fix: usage

packages/service/core/ai/llm/agentCall.ts

@@ -98,7 +98,10 @@ export const runAgentCall = async ({
     // 对请求的 requestMessages 进行压缩
     const taskDescription = currentStep.description || currentStep.title;
     if (taskDescription) {
-      requestMessages = await compressRequestMessages(requestMessages, model, taskDescription);
+      const result = await compressRequestMessages(requestMessages, model, taskDescription);
+      requestMessages = result.messages;
+      inputTokens += result.usage?.inputTokens || 0;
+      outputTokens += result.usage?.outputTokens || 0;
     }
 
     // Request LLM

packages/service/core/ai/llm/compress/index.ts

@@ -6,6 +6,8 @@ import { createLLMResponse } from '../request';
 import { ChatCompletionRequestMessageRoleEnum } from '@fastgpt/global/core/ai/constants';
 import type { ChatCompletionMessageParam } from '@fastgpt/global/core/ai/type';
 import { getCompressRequestMessagesPrompt } from './prompt';
+import type { ChatNodeUsageType } from '@fastgpt/global/support/wallet/bill/type';
+import { formatModelChars2Points } from '../../../../support/wallet/usage/utils';
 
 /**
  * Compress a single oversized tool response
@@ -210,8 +212,14 @@ export const compressRequestMessages = async (
   messages: ChatCompletionMessageParam[],
   model: LLMModelItemType,
   currentDescription: string
-): Promise<ChatCompletionMessageParam[]> => {
-  if (!messages || messages.length === 0) return messages;
+): Promise<{
+  messages: ChatCompletionMessageParam[];
+  usage?: ChatNodeUsageType;
+}> => {
+  if (!messages || messages.length === 0)
+    return {
+      messages
+    };
 
   const tokenCount = await countGptMessagesTokens(messages);
   const thresholds = calculateCompressionThresholds(model.maxContext);
@@ -225,7 +233,9 @@ export const compressRequestMessages = async (
 
   if (tokenCount <= maxTokenThreshold) {
     console.log('messages 无需压缩，共', messages.length, '条消息');
-    return messages;
+    return {
+      messages
+    };
   }
 
   addLog.info('Start compressing agent messages', {
@@ -243,7 +253,7 @@ export const compressRequestMessages = async (
   const userPrompt = '请执行压缩操作，严格按照JSON格式返回结果。';
 
   try {
-    const { answerText } = await createLLMResponse({
+    const { answerText, usage } = await createLLMResponse({
       body: {
         model,
         messages: [
@@ -263,22 +273,38 @@ export const compressRequestMessages = async (
 
     if (!answerText) {
       addLog.warn('Compression failed: empty response, return original messages');
-      return messages;
+      return { messages };
     }
 
+    const { totalPoints, modelName } = formatModelChars2Points({
+      model: model.model,
+      inputTokens: usage.inputTokens,
+      outputTokens: usage.outputTokens
+    });
+    const compressedUsage = {
+      moduleName: 'Agent 对话历史压缩',
+      model: modelName,
+      totalPoints,
+      inputTokens: usage.inputTokens,
+      outputTokens: usage.outputTokens
+    };
+
     const jsonMatch =
       answerText.match(/```json\s*([\s\S]*?)\s*```/) || answerText.match(/\{[\s\S]*\}/);
     if (!jsonMatch) {
       addLog.warn('Compression failed: cannot parse JSON, return original messages');
-      return messages;
+      return { messages, usage: compressedUsage };
     }
 
     const jsonText = jsonMatch[1] || jsonMatch[0];
-    const parsed = JSON.parse(jsonText);
+    const parsed = JSON.parse(jsonText) as {
+      compressed_messages: ChatCompletionMessageParam[];
+      compression_summary: string;
+    };
 
     if (!parsed.compressed_messages || !Array.isArray(parsed.compressed_messages)) {
       addLog.warn('Compression failed: invalid format, return original messages');
-      return messages;
+      return { messages, usage: compressedUsage };
     }
 
     const compressedTokens = await countGptMessagesTokens(parsed.compressed_messages);
@@ -289,9 +315,12 @@ export const compressRequestMessages = async (
       summary: parsed.compression_summary
     });
 
-    return parsed.compressed_messages as ChatCompletionMessageParam[];
+    return {
+      messages: parsed.compressed_messages,
+      usage: compressedUsage
+    };
   } catch (error) {
     addLog.error('Compression failed', error);
-    return messages;
+    return { messages };
   }
 };

packages/service/core/workflow/dispatch/ai/agent/index.ts

@@ -142,16 +142,16 @@ export const dispatchRunAgent = async (props: DispatchAgentModuleProps): Promise
 
     /* ===== Check task complexity ===== */
     const taskIsComplexity = await (async () => {
-      if (isCheckTaskComplexityStep) {
-        const res = await checkTaskComplexity({
-          model,
-          userChatInput
-        });
-        if (res.usage) {
-          usagePush([res.usage]);
-        }
-        return res.complex;
-      }
+      // if (isCheckTaskComplexityStep) {
+      //   const res = await checkTaskComplexity({
+      //     model,
+      //     userChatInput
+      //   });
+      //   if (res.usage) {
+      //     usagePush([res.usage]);
+      //   }
+      //   return res.complex;
+      // }
 
       // 对轮运行时候，代表都是进入复杂流程
       return true;

packages/service/core/workflow/dispatch/ai/agent/master/call.ts

@@ -62,15 +62,17 @@ export const stepCall = async ({
   } = props;
 
   // Get depends on step ids
-  const { depends, usage: dependsUsage } = await getStepDependon({
-    model,
-    steps,
-    step
-  });
-  if (dependsUsage) {
-    usagePush([dependsUsage]);
+  if (!step.depends_on) {
+    const { depends, usage: dependsUsage } = await getStepDependon({
+      model,
+      steps,
+      step
+    });
+    if (dependsUsage) {
+      usagePush([dependsUsage]);
+    }
+    step.depends_on = depends;
   }
-  step.depends_on = depends;
 
   // addLog.debug(`Step information`, steps);
   const systemPromptContent = await getMasterAgentSystemPrompt({

packages/service/core/workflow/dispatch/ai/agent/sub/plan/index.ts

@@ -230,17 +230,6 @@ export const dispatchReplanAgent = async ({
 }): Promise<DispatchPlanAgentResponse> => {
   const modelData = getLLMModel(model);
 
-  // 获取依赖的步骤
-  const { depends, usage: dependsUsage } = await getStepDependon({
-    model,
-    steps: plan.steps,
-    step: {
-      id: '',
-      title: '重新规划决策依据：需要依赖哪些步骤的判断',
-      description: '本步骤分析先前的执行结果，以确定重新规划时需要依赖哪些特定步骤。'
-    }
-  });
-  const replanSteps = plan.steps.filter((step) => depends.includes(step.id));
   const requestMessages: ChatCompletionMessageParam[] = [
     {
       role: 'system',
@@ -265,11 +254,25 @@ export const dispatchReplanAgent = async ({
       tool_call_id: lastMessages.tool_calls[0].id,
       content: userInput
     });
+    // TODO: 确认这里是否有问题
     requestMessages.push({
       role: 'assistant',
       content: '请基于以上收集的用户信息，对 PLAN 进行重新规划，并严格按照 JSON Schema 输出。'
     });
   } else {
+    // 获取依赖的步骤
+    const { depends, usage: dependsUsage } = await getStepDependon({
+      model,
+      steps: plan.steps,
+      step: {
+        id: '',
+        title: '重新规划决策依据：需要依赖哪些步骤的判断',
+        description: '本步骤分析先前的执行结果，以确定重新规划时需要依赖哪些特定步骤。'
+      }
+    });
+    // TODO: 推送
+    const replanSteps = plan.steps.filter((step) => depends.includes(step.id));
+
     requestMessages.push({
       role: 'user',
       // 根据需要 replanSteps 生成用户输入
@@ -306,15 +309,13 @@ export const dispatchReplanAgent = async ({
   if (!answerText && !toolCalls.length) {
     return Promise.reject(getEmptyResponseTip());
   }
-  console.log(JSON.stringify({ answerText, toolCalls }, null, 2), 'Replan response');
+
   /* 
     正常输出情况：
     1. text: 正常生成plan
     2. toolCall: 调用ask工具
     3. text + toolCall: 可能生成 plan + 调用ask工具
   */
-
-  // 获取生成的 plan
   const rePlan = (() => {
     if (!answerText) {
       return;
@@ -345,6 +346,7 @@ export const dispatchReplanAgent = async ({
           }
         };
       } else {
+        console.log(JSON.stringify({ answerText, toolCalls }, null, 2), 'Replan response');
         return {
           type: 'agentPlanAskQuery',
           params: {
@@ -353,6 +355,9 @@ export const dispatchReplanAgent = async ({
         };
       }
     }
+
+    // RePlan 没有主动交互，则强制触发 check
+    return PlanCheckInteractive;
   })();
 
   const { totalPoints, modelName } = formatModelChars2Points({

packages/service/core/workflow/dispatch/index.ts

@@ -484,7 +484,7 @@ export const runWorkflow = async (data: RunWorkflowProps): Promise<DispatchFlowR
       const dispatchData: ModuleDispatchProps<Record<string, any>> = {
         ...data,
         mcpClientMemory,
-        usagePush: this.usagePush,
+        usagePush: this.usagePush.bind(this),
         lastInteractive: data.lastInteractive?.entryNodeIds?.includes(node.nodeId)
           ? data.lastInteractive
           : undefined,