feat： 新增max_tokens和调整

backend/design/db_schema_pgsql.sql

@@ -412,6 +412,7 @@ CREATE TABLE biz_llm_models (
     model_code      VARCHAR(100),
     temperature     DECIMAL(3,2) DEFAULT 0.7,
     top_p           DECIMAL(3,2) DEFAULT 0.9,
+    max_tokens      BIGINT NOT NULL DEFAULT 30000,
     is_default      SMALLINT DEFAULT 0,
     sort_order      INTEGER NOT NULL DEFAULT 0,
     status          SMALLINT DEFAULT 1,

backend/src/main/java/com/imeeting/dto/biz/AiModelDTO.java

@@ -1,5 +1,6 @@
 package com.imeeting.dto.biz;
 
+import com.fasterxml.jackson.annotation.JsonProperty;
 import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Data;
 import java.math.BigDecimal;
@@ -32,6 +33,9 @@ public class AiModelDTO {
     private BigDecimal temperature;
     @Schema(description = "TopP 参数")
     private BigDecimal topP;
+    @JsonProperty("max_tokens")
+    @Schema(description = "最大输出 token 数")
+    private Long maxTokens;
     @Schema(description = "媒体配置")
     private Map<String, Object> mediaConfig;
     @Schema(description = "是否默认")

backend/src/main/java/com/imeeting/dto/biz/AiModelVO.java

@@ -1,5 +1,6 @@
 package com.imeeting.dto.biz;
 
+import com.fasterxml.jackson.annotation.JsonProperty;
 import io.swagger.v3.oas.annotations.media.Schema;
 import lombok.Data;
 import java.math.BigDecimal;
@@ -33,6 +34,9 @@ public class AiModelVO {
     private BigDecimal temperature;
     @Schema(description = "TopP 参数")
     private BigDecimal topP;
+    @JsonProperty("max_tokens")
+    @Schema(description = "最大输出 token 数")
+    private Long maxTokens;
     @Schema(description = "多媒体配置")
     private Map<String, Object> mediaConfig;
     @Schema(description = "是否为默认模型")

backend/src/main/java/com/imeeting/entity/biz/LlmModel.java

@@ -1,6 +1,7 @@
 package com.imeeting.entity.biz;
 
 import com.baomidou.mybatisplus.annotation.IdType;
+import com.baomidou.mybatisplus.annotation.TableField;
 import com.baomidou.mybatisplus.annotation.TableId;
 import com.baomidou.mybatisplus.annotation.TableName;
 import com.unisbase.entity.BaseEntity;
@@ -43,6 +44,10 @@ public class LlmModel extends BaseEntity {
     @Schema(description = "Top P参数")
     private BigDecimal topP;
 
+    @TableField("max_tokens")
+    @Schema(description = "最大输出 token 数")
+    private Long maxTokens;
+
     @Schema(description = "是否默认模型")
     private Integer isDefault;
 

backend/src/main/java/com/imeeting/service/biz/impl/AiModelServiceImpl.java

@@ -55,6 +55,7 @@ public class AiModelServiceImpl implements AiModelService {
   private static final String MEDIA_TENCENT_OFFLINE_MODEL_CODE = "tencentOfflineModelCode";
   private static final String MEDIA_TENCENT_REALTIME_MODEL_CODE = "tencentRealtimeModelCode";
     private static final int DEFAULT_SORT_ORDER = 0;
+    private static final long DEFAULT_LLM_MAX_TOKENS = 30000L;
     private static final String DEFAULT_LLM_API_PATH = "/v1/chat/completions";
     private static final String DEFAULT_ANTHROPIC_API_PATH = "/messages";
     private static final String CONNECTIVITY_TEST_SYSTEM_PROMPT = """
@@ -397,7 +398,7 @@ public class AiModelServiceImpl implements AiModelService {
         Map<String, Object> body = new LinkedHashMap<>();
         body.put("model", dto.getModelCode().trim());
         body.put("stream", false);
-        body.put("max_tokens", 32);
+        body.put("max_tokens", resolveConnectivityMaxTokens(dto.getMaxTokens()));
         if (dto.getTemperature() != null) {
             body.put("temperature", dto.getTemperature());
         }
@@ -425,7 +426,7 @@ public class AiModelServiceImpl implements AiModelService {
         Map<String, Object> body = new LinkedHashMap<>();
         body.put("model", dto.getModelCode().trim());
         body.put("system", CONNECTIVITY_TEST_SYSTEM_PROMPT);
-        body.put("max_tokens", 32);
+        body.put("max_tokens", resolveConnectivityMaxTokens(dto.getMaxTokens()));
         body.put("messages", List.of(
                 Map.of("role", "user", "content", resolveTestMessage(dto.getTestMessage()))
         ));
@@ -461,7 +462,7 @@ public class AiModelServiceImpl implements AiModelService {
                         "parts", List.of(Map.of("text", resolveTestMessage(dto.getTestMessage())))
                 )
         ));
-        if (dto.getTemperature() != null || dto.getTopP() != null) {
+        if (dto.getTemperature() != null || dto.getTopP() != null || dto.getMaxTokens() != null) {
             Map<String, Object> generationConfig = new LinkedHashMap<>();
             if (dto.getTemperature() != null) {
                 generationConfig.put("temperature", dto.getTemperature());
@@ -469,6 +470,9 @@ public class AiModelServiceImpl implements AiModelService {
             if (dto.getTopP() != null) {
                 generationConfig.put("topP", dto.getTopP());
             }
+            if (dto.getMaxTokens() != null) {
+                generationConfig.put("maxOutputTokens", resolveConnectivityMaxTokens(dto.getMaxTokens()));
+            }
             body.put("generationConfig", generationConfig);
         }
 
@@ -874,6 +878,11 @@ public class AiModelServiceImpl implements AiModelService {
         if (Integer.valueOf(1).equals(dto.getIsDefault()) && !Integer.valueOf(1).equals(dto.getStatus())) {
             throw new RuntimeException("默认模型必须为启用状态");
         }
+        if (TYPE_LLM.equals(normalizeType(dto.getModelType()))
+                && dto.getMaxTokens() != null
+                && dto.getMaxTokens() <= 0) {
+            throw new RuntimeException("max_tokens 必须为正整数");
+        }
       validateTencentAsrConfig(dto);
 //        if ("custom".equals(normalizeProvider(dto.getProvider()))) {
 //            if (TYPE_ASR.equals(normalizeType(dto.getModelType()))) {
@@ -1178,12 +1187,21 @@ public class AiModelServiceImpl implements AiModelService {
         entity.setModelCode(dto.getModelCode());
         entity.setTemperature(dto.getTemperature() == null ? BigDecimal.valueOf(0.2) : dto.getTemperature());
         entity.setTopP(dto.getTopP() == null ? BigDecimal.valueOf(0.9) : dto.getTopP());
+        entity.setMaxTokens(resolveMaxTokens(dto.getMaxTokens()));
         entity.setIsDefault(dto.getIsDefault());
         entity.setStatus(dto.getStatus());
         entity.setSortOrder(normalizeSortOrder(dto.getSortOrder()));
         entity.setRemark(dto.getRemark());
     }
 
+    private Long resolveMaxTokens(Long maxTokens) {
+        return maxTokens == null ? DEFAULT_LLM_MAX_TOKENS : maxTokens;
+    }
+
+    private Long resolveConnectivityMaxTokens(Long maxTokens) {
+        return maxTokens == null ? 32L : maxTokens;
+    }
+
     private AiModelVO toAsrVO(AsrModel entity) {
       return toAsrVO(entity, null, true);
     }
@@ -1281,6 +1299,7 @@ public class AiModelServiceImpl implements AiModelService {
         vo.setModelCode(entity.getModelCode());
         vo.setTemperature(entity.getTemperature());
         vo.setTopP(entity.getTopP());
+        vo.setMaxTokens(resolveMaxTokens(entity.getMaxTokens()));
         vo.setIsDefault(entity.getIsDefault());
         vo.setStatus(entity.getStatus());
     vo.setScope(Long.valueOf(0L).equals(entity.getTenantId()) ? "PLATFORM" : "TENANT");

backend/src/main/java/com/imeeting/service/biz/impl/AiTaskServiceImpl.java

@@ -1325,7 +1325,7 @@ public class AiTaskServiceImpl extends ServiceImpl<AiTaskMapper, AiTask> impleme
         Map<String, Object> req = new HashMap<>();
         req.put("model", llmModel.getModelCode());
         req.put("temperature", llmModel.getTemperature());
-        req.put("max_tokens", 30000);
+        req.put("max_tokens", llmModel.getMaxTokens() == null ? 30000L : llmModel.getMaxTokens());
         req.put("messages", List.of(
                 Map.of("role", "system", "content", meetingSummaryPromptAssembler.buildSystemMessage(taskRecord.getTaskConfig())),
           Map.of("role", "user", "content", meetingSummaryPromptAssembler.buildUserMessage(taskRecord.getTaskConfig(), meeting, summarySource, userPrompt))

backend/src/main/java/com/imeeting/service/biz/impl/MeetingExternalSummaryWebhookTrigger.java

@@ -200,6 +200,7 @@ public class MeetingExternalSummaryWebhookTrigger {
         config.put("modelCode", model.getModelCode());
         config.put("temperature", model.getTemperature());
         config.put("topP", model.getTopP());
+        config.put("max_tokens", model.getMaxTokens() == null ? 30000L : model.getMaxTokens());
         return config;
     }
 

backend/src/main/java/com/imeeting/service/biz/impl/MeetingTranscriptChapterServiceImpl.java

@@ -335,7 +335,7 @@ public class MeetingTranscriptChapterServiceImpl implements MeetingTranscriptCha
             Map<String, Object> requestBody = new LinkedHashMap<>();
             requestBody.put("model", llmModel.getModelCode());
             requestBody.put("temperature", llmModel.getTemperature());
-            requestBody.put("max_tokens", 30000);
+            requestBody.put("max_tokens", llmModel.getMaxTokens() == null ? 30000L : llmModel.getMaxTokens());
             requestBody.put("messages", List.of(
               Map.of("role", "system", "content", renderChapterSystemPrompt()),
               Map.of("role", "user", "content", renderChapterUserPrompt(transcripts))
@@ -1015,7 +1015,7 @@ public class MeetingTranscriptChapterServiceImpl implements MeetingTranscriptCha
             Map<String, Object> requestBody = new LinkedHashMap<>();
             requestBody.put("model", llmModel.getModelCode());
             requestBody.put("temperature", llmModel.getTemperature());
-            requestBody.put("max_tokens", 30000);
+            requestBody.put("max_tokens", llmModel.getMaxTokens() == null ? 30000L : llmModel.getMaxTokens());
             requestBody.put("messages", List.of(
               Map.of("role", "system", "content", renderChapterSystemPrompt()),
               Map.of("role", "user", "content", renderChapterUserPrompt(transcripts))

backend/src/test/java/com/imeeting/service/biz/impl/AiModelServiceImplTest.java

@@ -481,6 +481,64 @@ class AiModelServiceImplTest {
         assertEquals(7, captor.getValue().getSortOrder());
     }
 
+    @Test
+    void saveModelShouldPersistMaxTokensForLlm() throws Exception {
+        AsrModelMapper asrModelMapper = mock(AsrModelMapper.class);
+        LlmModelMapper llmModelMapper = mock(LlmModelMapper.class);
+        when(llmModelMapper.insert(any(LlmModel.class))).thenReturn(1);
+
+        AiModelServiceImpl service = new AiModelServiceImpl(
+                objectMapper,
+                asrModelMapper,
+                llmModelMapper
+        );
+
+        AiModelDTO dto = new AiModelDTO();
+        dto.setModelType("LLM");
+        dto.setModelName("max-token-llm");
+        dto.setProvider("openai");
+        dto.setBaseUrl("http://127.0.0.1:9000");
+        dto.setApiPath("/v1/chat/completions");
+        dto.setModelCode("gpt-test");
+        dto.setMaxTokens(8192L);
+        dto.setIsDefault(0);
+        dto.setStatus(1);
+
+        AiModelVO result = service.saveModel(dto);
+
+        ArgumentCaptor<LlmModel> captor = ArgumentCaptor.forClass(LlmModel.class);
+        verify(llmModelMapper, times(1)).insert(captor.capture());
+        assertEquals(8192L, captor.getValue().getMaxTokens());
+        assertEquals(8192L, result.getMaxTokens());
+
+        JsonNode json = objectMapper.readTree(objectMapper.writeValueAsString(result));
+        assertEquals(8192L, json.path("max_tokens").asLong());
+        assertEquals(false, json.has("maxTokens"));
+    }
+
+    @Test
+    void saveModelShouldRejectNonPositiveMaxTokensForLlm() {
+        AiModelServiceImpl service = new AiModelServiceImpl(
+                objectMapper,
+                mock(AsrModelMapper.class),
+                mock(LlmModelMapper.class)
+        );
+
+        AiModelDTO dto = new AiModelDTO();
+        dto.setModelType("LLM");
+        dto.setModelName("invalid-max-token-llm");
+        dto.setProvider("openai");
+        dto.setBaseUrl("http://127.0.0.1:9000");
+        dto.setApiPath("/v1/chat/completions");
+        dto.setModelCode("gpt-test");
+        dto.setMaxTokens(0L);
+        dto.setIsDefault(0);
+        dto.setStatus(1);
+
+        RuntimeException ex = assertThrows(RuntimeException.class, () -> service.saveModel(dto));
+        assertEquals("max_tokens 必须为正整数", ex.getMessage());
+    }
+
     @Test
     void saveModelShouldRejectDisabledDefaultModel() {
         AiModelServiceImpl service = new AiModelServiceImpl(

frontend/src/api/business/aimodel.ts

@@ -13,6 +13,7 @@ export interface AiModelVO {
   wsUrl?: string;
   temperature?: number;
   topP?: number;
+  max_tokens?: number;
   mediaConfig?: Record<string, any>;
   isDefault: number;
   status: number;
@@ -47,6 +48,7 @@ export interface AiModelDTO {
   wsUrl?: string;
   temperature?: number;
   topP?: number;
+  max_tokens?: number;
   mediaConfig?: Record<string, any>;
   isDefault: number;
   status: number;
@@ -157,6 +159,7 @@ export const testLlmModelConnectivity = (data: {
   modelCode: string;
   temperature?: number;
   topP?: number;
+  max_tokens?: number;
   testMessage: string;
 }) => {
   return http.post<{ code: string; data: boolean; msg: string }>(

frontend/src/pages/business/AiModels.tsx

@@ -323,6 +323,7 @@ const AiModels: React.FC = () => {
                 : undefined,
       temperature: values.temperature,
       topP: values.topP,
+      max_tokens: values.max_tokens,
       isDefault: values.isDefaultChecked ? 1 : 0,
       status: values.statusChecked ? 1 : 0,
       sortOrder: values.sortOrder ?? 0,
@@ -347,8 +348,8 @@ const AiModels: React.FC = () => {
 
   const handleTestConnectivity = async () => {
     if (activeType === "LLM") {
-      const values = await form.validateFields(["provider", "baseUrl", "apiPath", "modelCode"]);
+      const values = await form.validateFields(["provider", "baseUrl", "apiPath", "modelCode", "max_tokens"]);
-      const extraValues = form.getFieldsValue(["apiKey", "temperature", "topP"]);
+      const extraValues = form.getFieldsValue(["apiKey", "temperature", "topP", "max_tokens"]);
       setConnectivityLoading(true);
       try {
         await testLlmModelConnectivity({
@@ -359,6 +360,7 @@ const AiModels: React.FC = () => {
           modelCode: values.modelCode,
           temperature: extraValues.temperature,
           topP: extraValues.topP,
+          max_tokens: extraValues.max_tokens,
           testMessage: DEFAULT_LLM_TEST_MESSAGE,
         });
         message.success("LLM 连通性测试成功");
@@ -773,16 +775,38 @@ const AiModels: React.FC = () => {
                 <Input />
               </Form.Item>
               <Row gutter={16} className="app-responsive-form-row">
-                <Col xs={24} md={12}>
+                <Col xs={24} md={8}>
                     <Form.Item name="temperature" label="Temperature">
                     <InputNumber min={0} max={2} step={0.1} style={{ width: "100%" }} />
                   </Form.Item>
                 </Col>
-                <Col xs={24} md={12}>
+                <Col xs={24} md={8}>
                     <Form.Item name="topP" label="Top P">
                     <InputNumber min={0} max={1} step={0.1} style={{ width: "100%" }} />
                   </Form.Item>
                 </Col>
+                <Col xs={24} md={8}>
+                    <Form.Item
+                      name="max_tokens"
+                      label="max_tokens"
+                      rules={[
+                        { required: true, message: "请输入 max_tokens" },
+                        {
+                          validator: (_, value) => {
+                            if (value === undefined || value === null || value === "") {
+                              return Promise.resolve();
+                            }
+                            if (Number.isInteger(value) && value > 0) {
+                              return Promise.resolve();
+                            }
+                            return Promise.reject(new Error("max_tokens 必须为正整数"));
+                          },
+                        },
+                      ]}
+                    >
+                    <InputNumber min={1} precision={0} style={{ width: "100%" }} />
+                  </Form.Item>
+                </Col>
               </Row>
             </>
           )}