature(robotLlm.getTemperature()) + .topP(robotLlm.getTopP()) + .topK(robotLlm.getTopK()) + .build()); + } + + public OllamaApi createOllamaApi(String apiUrl) { + return OllamaApi.builder() + .baseUrl(apiUrl) + .build(); + } + + /** + * 检查模型是否存在 + * + * @param modelName 模型名称 + * @return 如果模型存在返回true，否则返回false + */ + public Boolean isModelExists(OllamaRequest request) { + OllamaApi ollamaApi = createOllamaApi(request.getApiUrl()); + String modelName = request.getModel(); + Assert.hasText(modelName, "Model name must not be null or empty"); + try { + ollamaApi.showModel(new OllamaApi.ShowModelRequest(modelName)); + return true; + } catch (HttpClientErrorException e) { + if (e.getStatusCode() == HttpStatus.NOT_FOUND) { + return false; + } + log.error("检查模型是否存在时发生错误: {}, 状态码: {}", modelName, e.getStatusCode()); + } catch (Exception e) { + log.error("检查模型是否存在时发生未知错误: {}, 错误: {}", modelName, e.getMessage()); + } + return false; + } + + protected void processPromptWebsocket(Prompt prompt, RobotProtobuf robot, MessageProtobuf messageProtobufQuery, + MessageProtobuf messageProtobufReply, String fullPromptContent) { + // 从robot中获取llm配置 + RobotLlm llm = robot.getLlm(); + log.info("Ollama API websocket fullPromptContent: {}", fullPromptContent); + if (llm == null) { + log.info("Ollama API not available"); + sendMessageWebsocket(MessageTypeEnum.ERROR, "Ollama service is not available", messageProtobufReply); + return; + } + + // 获取适当的模型实例 + OllamaChatModel chatModel = bytedeskOllamaChatModel; + if (chatModel == null) { + log.info("Ollama API not available"); + sendMessageWebsocket(MessageTypeEnum.ERROR, "Ollama service is not available", messageProtobufReply); + return; + } + + long startTime = System.currentTimeMillis(); + final boolean[] success = { false }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; + + try { + chatModel.stream(prompt).subscribe( + response -> { + if (response != null) { + log.info("Ollama API response metadata: {}", response.getMetadata()); + List generations = response.getResults(); + for (Generation generation : generations) { + AssistantMessage assistantMessage = generation.getOutput(); + String textContent = assistantMessage.getText(); + log.info("Ollama API Websocket response text: {}", textContent); + + sendMessageWebsocket(MessageTypeEnum.STREAM, textContent, messageProtobufReply); + } + // 提取token使用情况 + tokenUsage[0] = extractTokenUsage(response); + success[0] = true; + } + }, + error -> { + log.error("Ollama API error: ", error); + sendMessageWebsocket(MessageTypeEnum.ERROR, "服务暂时不可用，请稍后重试", messageProtobufReply); + success[0] = false; + }, + () -> { + log.info("Chat stream completed"); + // 记录token使用情况 + long responseTime = System.currentTimeMillis() - startTime; + String modelType = (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() + : "llama2"; + recordAiTokenUsage(robot, LlmConsts.OLLAMA, modelType, + tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), success[0], + responseTime); + }); + } catch (Exception e) { + log.error("Error processing Ollama prompt", e); + sendMessageWebsocket(MessageTypeEnum.ERROR, "服务暂时不可用，请稍后重试", messageProtobufReply); + success[0] = false; + // 记录token使用情况 + long responseTime = System.currentTimeMillis() - startTime; + String modelType = (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() : "llama2"; + recordAiTokenUsage(robot, LlmConsts.OLLAMA, modelType, + tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), success[0], responseTime); + } + } + + protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { + long startTime = System.currentTimeMillis(); + boolean success = false; + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); + + log.info("Ollama API sync fullPromptContent: {}", fullPromptContent); + + // 从robot中获取llm配置 + RobotLlm llm = robot.getLlm(); + log.info("Ollama API websocket fullPromptContent: {}", fullPromptContent); + + if (llm == null) { + log.info("Ollama API not available"); + return "Ollama service is not available"; + } + + // 获取适当的模型实例 + OllamaChatModel chatModel = bytedeskOllamaChatModel; + + try { + try { + // 如果有robot参数，尝试创建自定义选项 + if (robot != null && robot.getLlm() != null) { + // 创建自定义选项 + OllamaOptions customOptions = createOllamaOptions(robot.getLlm()); + if (customOptions != null) { + // 使用自定义选项创建Prompt + Prompt prompt = new Prompt(message, customOptions); + var response = chatModel.call(prompt); + log.info("Ollama API Sync response metadata: {}", response.getMetadata()); + tokenUsage = extractTokenUsage(response); + success = true; + return extractTextFromResponse(response); + } + } + var response = chatModel.call(message); + tokenUsage = extractTokenUsage(response); + success = true; + return extractTextFromResponse(response); + } catch (Exception e) { + log.error("Ollama API sync error", e); + success = false; + return "服务暂时不可用，请稍后重试"; + } + + } catch (Exception e) { + log.error("Ollama API sync error", e); + success = false; + return "服务暂时不可用，请稍后重试"; + } finally { + // 记录token使用情况 + long responseTime = System.currentTimeMillis() - startTime; + String modelType = (robot != null && robot.getLlm() != null + && StringUtils.hasText(robot.getLlm().getTextModel())) + ? robot.getLlm().getTextModel() + : "llama2"; + recordAiTokenUsage(robot, LlmConsts.OLLAMA, modelType, + tokenUsage.getPromptTokens(), tokenUsage.getCompletionTokens(), success, responseTime); + } + } + + protected void processPromptSse(Prompt prompt, RobotProtobuf robot, MessageProtobuf messageProtobufQuery, + MessageProtobuf messageProtobufReply, SseEmitter emitter, String fullPromptContent) { + Assert.notNull(emitter, "SseEmitter must not be null"); + // 从robot中获取llm配置 + RobotLlm llm = robot.getLlm(); + log.info("Ollama API SSE fullPromptContent: {}", fullPromptContent); + + if (llm == null) { + log.info("Ollama API not available"); + sendStreamEndMessage(messageProtobufQuery, messageProtobufReply, emitter, 0, 0, 0, fullPromptContent, + LlmConsts.OLLAMA, + (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() : "llama2"); + return; + } + + // 获取适当的模型实例 + OllamaChatModel chatModel = bytedeskOllamaChatModel; + + if (chatModel == null) { + log.info("Ollama API not available"); + // 使用sendStreamEndMessage方法替代重复的代码 + sendStreamEndMessage(messageProtobufQuery, messageProtobufReply, emitter, 0, 0, 0, fullPromptContent, + LlmConsts.OLLAMA, + (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() : "llama2"); + return; + } + + long startTime = System.currentTimeMillis(); + final boolean[] success = { false }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; + + try { + // 发送初始消息，告知用户请求已收到，正在处理 + sendStreamStartMessage(messageProtobufReply, emitter, "正在思考中..."); + + chatModel.stream(prompt).subscribe( + response -> { + try { + if (response != null && !isEmitterCompleted(emitter)) { + List generations = response.getResults(); + for (Generation generation : generations) { + AssistantMessage assistantMessage = generation.getOutput(); + String textContent = assistantMessage.getText(); + log.info("Ollama API SSE response text: {}", textContent); + + sendStreamMessage(messageProtobufQuery, messageProtobufReply, emitter, textContent); + } + // 提取token使用情况 + tokenUsage[0] = extractTokenUsage(response); + success[0] = true; + } + } catch (Exception e) { + log.error("Ollama API SSE error 1: ", e); + handleSseError(e, messageProtobufQuery, messageProtobufReply, emitter); + success[0] = false; + } + }, + error -> { + log.error("Ollama API SSE error 2: ", error); + handleSseError(error, messageProtobufQuery, messageProtobufReply, emitter); + success[0] = false; + }, + () -> { + log.info("Ollama API SSE complete"); + // 发送流结束消息，包含token使用情况和prompt内容 + sendStreamEndMessage(messageProtobufQuery, messageProtobufReply, emitter, + tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), + tokenUsage[0].getTotalTokens(), fullPromptContent, LlmConsts.OLLAMA, + (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() + : "llama2"); + // 记录token使用情况 + long responseTime = System.currentTimeMillis() - startTime; + String modelType = (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() + : "llama2"; + recordAiTokenUsage(robot, LlmConsts.OLLAMA, modelType, + tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), success[0], + responseTime); + }); + } catch (Exception e) { + log.error("Error starting Ollama stream 4", e); + handleSseError(e, messageProtobufQuery, messageProtobufReply, emitter); + success[0] = false; + // 记录token使用情况 + long responseTime = System.currentTimeMillis() - startTime; + String modelType = (llm != null && StringUtils.hasText(llm.getTextModel())) ? llm.getTextModel() : "llama2"; + recordAiTokenUsage(robot, LlmConsts.OLLAMA, modelType, + tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), success[0], responseTime); + } + } + +} diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaController.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaController.java index bf391f35ed..f40fd1029e 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaController.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaController.java @@ -3,7 +3,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2024-05-31 09:50:56 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-08-20 21:25:31 + * @LastEditTime: 2025-08-21 12:40:17 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -23,6 +23,7 @@ import org.springframework.ai.chat.prompt.Prompt; import org.springframework.ai.ollama.OllamaChatModel; import org.springframework.ai.ollama.api.OllamaApi; import org.springframework.ai.ollama.api.OllamaOptions; +import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; import org.springframework.http.MediaType; import org.springframework.http.ResponseEntity; import org.springframework.web.bind.annotation.GetMapping; @@ -43,7 +44,7 @@ import reactor.core.publisher.Flux; @RestController @RequestMapping("/api/v1/ollama") @RequiredArgsConstructor -// @ConditionalOnProperty(prefix = "spring.ai.ollama.chat", name = "enabled", havingValue = "true", matchIfMissing = false) +@ConditionalOnProperty(prefix = "spring.ai.ollama.chat", name = "enabled", havingValue = "true", matchIfMissing = false) public class SpringAIOllamaController { private final SpringAIOllamaService springAIOllamaService; diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaService.java index a6df0a5348..0deaa40364 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/ollama/SpringAIOllamaService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-26 16:59:14 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-08-20 10:57:53 + * @LastEditTime: 2025-08-21 12:45:56 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -36,6 +36,7 @@ import com.bytedesk.ai.provider.LlmProviderRestService; import com.bytedesk.ai.robot.RobotLlm; import com.bytedesk.ai.robot.RobotProtobuf; import com.bytedesk.ai.springai.service.BaseSpringAIService; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; @@ -154,7 +155,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = { false }; - final TokenUsage[] tokenUsage = { new TokenUsage(0, 0, 0) }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; try { chatModel.stream(prompt).subscribe( @@ -165,6 +166,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { for (Generation generation : generations) { AssistantMessage assistantMessage = generation.getOutput(); String textContent = assistantMessage.getText(); + log.info("Ollama API Websocket response text: {}", textContent); sendMessageWebsocket(MessageTypeEnum.STREAM, textContent, messageProtobufReply); } @@ -204,7 +206,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("Ollama API sync fullPromptContent: {}", fullPromptContent); @@ -230,6 +232,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { // 使用自定义选项创建Prompt Prompt prompt = new Prompt(message, customOptions); var response = chatModel.call(prompt); + log.info("Ollama API Sync response metadata: {}", response.getMetadata()); tokenUsage = extractTokenUsage(response); success = true; return extractTextFromResponse(response); @@ -291,7 +294,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = { false }; - final TokenUsage[] tokenUsage = { new TokenUsage(0, 0, 0) }; + final ChatTokenUsage[] tokenUsage = { new ChatTokenUsage(0, 0, 0) }; try { // 发送初始消息，告知用户请求已收到，正在处理 @@ -305,6 +308,7 @@ public class SpringAIOllamaService extends BaseSpringAIService { for (Generation generation : generations) { AssistantMessage assistantMessage = generation.getOutput(); String textContent = assistantMessage.getText(); + log.info("Ollama API SSE response text: {}", textContent); sendStreamMessage(messageProtobufQuery, messageProtobufReply, emitter, textContent); } diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openai/SpringAIOpenaiService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openai/SpringAIOpenaiService.java index bf68f2fba6..70669f495f 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openai/SpringAIOpenaiService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openai/SpringAIOpenaiService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 11:38:24 + * @LastEditTime: 2025-08-21 12:46:49 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -31,6 +31,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -82,7 +83,7 @@ public class SpringAIOpenaiService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 openaiChatModel.stream(requestPrompt).subscribe( @@ -120,7 +121,7 @@ public class SpringAIOpenaiService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("OpenAI API sync fullPromptContent: {}", fullPromptContent); @@ -188,7 +189,7 @@ public class SpringAIOpenaiService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; openaiChatModel.stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openrouter/SpringAIOpenrouterService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openrouter/SpringAIOpenrouterService.java index aeda40b020..db607ce9b8 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openrouter/SpringAIOpenrouterService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/openrouter/SpringAIOpenrouterService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 11:38:40 + * @LastEditTime: 2025-08-21 12:46:55 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -31,6 +31,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -83,7 +84,7 @@ public class SpringAIOpenrouterService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 openrouterChatModel.stream(requestPrompt).subscribe( @@ -122,7 +123,7 @@ public class SpringAIOpenrouterService extends BaseSpringAIService { log.info("SpringAIOpenrouterService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (openrouterChatModel == null) { @@ -184,7 +185,7 @@ public class SpringAIOpenrouterService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; openrouterChatModel.stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/siliconflow/SpringAISiliconFlowService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/siliconflow/SpringAISiliconFlowService.java index 6a55bfb6de..b903d8d32d 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/siliconflow/SpringAISiliconFlowService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/siliconflow/SpringAISiliconFlowService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 14:17:40 + * @LastEditTime: 2025-08-21 12:47:01 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -32,6 +32,7 @@ import org.springframework.util.StringUtils; import org.springframework.web.servlet.mvc.method.annotation.SseEmitter; import java.util.List; import java.util.Optional; +import com.bytedesk.ai.springai.service.ChatTokenUsage; /** * @author: https://github.com/fzj111 @@ -85,7 +86,7 @@ public class SpringAISiliconFlowService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 siliconFlowChatModel.get().stream(requestPrompt).subscribe( @@ -126,7 +127,7 @@ public class SpringAISiliconFlowService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("SiliconFlow API sync fullPromptContent: {}", fullPromptContent); @@ -197,7 +198,7 @@ public class SpringAISiliconFlowService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; siliconFlowChatModel.get().stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/tencent/SpringAITencentService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/tencent/SpringAITencentService.java index a3ce157132..344cacba9d 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/tencent/SpringAITencentService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/tencent/SpringAITencentService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 14:17:58 + * @LastEditTime: 2025-08-21 12:47:07 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -32,6 +32,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; @@ -83,7 +84,7 @@ public class SpringAITencentService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 tencentChatModel.stream(requestPrompt).subscribe( @@ -123,7 +124,7 @@ public class SpringAITencentService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("Tencent API sync fullPromptContent: {}", fullPromptContent); @@ -192,7 +193,7 @@ public class SpringAITencentService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; tencentChatModel.stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/volcengine/SpringAIVolcengineService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/volcengine/SpringAIVolcengineService.java index 5b78c050f2..013c631427 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/volcengine/SpringAIVolcengineService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/volcengine/SpringAIVolcengineService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-28 11:44:03 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 14:16:43 + * @LastEditTime: 2025-08-21 12:47:12 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -32,6 +32,7 @@ import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; import lombok.extern.slf4j.Slf4j; +import com.bytedesk.ai.springai.service.ChatTokenUsage; @Slf4j @Service @@ -82,7 +83,7 @@ public class SpringAIVolcengineService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; // 使用同一个ChatModel实例，但传入不同的选项 volcengineChatModel.stream(requestPrompt).subscribe( @@ -121,7 +122,7 @@ public class SpringAIVolcengineService extends BaseSpringAIService { log.info("SpringAIVolcengineService processPromptSync with full prompt content: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); try { if (volcengineChatModel == null) { @@ -191,7 +192,7 @@ public class SpringAIVolcengineService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; volcengineChatModel.stream(requestPrompt).subscribe( response -> { diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/zhipuai/SpringAIZhipuaiService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/zhipuai/SpringAIZhipuaiService.java index 5b6fda2a60..792466d158 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/zhipuai/SpringAIZhipuaiService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/providers/zhipuai/SpringAIZhipuaiService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-26 16:58:56 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 13:58:57 + * @LastEditTime: 2025-08-21 12:47:17 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -35,6 +35,7 @@ import com.bytedesk.ai.springai.service.BaseSpringAIService; import com.bytedesk.core.constant.LlmConsts; import com.bytedesk.core.message.MessageProtobuf; import com.bytedesk.core.message.MessageTypeEnum; +import com.bytedesk.ai.springai.service.ChatTokenUsage; import lombok.extern.slf4j.Slf4j; import reactor.core.publisher.Flux; @@ -117,7 +118,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; chatModel.stream(prompt).subscribe( response -> { @@ -160,7 +161,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); log.info("Zhipuai API sync fullPromptContent: {}", fullPromptContent); @@ -263,7 +264,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; final ChatResponse[] lastResponse = {null}; Flux responseFlux = chatModel.stream(prompt); @@ -341,16 +342,16 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { * @param response ChatResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractZhipuaiTokenUsage(org.springframework.ai.chat.model.ChatResponse response) { + private ChatTokenUsage extractZhipuaiTokenUsage(org.springframework.ai.chat.model.ChatResponse response) { try { if (response == null) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var metadata = response.getMetadata(); if (metadata == null) { log.warn("Zhipuai API response metadata is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Zhipuai API manual token extraction - metadata: {}", metadata); @@ -440,7 +441,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API manual token extraction result from string parsing - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } @@ -491,7 +492,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API manual token extraction result - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } @@ -505,7 +506,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { // 方法4: 如果手动提取失败，尝试使用原始的extractTokenUsage方法作为后备 log.info("Zhipuai API manual extraction failed, trying original extractTokenUsage method"); - TokenUsage fallbackUsage = extractTokenUsage(response); + ChatTokenUsage fallbackUsage = extractTokenUsage(response); if (fallbackUsage.getTotalTokens() > 0) { log.info("Zhipuai API fallback extraction successful: {}", fallbackUsage); return fallbackUsage; @@ -513,7 +514,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { // 方法5: 如果所有方法都失败，尝试估算token使用量 log.info("Zhipuai API all extraction methods failed, attempting to estimate token usage"); - TokenUsage estimatedUsage = estimateTokenUsageFromResponse(response); + ChatTokenUsage estimatedUsage = estimateTokenUsageFromResponse(response); if (estimatedUsage.getTotalTokens() > 0) { log.info("Zhipuai API estimated token usage: {}", estimatedUsage); return estimatedUsage; @@ -525,7 +526,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { // 如果手动提取出错，尝试使用原始的extractTokenUsage方法作为后备 try { log.info("Zhipuai API manual extraction error, trying original extractTokenUsage method"); - TokenUsage fallbackUsage = extractTokenUsage(response); + ChatTokenUsage fallbackUsage = extractTokenUsage(response); if (fallbackUsage.getTotalTokens() > 0) { log.info("Zhipuai API fallback extraction successful after error: {}", fallbackUsage); return fallbackUsage; @@ -536,7 +537,7 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { } log.warn("Zhipuai API all token extraction methods failed, returning zeros"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } /** @@ -545,10 +546,10 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { * @param response ChatResponse对象 * @return 估算的TokenUsage对象 */ - private TokenUsage estimateTokenUsageFromResponse(org.springframework.ai.chat.model.ChatResponse response) { + private ChatTokenUsage estimateTokenUsageFromResponse(org.springframework.ai.chat.model.ChatResponse response) { try { if (response == null) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 获取输出文本 @@ -563,11 +564,11 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API estimated tokens - output: {} chars -> {} tokens, estimated prompt: {} tokens, total: {} tokens", outputText.length(), completionTokens, promptTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } catch (Exception e) { log.error("Error estimating token usage", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -639,15 +640,15 @@ public class SpringAIZhipuaiService extends BaseSpringAIService { } // 测试手动token提取 - TokenUsage manualUsage = extractZhipuaiTokenUsage(response); + ChatTokenUsage manualUsage = extractZhipuaiTokenUsage(response); log.info("Zhipuai API test manual token extraction result: {}", manualUsage); // 测试原始token提取 - TokenUsage originalUsage = extractTokenUsage(response); + ChatTokenUsage originalUsage = extractTokenUsage(response); log.info("Zhipuai API test original token extraction result: {}", originalUsage); // 测试token估算功能 - TokenUsage estimatedUsage = estimateTokenUsageFromResponse(response); + ChatTokenUsage estimatedUsage = estimateTokenUsageFromResponse(response); log.info("Zhipuai API test estimated token usage result: {}", estimatedUsage); // 测试token估算算法 diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/service/BaseSpringAIService.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/service/BaseSpringAIService.java index dc7e823095..b8a425e52f 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/springai/service/BaseSpringAIService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/service/BaseSpringAIService.java @@ -721,11 +721,6 @@ public abstract class BaseSpringAIService implements SpringAIService { processPromptSse(aiPrompt, robot, messageProtobufQuery, messageProtobufReply, emitter, fullPromptContent); } - // private String createAndProcessPromptSync(String query, String context, RobotProtobuf robot, - // MessageProtobuf messageProtobufQuery, MessageProtobuf messageProtobufReply) { - // return createAndProcessPromptSyncWithPrompt(query, context, robot, messageProtobufQuery, messageProtobufReply).getResponse(); - // } - private PromptResult createAndProcessPromptSyncWithPrompt(String query, String context, RobotProtobuf robot, MessageProtobuf messageProtobufQuery, MessageProtobuf messageProtobufReply) { @@ -1145,7 +1140,7 @@ public abstract class BaseSpringAIService implements SpringAIService { * @param response ChatResponse from AI service * @return TokenUsage object containing prompt and completion tokens */ - protected TokenUsage extractTokenUsage(Object response) { + protected ChatTokenUsage extractTokenUsage(Object response) { try { if (response instanceof org.springframework.ai.chat.model.ChatResponse) { org.springframework.ai.chat.model.ChatResponse chatResponse = (org.springframework.ai.chat.model.ChatResponse) response; @@ -1306,7 +1301,7 @@ public abstract class BaseSpringAIService implements SpringAIService { log.info("BaseSpringAIService extractTokenUsage extracted tokens - prompt: {}, completion: {}, total: {}", prompt, completion, total); - return new TokenUsage(prompt, completion, total); + return new ChatTokenUsage(prompt, completion, total); } } } catch (Exception e) { @@ -1315,7 +1310,7 @@ public abstract class BaseSpringAIService implements SpringAIService { // 如果无法提取，返回默认值 log.warn("BaseSpringAIService extractTokenUsage could not extract token usage, returning zeros"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } /** @@ -1421,21 +1416,21 @@ public abstract class BaseSpringAIService implements SpringAIService { /** * Token usage data class */ - protected static class TokenUsage { - private final long promptTokens; - private final long completionTokens; - private final long totalTokens; + // protected static class TokenUsage { + // private final long promptTokens; + // private final long completionTokens; + // private final long totalTokens; - public TokenUsage(long promptTokens, long completionTokens, long totalTokens) { - this.promptTokens = promptTokens; - this.completionTokens = completionTokens; - this.totalTokens = totalTokens; - } + // public TokenUsage(long promptTokens, long completionTokens, long totalTokens) { + // this.promptTokens = promptTokens; + // this.completionTokens = completionTokens; + // this.totalTokens = totalTokens; + // } - public long getPromptTokens() { return promptTokens; } - public long getCompletionTokens() { return completionTokens; } - public long getTotalTokens() { return totalTokens; } - } + // public long getPromptTokens() { return promptTokens; } + // public long getCompletionTokens() { return completionTokens; } + // public long getTotalTokens() { return totalTokens; } + // } // 带prompt参数的抽象方法重载 protected abstract void processPromptWebsocket(Prompt prompt, RobotProtobuf robot, MessageProtobuf messageProtobufQuery, diff --git a/modules/ai/src/main/java/com/bytedesk/ai/springai/service/ChatTokenUsage.java b/modules/ai/src/main/java/com/bytedesk/ai/springai/service/ChatTokenUsage.java new file mode 100644 index 0000000000..ba224a4625 --- /dev/null +++ b/modules/ai/src/main/java/com/bytedesk/ai/springai/service/ChatTokenUsage.java @@ -0,0 +1,29 @@ +/* + * @Author: jackning 270580156@qq.com + * @Date: 2025-08-21 12:42:33 + * @LastEditors: jackning 270580156@qq.com + * @LastEditTime: 2025-08-21 13:05:30 + * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk + * Please be aware of the BSL license restrictions before installing Bytedesk IM – + * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. + * Business Source License 1.1: https://github.com/Bytedesk/bytedesk/blob/main/LICENSE + * contact: 270580156@qq.com + * + * Copyright (c) 2025 by bytedesk.com, All Rights Reserved. + */ +package com.bytedesk.ai.springai.service; + +import lombok.AllArgsConstructor; +import lombok.Getter; +import lombok.NoArgsConstructor; +import lombok.Setter; + +@Getter +@Setter +@AllArgsConstructor +@NoArgsConstructor +public class ChatTokenUsage { + private long promptTokens; + private long completionTokens; + private long totalTokens; +} diff --git a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatConfig.java b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatConfig.java index b27816b662..632f172471 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatConfig.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatConfig.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-19 09:39:15 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 09:33:50 + * @LastEditTime: 2025-08-21 12:22:06 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -41,7 +41,7 @@ public class ZhipuaiChatConfig { @Value("${spring.ai.zhipuai.api-key:}") private String apiKey; - @Value("${spring.ai.zhipuai.chat.options.model:glm-4}") + @Value("${spring.ai.zhipuai.chat.options.model:glm-4-flash}") private String model; @Value("${spring.ai.zhipuai.chat.options.temperature:0.7}") diff --git a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatService.java b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatService.java new file mode 100644 index 0000000000..332062e279 --- /dev/null +++ b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiChatService.java @@ -0,0 +1,624 @@ +/* + * @Author: jackning 270580156@qq.com + * @Date: 2025-08-21 12:26:02 + * @LastEditors: jackning 270580156@qq.com + * @LastEditTime: 2025-08-21 12:37:21 + * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk + * Please be aware of the BSL license restrictions before installing Bytedesk IM – + * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. + * Business Source License 1.1: https://github.com/Bytedesk/bytedesk/blob/main/LICENSE + * contact: 270580156@qq.com + * + * Copyright (c) 2025 by bytedesk.com, All Rights Reserved. + */ +package com.bytedesk.ai.zhipuai; + +import java.io.File; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.List; +import java.util.Map; + +import org.springframework.beans.factory.annotation.Autowired; +import org.springframework.beans.factory.annotation.Qualifier; +import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; +import org.springframework.stereotype.Service; + +import com.zhipu.oapi.ClientV4; +import com.zhipu.oapi.Constants; +import com.zhipu.oapi.service.v4.model.ChatCompletionRequest; +import com.zhipu.oapi.service.v4.model.ChatFunction; +import com.zhipu.oapi.service.v4.model.ChatMessage; +import com.zhipu.oapi.service.v4.model.ChatMessageAccumulator; +import com.zhipu.oapi.service.v4.model.ChatMessageRole; +import com.zhipu.oapi.service.v4.model.ChatMeta; +import com.zhipu.oapi.service.v4.model.ChatTool; +import com.zhipu.oapi.service.v4.model.ChatToolType; +import com.zhipu.oapi.service.v4.model.ModelApiResponse; +import com.zhipu.oapi.service.v4.model.ModelData; +import com.zhipu.oapi.service.v4.model.QueryModelResultRequest; +import com.zhipu.oapi.service.v4.model.QueryModelResultResponse; +import com.zhipu.oapi.service.v4.model.WebSearch; + +import lombok.extern.slf4j.Slf4j; +import reactor.core.publisher.Flux; + +@Slf4j +@Service +@ConditionalOnProperty(prefix = "spring.ai.zhipuai.chat", name = "enabled", havingValue = "true", matchIfMissing = false) +public class ZhipuaiChatService { + + @Autowired + @Qualifier("zhipuaiChatClient") + private ClientV4 client; + + @Autowired + private ZhipuaiChatConfig zhipuaiChatConfig; + + /** + * 角色扮演聊天 + */ + public String rolePlayChat(String message, String userInfo, String botInfo, String botName, String userName) { + // 添加请求日志 + log.info("Zhipuai API role play request - message length: {}, userInfo: {}, botInfo: {}, botName: {}, userName: {}", + message.length(), userInfo, botInfo, botName, userName); + + long startTime = System.currentTimeMillis(); + + // 使用默认client进行角色扮演聊天 + ClientV4 chatClient = client; + + try { + if (client == null) { + log.error("Zhipuai API client is null"); + return "Zhipuai client is not available"; + } + + List messages = new ArrayList<>(); + ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); + messages.add(chatMessage); + + ChatMeta meta = new ChatMeta(); + meta.setUser_info(userInfo); + meta.setBot_info(botInfo); + meta.setBot_name(botName); + meta.setUser_name(userName); + + String requestId = String.format("roleplay-%d", System.currentTimeMillis()); + + ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + .model(Constants.ModelCharGLM3) + .stream(Boolean.FALSE) + .invokeMethod(Constants.invokeMethod) + .messages(messages) + .meta(meta) + .requestId(requestId) + .build(); + + log.info("Zhipuai API role play invoking model with requestId: {}", requestId); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + if (response.isSuccess() && response.getData() != null) { + log.info("Zhipuai API role play response success"); + + // 提取token使用情况 + // TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); + // log.info("Zhipuai API role play tokenUsage: {}", tokenUsage); + + Object content = response.getData().getChoices().get(0).getMessage().getContent(); + return content != null ? content.toString() : null; + } else { + log.error("Zhipuai API role play error: {}", response.getError()); + return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); + } + } catch (Exception e) { + log.error("Zhipuai API role play error: ", e); + return "Error: " + e.getMessage(); + } finally { + long responseTime = System.currentTimeMillis() - startTime; + log.info("Zhipuai API role play completed in {}ms", responseTime); + } + } + + /** + * Function Calling 聊天 + */ + public String functionCallingChat(String message, List functions) { + return functionCallingChat(message, null, null, functions); + } + + /** + * Function Calling 聊天（带自定义参数） + */ + public String functionCallingChat(String message, String model, Double temperature, List functions) { + // 添加请求日志 + log.info("Zhipuai API function calling request - message length: {}, model: {}, temperature: {}, functions count: {}", + message.length(), model, temperature, functions != null ? functions.size() : 0); + + long startTime = System.currentTimeMillis(); + + // 使用默认client进行函数调用聊天 + ClientV4 chatClient = client; + + try { + + List messages = new ArrayList<>(); + ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); + messages.add(chatMessage); + + List chatToolList = new ArrayList<>(); + if (functions != null) { + for (ChatFunction function : functions) { + ChatTool chatTool = new ChatTool(); + chatTool.setType(ChatToolType.FUNCTION.value()); + chatTool.setFunction(function); + chatToolList.add(chatTool); + } + } + + String requestId = String.format("function-%d", System.currentTimeMillis()); + + ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + .model(model != null ? model : zhipuaiChatConfig.getModel()) + .stream(Boolean.FALSE) + .invokeMethod(Constants.invokeMethod) + .messages(messages) + .requestId(requestId) + .temperature(temperature != null ? temperature.floatValue() : (float) zhipuaiChatConfig.getTemperature()) + .tools(chatToolList) + .toolChoice("auto") + .build(); + + log.info("Zhipuai API function calling invoking model with requestId: {}", requestId); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + if (response.isSuccess() && response.getData() != null) { + log.info("Zhipuai API function calling response success"); + + // 提取token使用情况 + // TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); + // log.info("Zhipuai API function calling tokenUsage: {}", tokenUsage); + + Object content = response.getData().getChoices().get(0).getMessage().getContent(); + return content != null ? content.toString() : null; + } else { + log.error("Zhipuai API function calling error: {}", response.getError()); + return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); + } + } catch (Exception e) { + log.error("Zhipuai API function calling error: ", e); + return "Error: " + e.getMessage(); + } finally { + long responseTime = System.currentTimeMillis() - startTime; + log.info("Zhipuai API function calling completed in {}ms", responseTime); + } + } + + /** + * 流式 Function Calling 聊天 + */ + public Flux functionCallingChatStream(String message, List functions) { + return functionCallingChatStream(message, null, null, functions); + } + + /** + * 流式 Function Calling 聊天（带自定义参数） + */ + public Flux functionCallingChatStream(String message, String model, Double temperature, List functions) { + // 使用默认client进行流式函数调用聊天 + ClientV4 chatClient = client; + + try { + + List messages = new ArrayList<>(); + ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); + messages.add(chatMessage); + + List chatToolList = new ArrayList<>(); + if (functions != null) { + for (ChatFunction function : functions) { + ChatTool chatTool = new ChatTool(); + chatTool.setType(ChatToolType.FUNCTION.value()); + chatTool.setFunction(function); + chatToolList.add(chatTool); + } + } + + String requestId = String.format("function-stream-%d", System.currentTimeMillis()); + + ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + .model(model != null ? model : zhipuaiChatConfig.getModel()) + .stream(Boolean.TRUE) + .messages(messages) + .requestId(requestId) + .temperature(temperature != null ? temperature.floatValue() : (float) zhipuaiChatConfig.getTemperature()) + .tools(chatToolList) + .toolChoice("auto") + .build(); + + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + if (response.isSuccess()) { + return Flux.from(mapStreamToAccumulator(response.getFlowable()).map(accumulator -> { + log.info("Zhipuai API function calling accumulator received: {}", accumulator); + log.info("Zhipuai API function calling accumulator class: {}", accumulator.getClass().getName()); + + Object delta = accumulator.getDelta(); + log.info("Zhipuai API function calling delta: {}", delta); + log.info("Zhipuai API function calling delta class: {}", delta != null ? delta.getClass().getName() : "null"); + + // 处理tool_calls（如果有的话） + if (delta != null && delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { + com.zhipu.oapi.service.v4.model.ChatMessage deltaMessage = (com.zhipu.oapi.service.v4.model.ChatMessage) delta; + if (deltaMessage.getTool_calls() != null) { + log.info("Zhipuai API function calling tool_calls: {}", deltaMessage.getTool_calls()); + } + } + + if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { + Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); + log.info("Zhipuai API function calling content: {}", content); + return content != null ? content.toString() : ""; + } else if (delta != null) { + String deltaStr = delta.toString(); + log.info("Zhipuai API function calling delta as string: {}", deltaStr); + + // 尝试从JSON字符串中提取content字段 + // String extractedContent = extractContentFromDeltaString(deltaStr); + // if (extractedContent != null && !extractedContent.isEmpty()) { + // log.info("Zhipuai API function calling extracted content: {}", extractedContent); + // return extractedContent; + // } else if (!isEmptyAssistantMessage(deltaStr)) { + // return deltaStr; + // } else { + // return ""; + // } + } + return ""; + })); + } else { + log.error("Zhipuai API error: {}", response.getError()); + return Flux.just("Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error")); + } + } catch (Exception e) { + log.error("Error in functionCallingChatStream", e); + return Flux.just("Error: " + e.getMessage()); + } + } + + /** + * 图像生成 - 暂不支持，需要等待SDK更新 + */ + public String generateImage(String prompt) { + return "Image generation is not supported in current SDK version"; + } + + /** + * 图像生成（带请求ID）- 暂不支持，需要等待SDK更新 + */ + public String generateImage(String prompt, String requestId) { + return "Image generation is not supported in current SDK version"; + } + + /** + * 向量嵌入 - 暂不支持，需要等待SDK更新 + */ + public List getEmbedding(String text) { + log.warn("Embedding is not supported in current SDK version"); + return new ArrayList<>(); + } + + /** + * 批量向量嵌入 - 暂不支持，需要等待SDK更新 + */ + public List> getEmbeddings(List texts) { + log.warn("Embeddings is not supported in current SDK version"); + return new ArrayList<>(); + } + + /** + * 语音合成 - 暂不支持，需要等待SDK更新 + */ + public File generateSpeech(String text, String voice, String responseFormat) { + log.warn("Speech synthesis is not supported in current SDK version"); + return null; + } + + /** + * 自定义语音合成 - 暂不支持，需要等待SDK更新 + */ + public File generateCustomSpeech(String text, String voiceText, File voiceData, String responseFormat) { + log.warn("Custom voice synthesis is not supported in current SDK version"); + return null; + } + + /** + * 文件上传 - 暂不支持，需要等待SDK更新 + */ + public String uploadFile(String filePath, String purpose) { + return "File upload is not supported in current SDK version"; + } + + /** + * 查询文件列表 - 暂不支持，需要等待SDK更新 + */ + public List> queryFiles() { + log.warn("File query is not supported in current SDK version"); + return new ArrayList<>(); + } + + /** + * 下载文件内容 - 暂不支持，需要等待SDK更新 + */ + public File downloadFile(String fileId, String outputPath) { + log.warn("File download is not supported in current SDK version"); + return null; + } + + /** + * 创建微调任务 - 暂不支持，需要等待SDK更新 + */ + public String createFineTuningJob(String model, String trainingFile) { + return "Fine-tuning is not supported in current SDK version"; + } + + /** + * 查询微调任务 - 暂不支持，需要等待SDK更新 + */ + public Map queryFineTuningJob(String jobId) { + log.warn("Fine-tuning query is not supported in current SDK version"); + return new HashMap<>(); + } + + /** + * 异步聊天 + */ + public String chatAsync(String message) { + // 添加请求日志 + log.info("Zhipuai API async request - message length: {}", message.length()); + + long startTime = System.currentTimeMillis(); + + // 使用默认client进行异步聊天 + ClientV4 chatClient = client; + + try { + + List messages = new ArrayList<>(); + ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); + messages.add(chatMessage); + + ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + .model(zhipuaiChatConfig.getModel()) + .stream(Boolean.FALSE) + .invokeMethod(Constants.invokeMethodAsync) + .messages(messages) + .build(); + + log.info("Zhipuai API async invoking model"); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + if (response.isSuccess() && response.getData() != null) { + String taskId = response.getData().getId(); + log.info("Zhipuai API async task created with taskId: {}", taskId); + + // 轮询获取结果 + return pollAsyncResult(taskId); + } else { + log.error("Zhipuai API async error: {}", response.getError()); + return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); + } + } catch (Exception e) { + log.error("Zhipuai API async error: ", e); + return "Error: " + e.getMessage(); + } finally { + long responseTime = System.currentTimeMillis() - startTime; + log.info("Zhipuai API async completed in {}ms", responseTime); + } + } + + /** + * 轮询异步结果 + */ + private String pollAsyncResult(String taskId) { + log.info("Zhipuai API starting async result polling for taskId: {}", taskId); + + try { + int maxAttempts = 30; // 最多轮询30次 + int attempt = 0; + + while (attempt < maxAttempts) { + log.debug("Zhipuai API polling attempt {}/{} for taskId: {}", attempt + 1, maxAttempts, taskId); + + QueryModelResultRequest request = new QueryModelResultRequest(); + request.setTaskId(taskId); + + QueryModelResultResponse response = client.queryModelResult(request); + + if (response.isSuccess() && response.getData() != null) { + Object taskStatus = response.getData().getTaskStatus(); + log.debug("Zhipuai API task status: {} for taskId: {}", taskStatus, taskId); + + if ("SUCCESS".equals(taskStatus.toString())) { + log.info("Zhipuai API async task completed successfully for taskId: {}", taskId); + Object content = response.getData().getChoices().get(0).getMessage().getContent(); + return content != null ? content.toString() : null; + } else if ("FAILED".equals(taskStatus.toString())) { + log.error("Zhipuai API async task failed for taskId: {}", taskId); + return "Task failed"; + } + } else { + log.warn("Zhipuai API async polling response not successful for taskId: {}", taskId); + } + + attempt++; + Thread.sleep(2000); // 等待2秒后重试 + } + + log.error("Zhipuai API async task timeout after {} attempts for taskId: {}", maxAttempts, taskId); + return "Task timeout after " + maxAttempts + " attempts"; + } catch (Exception e) { + log.error("Zhipuai API error polling async result for taskId: {}", taskId, e); + return "Error: " + e.getMessage(); + } + } + + /** + * 带Web搜索的聊天 + */ + public String chatWithWebSearch(String message, String searchQuery) { + // 添加请求日志 + log.info("Zhipuai API web search request - message length: {}, searchQuery: {}", + message.length(), searchQuery); + + long startTime = System.currentTimeMillis(); + + // 使用默认client进行网络搜索聊天 + ClientV4 chatClient = client; + + try { + + List messages = new ArrayList<>(); + ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); + messages.add(chatMessage); + + List chatToolList = new ArrayList<>(); + + // 添加Web搜索工具 + ChatTool webSearchTool = new ChatTool(); + webSearchTool.setType(ChatToolType.WEB_SEARCH.value()); + WebSearch webSearch = new WebSearch(); + webSearch.setSearch_query(searchQuery); + webSearch.setSearch_result(true); + webSearch.setEnable(true); + webSearchTool.setWeb_search(webSearch); + chatToolList.add(webSearchTool); + + String requestId = String.format("websearch-%d", System.currentTimeMillis()); + + ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + .model(zhipuaiChatConfig.getModel()) + .stream(Boolean.FALSE) + .invokeMethod(Constants.invokeMethod) + .messages(messages) + .requestId(requestId) + .tools(chatToolList) + .toolChoice("auto") + .build(); + + log.info("Zhipuai API web search invoking model with requestId: {}", requestId); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + if (response.isSuccess() && response.getData() != null) { + log.info("Zhipuai API web search response success"); + + // 提取token使用情况 + // TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); + // log.info("Zhipuai API web search tokenUsage: {}", tokenUsage); + + Object content = response.getData().getChoices().get(0).getMessage().getContent(); + return content != null ? content.toString() : null; + } else { + log.error("Zhipuai API web search error: {}", response.getError()); + return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); + } + } catch (Exception e) { + log.error("Zhipuai API web search error: ", e); + return "Error: " + e.getMessage(); + } finally { + long responseTime = System.currentTimeMillis() - startTime; + log.info("Zhipuai API web search completed in {}ms", responseTime); + } + } + + /** + * 语音模型聊天 - 暂不支持，需要等待SDK更新 + */ + public String chatWithVoice(String message) { + log.warn("Voice chat is not supported in current SDK version"); + return "Voice chat is not supported in current SDK version"; + } + + + /** + * 将流式响应转换为Accumulator，参考官方示例 + */ + private io.reactivex.Flowable mapStreamToAccumulator(io.reactivex.Flowable flowable) { + return flowable.map(chunk -> { + return new ChatMessageAccumulator( + chunk.getChoices().get(0).getDelta(), + null, + chunk.getChoices().get(0), + chunk.getUsage(), + chunk.getCreated(), + chunk.getId() + ); + }); + } + + /** + * 测试流式响应功能 + * 用于调试流式响应问题 + */ + public void testStreamResponse() { + // 使用默认client进行测试 + // ClientV4 chatClient = client; + + try { + log.info("Zhipuai API testing stream response..."); + + // // 创建一个简单的测试请求 + // String testMessage = "Hello, this is a test message for stream response."; + // ChatCompletionRequest chatCompletionRequest = createDynamicRequest(null, testMessage, true); + + // log.info("Zhipuai API making stream test call with message: {}", testMessage); + + // // 调用API + // ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); + + // log.info("Zhipuai API stream test response success: {}", response.isSuccess()); + + // if (response.isSuccess()) { + // log.info("Zhipuai API stream test starting flowable processing"); + + // final int[] messageCount = {0}; + + // // 使用AtomicBoolean来标记是否是第一个消息，参考官方示例 + // // java.util.concurrent.atomic.AtomicBoolean isFirst = new java.util.concurrent.atomic.AtomicBoolean(true); + + // mapStreamToAccumulator(response.getFlowable()) + // .doOnNext(accumulator -> { + // messageCount[0]++; + // log.info("Zhipuai API stream test message #{}: accumulator={}", messageCount[0], accumulator); + // log.info("Zhipuai API stream test message #{}: accumulator class={}", messageCount[0], accumulator.getClass().getName()); + + // Object delta = accumulator.getDelta(); + // log.info("Zhipuai API stream test message #{}: delta={}", messageCount[0], delta); + // log.info("Zhipuai API stream test message #{}: delta class={}", messageCount[0], delta != null ? delta.getClass().getName() : "null"); + + // if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { + // Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); + // log.info("Zhipuai API stream test message #{}: content={}", messageCount[0], content); + // } else { + // log.info("Zhipuai API stream test message #{}: delta is not ChatMessage", messageCount[0]); + // } + // }) + // .doOnComplete(() -> { + // log.info("Zhipuai API stream test completed, total messages: {}", messageCount[0]); + // }) + // .doOnError(error -> { + // log.error("Zhipuai API stream test error: ", error); + // }) + // .subscribe(); + // } else { + // log.error("Zhipuai API stream test failed: {}", response.getError()); + // } + + } catch (Exception e) { + log.error("Zhipuai API test stream response error", e); + } + } + + + +} diff --git a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiController.java b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiController.java index 2026580336..380d2574a4 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiController.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiController.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-19 09:39:15 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 15:45:45 + * @LastEditTime: 2025-08-21 12:38:11 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -49,6 +49,7 @@ public class ZhipuaiController { private final BytedeskProperties bytedeskProperties; private final ZhipuaiService zhipuaiService; + private final ZhipuaiChatService zhipuaiChatService; private final ExecutorService executorService = Executors.newCachedThreadPool(); /** @@ -140,7 +141,7 @@ public class ZhipuaiController { } try { - String result = zhipuaiService.rolePlayChat(message, userInfo, botInfo, botName, userName); + String result = zhipuaiChatService.rolePlayChat(message, userInfo, botInfo, botName, userName); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in role play chat", e); @@ -191,7 +192,7 @@ public class ZhipuaiController { functions.add(function); } - String result = zhipuaiService.functionCallingChat(message, functions); + String result = zhipuaiChatService.functionCallingChat(message, functions); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in function calling chat", e); @@ -224,7 +225,7 @@ public class ZhipuaiController { functions.add(function); } - return zhipuaiService.functionCallingChatStream(message, functions); + return zhipuaiChatService.functionCallingChatStream(message, functions); } catch (Exception e) { log.error("Error in function calling chat stream", e); return Flux.just("Error: " + e.getMessage()); @@ -303,7 +304,7 @@ public class ZhipuaiController { functions.add(weatherFunction); String message = "请告诉我" + city + "的天气情况"; - String result = zhipuaiService.functionCallingChat(message, functions); + String result = zhipuaiChatService.functionCallingChat(message, functions); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in weather function call", e); @@ -351,7 +352,7 @@ public class ZhipuaiController { functions.add(flightFunction); String message = "请查询从" + from + "到" + to + "的航班价格"; - String result = zhipuaiService.functionCallingChat(message, functions); + String result = zhipuaiChatService.functionCallingChat(message, functions); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in flight function call", e); @@ -377,7 +378,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Prompt is required")); } - String result = zhipuaiService.generateImage(prompt, requestId); + String result = zhipuaiChatService.generateImage(prompt, requestId); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in image generation", e); @@ -402,7 +403,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Text is required")); } - List result = zhipuaiService.getEmbedding(text); + List result = zhipuaiChatService.getEmbedding(text); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in embedding", e); @@ -428,7 +429,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Texts are required")); } - List> result = zhipuaiService.getEmbeddings(texts); + List> result = zhipuaiChatService.getEmbeddings(texts); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in embeddings", e); @@ -455,7 +456,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Text is required")); } - File result = zhipuaiService.generateSpeech(text, voice, responseFormat); + File result = zhipuaiChatService.generateSpeech(text, voice, responseFormat); if (result != null) { return ResponseEntity.ok(JsonResult.success("Speech generated: " + result.getAbsolutePath())); } else { @@ -496,7 +497,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Voice data file not found")); } - File result = zhipuaiService.generateCustomSpeech(text, voiceText, voiceData, responseFormat); + File result = zhipuaiChatService.generateCustomSpeech(text, voiceText, voiceData, responseFormat); if (result != null) { return ResponseEntity.ok(JsonResult.success("Custom voice generated: " + result.getAbsolutePath())); } else { @@ -530,7 +531,7 @@ public class ZhipuaiController { purpose = "fine-tune"; } - String result = zhipuaiService.uploadFile(filePath, purpose); + String result = zhipuaiChatService.uploadFile(filePath, purpose); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in file upload", e); @@ -549,7 +550,7 @@ public class ZhipuaiController { } try { - List> result = zhipuaiService.queryFiles(); + List> result = zhipuaiChatService.queryFiles(); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in query files", e); @@ -579,7 +580,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Output path is required")); } - File result = zhipuaiService.downloadFile(fileId, outputPath); + File result = zhipuaiChatService.downloadFile(fileId, outputPath); if (result != null) { return ResponseEntity.ok(JsonResult.success("File downloaded: " + result.getAbsolutePath())); } else { @@ -613,7 +614,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Training file is required")); } - String result = zhipuaiService.createFineTuningJob(model, trainingFile); + String result = zhipuaiChatService.createFineTuningJob(model, trainingFile); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in create fine-tuning job", e); @@ -632,7 +633,7 @@ public class ZhipuaiController { } try { - Map result = zhipuaiService.queryFineTuningJob(jobId); + Map result = zhipuaiChatService.queryFineTuningJob(jobId); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in query fine-tuning job", e); @@ -657,7 +658,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Message is required")); } - String result = zhipuaiService.chatAsync(message); + String result = zhipuaiChatService.chatAsync(message); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in async chat", e); @@ -687,7 +688,7 @@ public class ZhipuaiController { searchQuery = message; // 如果没有指定搜索查询，使用消息作为搜索查询 } - String result = zhipuaiService.chatWithWebSearch(message, searchQuery); + String result = zhipuaiChatService.chatWithWebSearch(message, searchQuery); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in web search chat", e); @@ -712,7 +713,7 @@ public class ZhipuaiController { return ResponseEntity.ok(JsonResult.error("Message is required")); } - String result = zhipuaiService.chatWithVoice(message); + String result = zhipuaiChatService.chatWithVoice(message); return ResponseEntity.ok(JsonResult.success(result)); } catch (Exception e) { log.error("Error in voice chat", e); @@ -724,39 +725,39 @@ public class ZhipuaiController { * 测试流式响应功能 * GET http://127.0.0.1:9003/zhipuai/test-stream */ - @GetMapping("/test-stream") - public ResponseEntity> testStreamResponse() { - if (!bytedeskProperties.getDebug()) { - return ResponseEntity.ok(JsonResult.error("Zhipuai service is not available")); - } + // @GetMapping("/test-stream") + // public ResponseEntity> testStreamResponse() { + // if (!bytedeskProperties.getDebug()) { + // return ResponseEntity.ok(JsonResult.error("Zhipuai service is not available")); + // } - try { - zhipuaiService.testStreamResponse(); - return ResponseEntity.ok(JsonResult.success("Stream response test completed. Check logs for details.")); - } catch (Exception e) { - log.error("Error testing stream response", e); - return ResponseEntity.ok(JsonResult.error("Error testing stream response: " + e.getMessage())); - } - } + // try { + // zhipuaiService.testStreamResponse(); + // return ResponseEntity.ok(JsonResult.success("Stream response test completed. Check logs for details.")); + // } catch (Exception e) { + // log.error("Error testing stream response", e); + // return ResponseEntity.ok(JsonResult.error("Error testing stream response: " + e.getMessage())); + // } + // } /** * 简单流式测试 - 完全按照官方示例代码实现 * GET http://127.0.0.1:9003/zhipuai/test-simple-stream */ - @GetMapping("/test-simple-stream") - public ResponseEntity> testSimpleStream() { - if (!bytedeskProperties.getDebug()) { - return ResponseEntity.ok(JsonResult.error("Zhipuai service is not available")); - } + // @GetMapping("/test-simple-stream") + // public ResponseEntity> testSimpleStream() { + // if (!bytedeskProperties.getDebug()) { + // return ResponseEntity.ok(JsonResult.error("Zhipuai service is not available")); + // } - try { - zhipuaiService.testSimpleStream(); - return ResponseEntity.ok(JsonResult.success("Simple stream test completed. Check logs for details.")); - } catch (Exception e) { - log.error("Error testing simple stream", e); - return ResponseEntity.ok(JsonResult.error("Error testing simple stream: " + e.getMessage())); - } - } + // try { + // zhipuaiChatService.testSimpleStream(); + // return ResponseEntity.ok(JsonResult.success("Simple stream test completed. Check logs for details.")); + // } catch (Exception e) { + // log.error("Error testing simple stream", e); + // return ResponseEntity.ok(JsonResult.error("Error testing simple stream: " + e.getMessage())); + // } + // } /** * 在 Bean 销毁时关闭线程池 diff --git a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiService.java b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiService.java index 05f1b39fd9..a9ef9dfa2f 100644 --- a/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiService.java +++ b/modules/ai/src/main/java/com/bytedesk/ai/zhipuai/ZhipuaiService.java @@ -2,7 +2,7 @@ * @Author: jackning 270580156@qq.com * @Date: 2025-02-19 09:39:15 * @LastEditors: jackning 270580156@qq.com - * @LastEditTime: 2025-07-16 14:44:47 + * @LastEditTime: 2025-08-21 12:47:50 * @Description: bytedesk.com https://github.com/Bytedesk/bytedesk * Please be aware of the BSL license restrictions before installing Bytedesk IM – * selling, reselling, or hosting Bytedesk IM as a service is a breach of the terms and automatically terminates your rights under the license. @@ -15,13 +15,15 @@ package com.bytedesk.ai.zhipuai; import java.util.ArrayList; import java.util.List; +import java.util.Optional; import org.springframework.ai.chat.prompt.Prompt; import org.springframework.beans.factory.annotation.Autowired; -import org.springframework.beans.factory.annotation.Qualifier; -import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; +// import org.springframework.boot.autoconfigure.condition.ConditionalOnProperty; import org.springframework.stereotype.Service; import org.springframework.web.servlet.mvc.method.annotation.SseEmitter; +import com.bytedesk.ai.provider.LlmProviderEntity; +import com.bytedesk.ai.provider.LlmProviderRestService; import com.bytedesk.ai.robot.RobotLlm; import com.bytedesk.ai.robot.RobotProtobuf; import com.bytedesk.ai.springai.service.BaseSpringAIService; @@ -33,21 +35,11 @@ import com.zhipu.oapi.Constants; import com.zhipu.oapi.service.v4.model.ChatCompletionRequest; import com.zhipu.oapi.service.v4.model.ChatMessage; import com.zhipu.oapi.service.v4.model.ChatMessageRole; -import com.zhipu.oapi.service.v4.model.ChatMeta; import com.zhipu.oapi.service.v4.model.ModelApiResponse; import com.zhipu.oapi.service.v4.model.ChatMessageAccumulator; -import com.zhipu.oapi.service.v4.model.ChatTool; -import com.zhipu.oapi.service.v4.model.ChatToolType; -import com.zhipu.oapi.service.v4.model.ChatFunction; import com.zhipu.oapi.service.v4.model.ModelData; -import com.zhipu.oapi.service.v4.model.QueryModelResultRequest; -import com.zhipu.oapi.service.v4.model.QueryModelResultResponse; -import com.zhipu.oapi.service.v4.model.WebSearch; -import java.io.File; -import java.util.HashMap; -import java.util.Map; import lombok.extern.slf4j.Slf4j; -import reactor.core.publisher.Flux; +import com.bytedesk.ai.springai.service.ChatTokenUsage; /** * 智谱AI服务类 @@ -56,15 +48,10 @@ import reactor.core.publisher.Flux; */ @Slf4j @Service -@ConditionalOnProperty(prefix = "spring.ai.zhipuai.chat", name = "enabled", havingValue = "true", matchIfMissing = false) public class ZhipuaiService extends BaseSpringAIService { @Autowired - @Qualifier("zhipuaiChatClient") - private ClientV4 client; - - @Autowired - private ZhipuaiChatConfig zhipuaiChatConfig; + private LlmProviderRestService llmProviderRestService; /** * 构造函数 @@ -73,10 +60,68 @@ public class ZhipuaiService extends BaseSpringAIService { super(); } + /** + * 根据机器人配置创建动态的ClientV4实例 + * + * @param llm 机器人LLM配置 + * @return 配置了特定参数的ClientV4 + */ + private ClientV4 createDynamicClient(RobotLlm llm) { + if (llm == null || llm.getTextProviderUid() == null) { + log.warn("RobotLlm or textProviderUid is null, using default client"); + // return client; // 使用默认的注入client + return null; + } + + Optional llmProviderOptional = llmProviderRestService.findByUid(llm.getTextProviderUid()); + if (llmProviderOptional.isEmpty()) { + log.warn("LlmProvider with uid {} not found, using default client", llm.getTextProviderUid()); + // return client; // 使用默认的注入client + return null; + } + + LlmProviderEntity provider = llmProviderOptional.get(); + String apiKey = provider.getApiKey(); + + if (apiKey == null || apiKey.trim().isEmpty()) { + log.warn("API key is not configured for provider {}, using default client", provider.getUid()); + // return client; // 使用默认的注入client + return null; + } + + try { + log.info("Creating dynamic Zhipuai client with provider: {} ({})", provider.getName(), provider.getUid()); + + return new ClientV4.Builder(apiKey) + .enableTokenCache() + // .networkConfig( + // zhipuaiChatConfig.getConnectionTimeout(), + // zhipuaiChatConfig.getReadTimeout(), + // zhipuaiChatConfig.getWriteTimeout(), + // zhipuaiChatConfig.getPingInterval(), + // TimeUnit.SECONDS + // ) + // .connectionPool(new okhttp3.ConnectionPool( + // zhipuaiChatConfig.getMaxIdleConnections(), + // zhipuaiChatConfig.getKeepAliveDuration(), + // TimeUnit.SECONDS + // )) + .build(); + } catch (Exception e) { + log.error("Failed to create dynamic Zhipuai client for provider {}, using default client", provider.getUid(), e); + // return client; // 使用默认的注入client + return null; + } + } + /** * 根据机器人配置创建动态的聊天选项 */ private ChatCompletionRequest createDynamicRequest(RobotLlm llm, String message, boolean stream) { + if (llm == null || llm.getTextModel() == null) { + log.warn("RobotLlm or textModel is null, using default model"); + throw new IllegalArgumentException("RobotLlm or textModel cannot be null"); + } List messages = new ArrayList<>(); ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); messages.add(chatMessage); @@ -84,12 +129,12 @@ public class ZhipuaiService extends BaseSpringAIService { String requestId = String.format("zhipuai-%d", System.currentTimeMillis()); ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(llm != null && llm.getTextModel() != null ? llm.getTextModel() : zhipuaiChatConfig.getModel()) + .model(llm.getTextModel()) .stream(stream) .invokeMethod(Constants.invokeMethod) .messages(messages) .requestId(requestId) - .temperature(llm != null ? llm.getTemperature().floatValue() : (float) zhipuaiChatConfig.getTemperature()) + .temperature(llm.getTemperature().floatValue()) .build(); // 添加请求日志 @@ -130,12 +175,12 @@ public class ZhipuaiService extends BaseSpringAIService { String requestId = String.format("zhipuai-%d", System.currentTimeMillis()); ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(llm != null && llm.getTextModel() != null ? llm.getTextModel() : zhipuaiChatConfig.getModel()) + .model(llm.getTextModel()) .stream(stream) .invokeMethod(Constants.invokeMethod) .messages(messages) .requestId(requestId) - .temperature(llm != null ? llm.getTemperature().floatValue() : (float) zhipuaiChatConfig.getTemperature()) + .temperature(llm.getTemperature().floatValue()) .build(); // 添加请求日志 @@ -152,11 +197,11 @@ public class ZhipuaiService extends BaseSpringAIService { * @param response ModelApiResponse对象 * @return TokenUsage对象 */ - private TokenUsage extractZhipuaiTokenUsage(ModelApiResponse response) { + private ChatTokenUsage extractZhipuaiTokenUsage(ModelApiResponse response) { try { if (response == null) { log.warn("Zhipuai API response is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } log.info("Zhipuai API manual token extraction - response success: {}, has data: {}", @@ -164,7 +209,7 @@ public class ZhipuaiService extends BaseSpringAIService { if (!response.isSuccess() || response.getData() == null) { log.warn("Zhipuai API response is not successful or has no data"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 尝试从response.getData()中获取usage信息 @@ -213,7 +258,7 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API manual token extraction result - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } } } catch (Exception e) { @@ -291,7 +336,7 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API manual token extraction result from string parsing - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } // 如果所有方法都失败，尝试估算token使用量 @@ -300,7 +345,7 @@ public class ZhipuaiService extends BaseSpringAIService { } catch (Exception e) { log.error("Error in manual Zhipuai token extraction", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -310,10 +355,10 @@ public class ZhipuaiService extends BaseSpringAIService { * @param response ModelApiResponse对象 * @return 估算的TokenUsage对象 */ - private TokenUsage estimateTokenUsageFromResponse(ModelApiResponse response) { + private ChatTokenUsage estimateTokenUsageFromResponse(ModelApiResponse response) { try { if (response == null || response.getData() == null) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 获取输出文本 @@ -334,11 +379,11 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API estimated tokens - output: {} chars -> {} tokens, estimated prompt: {} tokens, total: {} tokens", outputText.length(), completionTokens, promptTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } catch (Exception e) { log.error("Error estimating token usage", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -379,10 +424,10 @@ public class ZhipuaiService extends BaseSpringAIService { * @param message 输入消息 * @return 估算的TokenUsage对象 */ - private TokenUsage estimateTokenUsageFromMessage(String message) { + private ChatTokenUsage estimateTokenUsageFromMessage(String message) { try { if (message == null || message.isEmpty()) { - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } // 估算输入token数量 @@ -395,11 +440,11 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API estimated token usage from message - input: {} chars -> {} tokens, estimated output: {} tokens, total: {} tokens", message.length(), promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } catch (Exception e) { log.error("Error estimating token usage from message", e); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } } @@ -414,28 +459,19 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API websocket prompt: {}", prompt); log.info("Zhipuai API websocket fullPromptContent: {}", fullPromptContent); - // 添加请求日志 - log.info("Zhipuai API websocket request - model: {}, prompt instructions count: {}, robot: {}", - (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(), - prompt != null && prompt.getInstructions() != null ? prompt.getInstructions().size() : 0, - robot != null ? robot.getUid() : "null"); - long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; final ChatMessageAccumulator[] finalAccumulator = {null}; - try { - if (client == null) { - log.error("Zhipuai API client is null"); - sendMessageWebsocket(MessageTypeEnum.ERROR, "Zhipuai client is not available", messageProtobufReply); - return; - } + // 获取适当的client实例 + ClientV4 chatClient = createDynamicClient(llm); + try { ChatCompletionRequest chatCompletionRequest = createDynamicRequestFromPrompt(llm, prompt, true); log.info("Zhipuai API invoking model with requestId: {}", chatCompletionRequest.getRequestId()); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); if (response.isSuccess()) { log.info("Zhipuai API websocket response success, starting stream processing"); @@ -520,7 +556,7 @@ public class ZhipuaiService extends BaseSpringAIService { // 记录token使用情况 long responseTime = System.currentTimeMillis() - startTime; - String modelType = (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(); + String modelType = llm.getTextModel(); log.info("Zhipuai API websocket recording token usage - prompt: {}, completion: {}, total: {}, model: {}, responseTime: {}ms", tokenUsage[0].getPromptTokens(), tokenUsage[0].getCompletionTokens(), tokenUsage[0].getTotalTokens(), modelType, responseTime); recordAiTokenUsage(robot, LlmConsts.ZHIPUAI, modelType, @@ -550,27 +586,27 @@ public class ZhipuaiService extends BaseSpringAIService { */ @Override protected String processPromptSync(String message, RobotProtobuf robot, String fullPromptContent) { + if (robot == null || robot.getLlm() == null || robot.getLlm().getTextModel() == null) { + log.error("Robot or RobotLlm is null, cannot process prompt sync"); + return "Error: Robot or RobotLlm is not configured"; + } + // 从robot中获取llm配置 + RobotLlm llm = robot.getLlm(); // 添加请求日志 - RobotLlm llm = robot != null ? robot.getLlm() : null; - log.info("Zhipuai API sync request - model: {}, message length: {}, robot: {}", - (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(), - message.length(), robot != null ? robot.getUid() : "null"); log.info("Zhipuai API sync fullPromptContent: {}", fullPromptContent); long startTime = System.currentTimeMillis(); boolean success = false; - TokenUsage tokenUsage = new TokenUsage(0, 0, 0); + ChatTokenUsage tokenUsage = new ChatTokenUsage(0, 0, 0); + + // 获取适当的client实例 + ClientV4 chatClient = createDynamicClient(llm); try { - if (client == null) { - log.error("Zhipuai API client is null"); - return "Zhipuai client is not available"; - } - ChatCompletionRequest chatCompletionRequest = createDynamicRequest(llm, message, false); log.info("Zhipuai API invoking sync model with requestId: {}", chatCompletionRequest.getRequestId()); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); if (response.isSuccess() && response.getData() != null) { log.info("Zhipuai API sync response success"); @@ -592,7 +628,7 @@ public class ZhipuaiService extends BaseSpringAIService { } finally { // 记录token使用情况 long responseTime = System.currentTimeMillis() - startTime; - String modelType = (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(); + String modelType = llm.getTextModel(); log.info("Zhipuai API sync recording token usage - prompt: {}, completion: {}, total: {}, model: {}, responseTime: {}ms", tokenUsage.getPromptTokens(), tokenUsage.getCompletionTokens(), tokenUsage.getTotalTokens(), modelType, responseTime); recordAiTokenUsage(robot, LlmConsts.ZHIPUAI, modelType, @@ -606,38 +642,34 @@ public class ZhipuaiService extends BaseSpringAIService { @Override protected void processPromptSse(Prompt prompt, RobotProtobuf robot, MessageProtobuf messageProtobufQuery, MessageProtobuf messageProtobufReply, SseEmitter emitter, String fullPromptContent) { + if (robot == null || robot.getLlm() == null || robot.getLlm().getTextModel() == null) { + log.error("Robot or RobotLlm is null, cannot process prompt SSE"); + sendSseMessage("服务暂时不可用，请稍后重试", robot, messageProtobufQuery, messageProtobufReply, emitter); + return; + } // if (robot) // 从robot中获取llm配置 RobotLlm llm = robot.getLlm(); log.info("Zhipuai API SSE prompt: {}", prompt); log.info("Zhipuai API SSE fullPromptContent: {}", fullPromptContent); - - // 添加请求日志 - log.info("Zhipuai API SSE request - model: {}, prompt instructions count: {}, robot: {}", - (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(), - prompt != null && prompt.getInstructions() != null ? prompt.getInstructions().size() : 0, - robot != null ? robot.getUid() : "null"); - + // 发送起始消息 sendStreamStartMessage(messageProtobufReply, emitter, "正在思考中..."); long startTime = System.currentTimeMillis(); final boolean[] success = {false}; - final TokenUsage[] tokenUsage = {new TokenUsage(0, 0, 0)}; + final ChatTokenUsage[] tokenUsage = {new ChatTokenUsage(0, 0, 0)}; final ChatMessageAccumulator[] finalAccumulator = {null}; - final String modelType = (llm != null && llm.getTextModel() != null) ? llm.getTextModel() : zhipuaiChatConfig.getModel(); + final String modelType = llm.getTextModel(); + + // 获取适当的client实例 + ClientV4 chatClient = createDynamicClient(llm); try { - if (client == null) { - log.error("Zhipuai API client is null"); - handleSseError(new Exception("Zhipuai client is not available"), messageProtobufQuery, messageProtobufReply, emitter); - return; - } - ChatCompletionRequest chatCompletionRequest = createDynamicRequestFromPrompt(llm, prompt, true); log.info("Zhipuai API invoking SSE model with requestId: {}", chatCompletionRequest.getRequestId()); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); + ModelApiResponse response = chatClient.invokeModelApi(chatCompletionRequest); if (response.isSuccess()) { log.info("Zhipuai API SSE response success, starting stream processing"); @@ -653,10 +685,6 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API SSE Response: "); } - log.info("Zhipuai API SSE accumulator received: {}", accumulator); - log.info("Zhipuai API SSE accumulator class: {}", accumulator.getClass().getName()); - log.info("Zhipuai API SSE accumulator tokenUsage: {}", accumulator.getUsage()); - // 保存最新的accumulator用于token统计 finalAccumulator[0] = accumulator; @@ -782,490 +810,7 @@ public class ZhipuaiService extends BaseSpringAIService { return fullPrompt.toString().trim(); } - /** - * 角色扮演聊天 - */ - public String rolePlayChat(String message, String userInfo, String botInfo, String botName, String userName) { - // 添加请求日志 - log.info("Zhipuai API roleplay request - message length: {}, userInfo: {}, botInfo: {}, botName: {}, userName: {}", - message.length(), userInfo, botInfo, botName, userName); - - long startTime = System.currentTimeMillis(); - - try { - if (client == null) { - log.error("Zhipuai API client is null"); - return "Zhipuai client is not available"; - } - - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); - messages.add(chatMessage); - - ChatMeta meta = new ChatMeta(); - meta.setUser_info(userInfo); - meta.setBot_info(botInfo); - meta.setBot_name(botName); - meta.setUser_name(userName); - - String requestId = String.format("roleplay-%d", System.currentTimeMillis()); - - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(Constants.ModelCharGLM3) - .stream(Boolean.FALSE) - .invokeMethod(Constants.invokeMethod) - .messages(messages) - .meta(meta) - .requestId(requestId) - .build(); - - log.info("Zhipuai API roleplay invoking model with requestId: {}", requestId); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - if (response.isSuccess() && response.getData() != null) { - log.info("Zhipuai API roleplay response success"); - - // 提取token使用情况 - TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); - log.info("Zhipuai API roleplay tokenUsage: {}", tokenUsage); - - Object content = response.getData().getChoices().get(0).getMessage().getContent(); - return content != null ? content.toString() : null; - } else { - log.error("Zhipuai API roleplay error: {}", response.getError()); - return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); - } - } catch (Exception e) { - log.error("Zhipuai API roleplay error: ", e); - return "Error: " + e.getMessage(); - } finally { - long responseTime = System.currentTimeMillis() - startTime; - log.info("Zhipuai API roleplay completed in {}ms", responseTime); - } - } - - /** - * Function Calling 聊天 - */ - public String functionCallingChat(String message, List functions) { - return functionCallingChat(message, null, null, functions); - } - - /** - * Function Calling 聊天（带自定义参数） - */ - public String functionCallingChat(String message, String model, Double temperature, List functions) { - // 添加请求日志 - log.info("Zhipuai API function calling request - message length: {}, model: {}, temperature: {}, functions count: {}", - message.length(), model, temperature, functions != null ? functions.size() : 0); - - long startTime = System.currentTimeMillis(); - - try { - if (client == null) { - log.error("Zhipuai API client is null"); - return "Zhipuai client is not available"; - } - - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); - messages.add(chatMessage); - - List chatToolList = new ArrayList<>(); - if (functions != null) { - for (ChatFunction function : functions) { - ChatTool chatTool = new ChatTool(); - chatTool.setType(ChatToolType.FUNCTION.value()); - chatTool.setFunction(function); - chatToolList.add(chatTool); - } - } - - String requestId = String.format("function-%d", System.currentTimeMillis()); - - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(model != null ? model : zhipuaiChatConfig.getModel()) - .stream(Boolean.FALSE) - .invokeMethod(Constants.invokeMethod) - .messages(messages) - .requestId(requestId) - .temperature(temperature != null ? temperature.floatValue() : (float) zhipuaiChatConfig.getTemperature()) - .tools(chatToolList) - .toolChoice("auto") - .build(); - - log.info("Zhipuai API function calling invoking model with requestId: {}", requestId); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - if (response.isSuccess() && response.getData() != null) { - log.info("Zhipuai API function calling response success"); - - // 提取token使用情况 - TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); - log.info("Zhipuai API function calling tokenUsage: {}", tokenUsage); - - Object content = response.getData().getChoices().get(0).getMessage().getContent(); - return content != null ? content.toString() : null; - } else { - log.error("Zhipuai API function calling error: {}", response.getError()); - return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); - } - } catch (Exception e) { - log.error("Zhipuai API function calling error: ", e); - return "Error: " + e.getMessage(); - } finally { - long responseTime = System.currentTimeMillis() - startTime; - log.info("Zhipuai API function calling completed in {}ms", responseTime); - } - } - - /** - * 流式 Function Calling 聊天 - */ - public Flux functionCallingChatStream(String message, List functions) { - return functionCallingChatStream(message, null, null, functions); - } - - /** - * 流式 Function Calling 聊天（带自定义参数） - */ - public Flux functionCallingChatStream(String message, String model, Double temperature, List functions) { - try { - if (client == null) { - return Flux.just("Zhipuai client is not available"); - } - - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); - messages.add(chatMessage); - - List chatToolList = new ArrayList<>(); - if (functions != null) { - for (ChatFunction function : functions) { - ChatTool chatTool = new ChatTool(); - chatTool.setType(ChatToolType.FUNCTION.value()); - chatTool.setFunction(function); - chatToolList.add(chatTool); - } - } - - String requestId = String.format("function-stream-%d", System.currentTimeMillis()); - - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(model != null ? model : zhipuaiChatConfig.getModel()) - .stream(Boolean.TRUE) - .messages(messages) - .requestId(requestId) - .temperature(temperature != null ? temperature.floatValue() : (float) zhipuaiChatConfig.getTemperature()) - .tools(chatToolList) - .toolChoice("auto") - .build(); - - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - if (response.isSuccess()) { - return Flux.from(mapStreamToAccumulator(response.getFlowable()).map(accumulator -> { - log.info("Zhipuai API function calling accumulator received: {}", accumulator); - log.info("Zhipuai API function calling accumulator class: {}", accumulator.getClass().getName()); - - Object delta = accumulator.getDelta(); - log.info("Zhipuai API function calling delta: {}", delta); - log.info("Zhipuai API function calling delta class: {}", delta != null ? delta.getClass().getName() : "null"); - - // 处理tool_calls（如果有的话） - if (delta != null && delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { - com.zhipu.oapi.service.v4.model.ChatMessage deltaMessage = (com.zhipu.oapi.service.v4.model.ChatMessage) delta; - if (deltaMessage.getTool_calls() != null) { - log.info("Zhipuai API function calling tool_calls: {}", deltaMessage.getTool_calls()); - } - } - - if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { - Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); - log.info("Zhipuai API function calling content: {}", content); - return content != null ? content.toString() : ""; - } else if (delta != null) { - String deltaStr = delta.toString(); - log.info("Zhipuai API function calling delta as string: {}", deltaStr); - - // 尝试从JSON字符串中提取content字段 - String extractedContent = extractContentFromDeltaString(deltaStr); - if (extractedContent != null && !extractedContent.isEmpty()) { - log.info("Zhipuai API function calling extracted content: {}", extractedContent); - return extractedContent; - } else if (!isEmptyAssistantMessage(deltaStr)) { - return deltaStr; - } else { - return ""; - } - } - return ""; - })); - } else { - log.error("Zhipuai API error: {}", response.getError()); - return Flux.just("Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error")); - } - } catch (Exception e) { - log.error("Error in functionCallingChatStream", e); - return Flux.just("Error: " + e.getMessage()); - } - } - - /** - * 图像生成 - 暂不支持，需要等待SDK更新 - */ - public String generateImage(String prompt) { - return "Image generation is not supported in current SDK version"; - } - - /** - * 图像生成（带请求ID）- 暂不支持，需要等待SDK更新 - */ - public String generateImage(String prompt, String requestId) { - return "Image generation is not supported in current SDK version"; - } - - /** - * 向量嵌入 - 暂不支持，需要等待SDK更新 - */ - public List getEmbedding(String text) { - log.warn("Embedding is not supported in current SDK version"); - return new ArrayList<>(); - } - - /** - * 批量向量嵌入 - 暂不支持，需要等待SDK更新 - */ - public List> getEmbeddings(List texts) { - log.warn("Embeddings is not supported in current SDK version"); - return new ArrayList<>(); - } - - /** - * 语音合成 - 暂不支持，需要等待SDK更新 - */ - public File generateSpeech(String text, String voice, String responseFormat) { - log.warn("Speech synthesis is not supported in current SDK version"); - return null; - } - - /** - * 自定义语音合成 - 暂不支持，需要等待SDK更新 - */ - public File generateCustomSpeech(String text, String voiceText, File voiceData, String responseFormat) { - log.warn("Custom voice synthesis is not supported in current SDK version"); - return null; - } - - /** - * 文件上传 - 暂不支持，需要等待SDK更新 - */ - public String uploadFile(String filePath, String purpose) { - return "File upload is not supported in current SDK version"; - } - - /** - * 查询文件列表 - 暂不支持，需要等待SDK更新 - */ - public List> queryFiles() { - log.warn("File query is not supported in current SDK version"); - return new ArrayList<>(); - } - - /** - * 下载文件内容 - 暂不支持，需要等待SDK更新 - */ - public File downloadFile(String fileId, String outputPath) { - log.warn("File download is not supported in current SDK version"); - return null; - } - - /** - * 创建微调任务 - 暂不支持，需要等待SDK更新 - */ - public String createFineTuningJob(String model, String trainingFile) { - return "Fine-tuning is not supported in current SDK version"; - } - - /** - * 查询微调任务 - 暂不支持，需要等待SDK更新 - */ - public Map queryFineTuningJob(String jobId) { - log.warn("Fine-tuning query is not supported in current SDK version"); - return new HashMap<>(); - } - - /** - * 异步聊天 - */ - public String chatAsync(String message) { - // 添加请求日志 - log.info("Zhipuai API async request - message length: {}", message.length()); - - long startTime = System.currentTimeMillis(); - - try { - if (client == null) { - log.error("Zhipuai API client is null"); - return "Zhipuai client is not available"; - } - - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); - messages.add(chatMessage); - - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(zhipuaiChatConfig.getModel()) - .stream(Boolean.FALSE) - .invokeMethod(Constants.invokeMethodAsync) - .messages(messages) - .build(); - - log.info("Zhipuai API async invoking model"); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - if (response.isSuccess() && response.getData() != null) { - String taskId = response.getData().getId(); - log.info("Zhipuai API async task created with taskId: {}", taskId); - - // 轮询获取结果 - return pollAsyncResult(taskId); - } else { - log.error("Zhipuai API async error: {}", response.getError()); - return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); - } - } catch (Exception e) { - log.error("Zhipuai API async error: ", e); - return "Error: " + e.getMessage(); - } finally { - long responseTime = System.currentTimeMillis() - startTime; - log.info("Zhipuai API async completed in {}ms", responseTime); - } - } - - /** - * 轮询异步结果 - */ - private String pollAsyncResult(String taskId) { - log.info("Zhipuai API starting async result polling for taskId: {}", taskId); - - try { - int maxAttempts = 30; // 最多轮询30次 - int attempt = 0; - - while (attempt < maxAttempts) { - log.debug("Zhipuai API polling attempt {}/{} for taskId: {}", attempt + 1, maxAttempts, taskId); - - QueryModelResultRequest request = new QueryModelResultRequest(); - request.setTaskId(taskId); - - QueryModelResultResponse response = client.queryModelResult(request); - - if (response.isSuccess() && response.getData() != null) { - Object taskStatus = response.getData().getTaskStatus(); - log.debug("Zhipuai API task status: {} for taskId: {}", taskStatus, taskId); - - if ("SUCCESS".equals(taskStatus.toString())) { - log.info("Zhipuai API async task completed successfully for taskId: {}", taskId); - Object content = response.getData().getChoices().get(0).getMessage().getContent(); - return content != null ? content.toString() : null; - } else if ("FAILED".equals(taskStatus.toString())) { - log.error("Zhipuai API async task failed for taskId: {}", taskId); - return "Task failed"; - } - } else { - log.warn("Zhipuai API async polling response not successful for taskId: {}", taskId); - } - - attempt++; - Thread.sleep(2000); // 等待2秒后重试 - } - - log.error("Zhipuai API async task timeout after {} attempts for taskId: {}", maxAttempts, taskId); - return "Task timeout after " + maxAttempts + " attempts"; - } catch (Exception e) { - log.error("Zhipuai API error polling async result for taskId: {}", taskId, e); - return "Error: " + e.getMessage(); - } - } - - /** - * 带Web搜索的聊天 - */ - public String chatWithWebSearch(String message, String searchQuery) { - // 添加请求日志 - log.info("Zhipuai API web search request - message length: {}, searchQuery: {}", - message.length(), searchQuery); - - long startTime = System.currentTimeMillis(); - - try { - if (client == null) { - log.error("Zhipuai API client is null"); - return "Zhipuai client is not available"; - } - - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), message); - messages.add(chatMessage); - - List chatToolList = new ArrayList<>(); - - // 添加Web搜索工具 - ChatTool webSearchTool = new ChatTool(); - webSearchTool.setType(ChatToolType.WEB_SEARCH.value()); - WebSearch webSearch = new WebSearch(); - webSearch.setSearch_query(searchQuery); - webSearch.setSearch_result(true); - webSearch.setEnable(true); - webSearchTool.setWeb_search(webSearch); - chatToolList.add(webSearchTool); - - String requestId = String.format("websearch-%d", System.currentTimeMillis()); - - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(zhipuaiChatConfig.getModel()) - .stream(Boolean.FALSE) - .invokeMethod(Constants.invokeMethod) - .messages(messages) - .requestId(requestId) - .tools(chatToolList) - .toolChoice("auto") - .build(); - - log.info("Zhipuai API web search invoking model with requestId: {}", requestId); - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - if (response.isSuccess() && response.getData() != null) { - log.info("Zhipuai API web search response success"); - - // 提取token使用情况 - TokenUsage tokenUsage = extractZhipuaiTokenUsage(response); - log.info("Zhipuai API web search tokenUsage: {}", tokenUsage); - - Object content = response.getData().getChoices().get(0).getMessage().getContent(); - return content != null ? content.toString() : null; - } else { - log.error("Zhipuai API web search error: {}", response.getError()); - return "Error: " + (response.getError() != null ? response.getError().getMessage() : "Unknown error"); - } - } catch (Exception e) { - log.error("Zhipuai API web search error: ", e); - return "Error: " + e.getMessage(); - } finally { - long responseTime = System.currentTimeMillis() - startTime; - log.info("Zhipuai API web search completed in {}ms", responseTime); - } - } - - /** - * 语音模型聊天 - 暂不支持，需要等待SDK更新 - */ - public String chatWithVoice(String message) { - log.warn("Voice chat is not supported in current SDK version"); - return "Voice chat is not supported in current SDK version"; - } - + /** * 将流式响应转换为Accumulator，参考官方示例 */ @@ -1379,11 +924,11 @@ public class ZhipuaiService extends BaseSpringAIService { /** * 从ChatMessageAccumulator中提取token使用情况 */ - private TokenUsage extractTokenUsageFromAccumulator(ChatMessageAccumulator accumulator) { + private ChatTokenUsage extractTokenUsageFromAccumulator(ChatMessageAccumulator accumulator) { try { if (accumulator == null || accumulator.getUsage() == null) { log.warn("Zhipuai API accumulator or usage is null"); - return new TokenUsage(0, 0, 0); + return new ChatTokenUsage(0, 0, 0); } var usage = accumulator.getUsage(); @@ -1435,143 +980,77 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API accumulator token extraction result - prompt: {}, completion: {}, total: {}", promptTokens, completionTokens, totalTokens); - return new TokenUsage(promptTokens, completionTokens, totalTokens); + return new ChatTokenUsage(promptTokens, completionTokens, totalTokens); } catch (Exception e) { log.error("Error extracting token usage from accumulator", e); - return new TokenUsage(0, 0, 0); - } - } - - /** - * 测试流式响应功能 - * 用于调试流式响应问题 - */ - public void testStreamResponse() { - try { - log.info("Zhipuai API testing stream response..."); - - if (client == null) { - log.error("Zhipuai API client is null"); - return; - } - - // 创建一个简单的测试请求 - String testMessage = "Hello, this is a test message for stream response."; - ChatCompletionRequest chatCompletionRequest = createDynamicRequest(null, testMessage, true); - - log.info("Zhipuai API making stream test call with message: {}", testMessage); - - // 调用API - ModelApiResponse response = client.invokeModelApi(chatCompletionRequest); - - log.info("Zhipuai API stream test response success: {}", response.isSuccess()); - - if (response.isSuccess()) { - log.info("Zhipuai API stream test starting flowable processing"); - - final int[] messageCount = {0}; - - // 使用AtomicBoolean来标记是否是第一个消息，参考官方示例 - // java.util.concurrent.atomic.AtomicBoolean isFirst = new java.util.concurrent.atomic.AtomicBoolean(true); - - mapStreamToAccumulator(response.getFlowable()) - .doOnNext(accumulator -> { - messageCount[0]++; - log.info("Zhipuai API stream test message #{}: accumulator={}", messageCount[0], accumulator); - log.info("Zhipuai API stream test message #{}: accumulator class={}", messageCount[0], accumulator.getClass().getName()); - - Object delta = accumulator.getDelta(); - log.info("Zhipuai API stream test message #{}: delta={}", messageCount[0], delta); - log.info("Zhipuai API stream test message #{}: delta class={}", messageCount[0], delta != null ? delta.getClass().getName() : "null"); - - if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { - Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); - log.info("Zhipuai API stream test message #{}: content={}", messageCount[0], content); - } else { - log.info("Zhipuai API stream test message #{}: delta is not ChatMessage", messageCount[0]); - } - }) - .doOnComplete(() -> { - log.info("Zhipuai API stream test completed, total messages: {}", messageCount[0]); - }) - .doOnError(error -> { - log.error("Zhipuai API stream test error: ", error); - }) - .subscribe(); - } else { - log.error("Zhipuai API stream test failed: {}", response.getError()); - } - - } catch (Exception e) { - log.error("Zhipuai API test stream response error", e); + return new ChatTokenUsage(0, 0, 0); } } + /** * 简单流式测试 - 完全按照官方示例代码实现 * 用于调试流式响应问题 */ - public void testSimpleStream() { - try { - log.info("Zhipuai API testing simple stream response..."); + // public void testSimpleStream() { + // // 使用默认client进行简单测试 + // ClientV4 chatClient = client; + + // try { + // log.info("Zhipuai API testing simple stream response..."); - if (client == null) { - log.error("Zhipuai API client is null"); - return; - } + // // 完全按照官方示例代码实现 + // List messages = new ArrayList<>(); + // ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), "What is the relationship between ZhipuAI and ChatGLM?"); + // messages.add(chatMessage); - // 完全按照官方示例代码实现 - List messages = new ArrayList<>(); - ChatMessage chatMessage = new ChatMessage(ChatMessageRole.USER.value(), "What is the relationship between ZhipuAI and ChatGLM?"); - messages.add(chatMessage); + // String requestId = String.format("your-request-id-%d", System.currentTimeMillis()); + // ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() + // .model(Constants.ModelChatGLM4) + // .stream(Boolean.TRUE) + // .messages(messages) + // .requestId(requestId) + // .build(); - String requestId = String.format("your-request-id-%d", System.currentTimeMillis()); - ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder() - .model(Constants.ModelChatGLM4) - .stream(Boolean.TRUE) - .messages(messages) - .requestId(requestId) - .build(); + // log.info("Zhipuai API making simple stream test call with requestId: {}", requestId); - log.info("Zhipuai API making simple stream test call with requestId: {}", requestId); + // ModelApiResponse sseModelApiResp = chatClient.invokeModelApi(chatCompletionRequest); - ModelApiResponse sseModelApiResp = client.invokeModelApi(chatCompletionRequest); - - if (sseModelApiResp.isSuccess()) { - log.info("Zhipuai API simple stream test response success"); + // if (sseModelApiResp.isSuccess()) { + // log.info("Zhipuai API simple stream test response success"); - // java.util.concurrent.atomic.AtomicBoolean isFirst = new java.util.concurrent.atomic.AtomicBoolean(true); - final int[] messageCount = {0}; + // // java.util.concurrent.atomic.AtomicBoolean isFirst = new java.util.concurrent.atomic.AtomicBoolean(true); + // final int[] messageCount = {0}; - mapStreamToAccumulator(sseModelApiResp.getFlowable()) - .doOnNext(accumulator -> { - messageCount[0]++; - log.info("Zhipuai API simple stream test message #{}: accumulator: {}", messageCount[0], accumulator); + // mapStreamToAccumulator(sseModelApiResp.getFlowable()) + // .doOnNext(accumulator -> { + // messageCount[0]++; + // log.info("Zhipuai API simple stream test message #{}: accumulator: {}", messageCount[0], accumulator); - Object delta = accumulator.getDelta(); - log.info("Zhipuai API simple stream test message #{}: delta: {}", messageCount[0], delta); + // Object delta = accumulator.getDelta(); + // log.info("Zhipuai API simple stream test message #{}: delta: {}", messageCount[0], delta); - if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { - Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); - log.info("Zhipuai API simple stream test message #{}: content: {}", messageCount[0], content); - } - }) - .doOnComplete(() -> { - log.info("Zhipuai API simple stream test completed, total messages: {}", messageCount[0]); - }) - .doOnError(error -> { - log.error("Zhipuai API simple stream test error: ", error); - }) - .subscribe(); - } else { - log.error("Zhipuai API simple stream test failed: {}", sseModelApiResp.getError()); - } + // if (delta instanceof com.zhipu.oapi.service.v4.model.ChatMessage) { + // Object content = ((com.zhipu.oapi.service.v4.model.ChatMessage) delta).getContent(); + // log.info("Zhipuai API simple stream test message #{}: content: {}", messageCount[0], content); + // } + // }) + // .doOnComplete(() -> { + // log.info("Zhipuai API simple stream test completed, total messages: {}", messageCount[0]); + // }) + // .doOnError(error -> { + // log.error("Zhipuai API simple stream test error: ", error); + // }) + // .subscribe(); + // } else { + // log.error("Zhipuai API simple stream test failed: {}", sseModelApiResp.getError()); + // } - } catch (Exception e) { - log.error("Zhipuai API test simple stream error", e); - } - } + // } catch (Exception e) { + // log.error("Zhipuai API test simple stream error", e); + // } + // } /** * 测试content提取功能 @@ -1620,11 +1099,6 @@ public class ZhipuaiService extends BaseSpringAIService { log.info("Zhipuai API health check started"); try { - if (client == null) { - log.error("Zhipuai API health check failed: client is null"); - return false; - } - // 发送一个简单的测试请求 String response = processPromptSync("Hello", null, ""); boolean isHealthy = response != null && !response.startsWith("Error");