V4.8.17 feature (#3493)

* split tokens into input and output (#3477)

* split tokens into input and output

* query extension & tool call & question guide

* fix

* perf: input and output tokens

* perf: tool call if else

* perf: remove code

* fix: extract usage count

* fix: qa usage count

---------

Co-authored-by: heheer <heheer@sealos.io>

packages/service/core/ai/functions/createQuestionGuide.ts

@@ -1,6 +1,6 @@
 import type { ChatCompletionMessageParam } from '@fastgpt/global/core/ai/type.d';
 import { createChatCompletion } from '../config';
-import { countGptMessagesTokens } from '../../../common/string/tiktoken/index';
+import { countGptMessagesTokens, countPromptTokens } from '../../../common/string/tiktoken/index';
 import { loadRequestMessages } from '../../chat/utils';
 import { llmCompletionsBodyFormat } from '../utils';
 import {
@@ -20,7 +20,8 @@ export async function createQuestionGuide({
   customPrompt?: string;
 }): Promise<{
   result: string[];
-  tokens: number;
+  inputTokens: number;
+  outputTokens: number;
 }> {
   const concatMessages: ChatCompletionMessageParam[] = [
     ...messages,
@@ -29,6 +30,10 @@ export async function createQuestionGuide({
       content: `${customPrompt || PROMPT_QUESTION_GUIDE}\n${PROMPT_QUESTION_GUIDE_FOOTER}`
     }
   ];
+  const requestMessages = await loadRequestMessages({
+    messages: concatMessages,
+    useVision: false
+  });
 
   const { response: data } = await createChatCompletion({
     body: llmCompletionsBodyFormat(
@@ -36,10 +41,7 @@ export async function createQuestionGuide({
         model,
         temperature: 0.1,
         max_tokens: 200,
-        messages: await loadRequestMessages({
-          messages: concatMessages,
-          useVision: false
-        }),
+        messages: requestMessages,
         stream: false
       },
       model
@@ -51,13 +53,15 @@ export async function createQuestionGuide({
   const start = answer.indexOf('[');
   const end = answer.lastIndexOf(']');
 
-  const tokens = await countGptMessagesTokens(concatMessages);
+  const inputTokens = await countGptMessagesTokens(requestMessages);
+  const outputTokens = await countPromptTokens(answer);
 
   if (start === -1 || end === -1) {
     addLog.warn('Create question guide error', { answer });
     return {
       result: [],
-      tokens: 0
+      inputTokens: 0,
+      outputTokens: 0
     };
   }
 
@@ -69,14 +73,16 @@ export async function createQuestionGuide({
   try {
     return {
       result: json5.parse(jsonStr),
-      tokens
+      inputTokens,
+      outputTokens
     };
   } catch (error) {
     console.log(error);
 
     return {
       result: [],
-      tokens: 0
+      inputTokens: 0,
+      outputTokens: 0
     };
   }
 }

packages/service/core/ai/functions/queryExtension.ts

@@ -1,7 +1,7 @@
 import { replaceVariable } from '@fastgpt/global/common/string/tools';
 import { createChatCompletion } from '../config';
 import { ChatItemType } from '@fastgpt/global/core/chat/type';
-import { countGptMessagesTokens } from '../../../common/string/tiktoken/index';
+import { countGptMessagesTokens, countPromptTokens } from '../../../common/string/tiktoken/index';
 import { chatValue2RuntimePrompt } from '@fastgpt/global/core/chat/adapt';
 import { getLLMModel } from '../model';
 import { llmCompletionsBodyFormat } from '../utils';
@@ -121,7 +121,8 @@ export const queryExtension = async ({
   rawQuery: string;
   extensionQueries: string[];
   model: string;
-  tokens: number;
+  inputTokens: number;
+  outputTokens: number;
 }> => {
   const systemFewShot = chatBg
     ? `Q: 对话背景。
@@ -166,7 +167,8 @@ A: ${chatBg}
       rawQuery: query,
       extensionQueries: [],
       model,
-      tokens: 0
+      inputTokens: 0,
+      outputTokens: 0
     };
   }
 
@@ -181,7 +183,8 @@ A: ${chatBg}
       rawQuery: query,
       extensionQueries: Array.isArray(queries) ? queries : [],
       model,
-      tokens: await countGptMessagesTokens(messages)
+      inputTokens: await countGptMessagesTokens(messages),
+      outputTokens: await countPromptTokens(answer)
     };
   } catch (error) {
     addLog.error(`Query extension error`, error);
@@ -189,7 +192,8 @@ A: ${chatBg}
       rawQuery: query,
       extensionQueries: [],
       model,
-      tokens: 0
+      inputTokens: 0,
+      outputTokens: 0
     };
   }
 };

packages/service/core/ai/model.ts

@@ -4,6 +4,7 @@ export const getLLMModel = (model?: string) => {
     global.llmModels[0]
   );
 };
+
 export const getDatasetModel = (model?: string) => {
   return (
     global.llmModels