labring · c121914yu · Dec 29, 2024 · Dec 26, 2024 · Dec 27, 2024 · Dec 27, 2024
diff --git a/packages/global/core/ai/model.d.ts b/packages/global/core/ai/model.d.ts
@@ -11,6 +11,8 @@ export type LLMModelItemType = {
   maxTemperature: number;
 
   charsPointsPrice: number; // 1k chars=n points
+  inputPrice?: number; // 1k tokens=n points
+  outputPrice?: number; // 1k tokens=n points
 
   censor?: boolean;
   vision?: boolean;

diff --git a/packages/global/core/workflow/runtime/type.d.ts b/packages/global/core/workflow/runtime/type.d.ts
@@ -107,7 +107,9 @@ export type DispatchNodeResponseType = {
   mergeSignId?: string;
 
   // bill
-  tokens?: number;
+  tokens?: number; // deprecated
+  inputTokens?: number;
+  outputTokens?: number;
   model?: string;
   contextTotalLen?: number;
   totalPoints?: number;

diff --git a/packages/global/support/wallet/bill/type.d.ts b/packages/global/support/wallet/bill/type.d.ts
@@ -23,10 +23,14 @@ export type BillSchemaType = {
 };
 
 export type ChatNodeUsageType = {
-  tokens?: number;
+  inputTokens?: number;
+  outputTokens?: number;
   totalPoints: number;
   moduleName: string;
   model?: string;
+
+  // deprecated
+  tokens?: number;
 };
 
 export type InvoiceType = {

diff --git a/packages/global/support/wallet/usage/type.d.ts b/packages/global/support/wallet/usage/type.d.ts
@@ -2,9 +2,13 @@ import { CreateUsageProps } from './api';
 import { UsageSourceEnum } from './constants';
 
 export type UsageListItemCountType = {
-  tokens?: number;
+  inputTokens?: number;
+  outputTokens?: number;
   charsLength?: number;
   duration?: number;
+
+  // deprecated
+  tokens?: number;
 };
 export type UsageListItemType = UsageListItemCountType & {
   moduleName: string;

diff --git a/packages/service/core/ai/model.ts b/packages/service/core/ai/model.ts
@@ -4,6 +4,9 @@ export const getLLMModel = (model?: string) => {
     global.llmModels[0]
   );
 };
+export const getLLMModelPriceType = () => {
+  return global.llmModels.some((item) => item.inputPrice || item.outputPrice);
+};
 export const getDatasetModel = (model?: string) => {
   return (
     global.llmModels

diff --git a/packages/service/core/workflow/dispatch/agent/classifyQuestion.ts b/packages/service/core/workflow/dispatch/agent/classifyQuestion.ts
@@ -49,7 +49,7 @@ export const dispatchClassifyQuestion = async (props: Props): Promise<CQResponse
 
   const chatHistories = getHistories(history, histories);
 
-  const { arg, tokens } = await completions({
+  const { arg, inputTokens, outputTokens } = await completions({
     ...props,
     histories: chatHistories,
     cqModel
@@ -59,7 +59,8 @@ export const dispatchClassifyQuestion = async (props: Props): Promise<CQResponse
 
   const { totalPoints, modelName } = formatModelChars2Points({
     model: cqModel.model,
-    tokens,
+    inputTokens: inputTokens,
+    outputTokens: outputTokens,
     modelType: ModelTypeEnum.llm
   });
 
@@ -72,7 +73,8 @@ export const dispatchClassifyQuestion = async (props: Props): Promise<CQResponse
       totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
       model: modelName,
       query: userChatInput,
-      tokens,
+      inputTokens: inputTokens,
+      outputTokens: outputTokens,
       cqList: agents,
       cqResult: result.value,
       contextTotalLen: chatHistories.length + 2
@@ -82,7 +84,8 @@ export const dispatchClassifyQuestion = async (props: Props): Promise<CQResponse
         moduleName: name,
         totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
         model: modelName,
-        tokens
+        inputTokens: inputTokens,
+        outputTokens: outputTokens
       }
     ]
   };
@@ -148,7 +151,8 @@ const completions = async ({
   }
 
   return {
-    tokens: await countMessagesTokens(messages),
+    inputTokens: data.usage?.prompt_tokens || 0,
+    outputTokens: data.usage?.completion_tokens || 0,
     arg: { type: id }
   };
 };
diff --git a/packages/service/core/workflow/dispatch/agent/extract.ts b/packages/service/core/workflow/dispatch/agent/extract.ts
@@ -59,7 +59,7 @@ export async function dispatchContentExtract(props: Props): Promise<Response> {
   const extractModel = getLLMModel(model);
   const chatHistories = getHistories(history, histories);
 
-  const { arg, tokens } = await (async () => {
+  const { arg, inputTokens, outputTokens } = await (async () => {
     if (extractModel.toolChoice) {
       return toolChoice({
         ...props,
@@ -114,7 +114,8 @@ export async function dispatchContentExtract(props: Props): Promise<Response> {
 
   const { totalPoints, modelName } = formatModelChars2Points({
     model: extractModel.model,
-    tokens,
+    inputTokens: inputTokens,
+    outputTokens: outputTokens,
     modelType: ModelTypeEnum.llm
   });
 
@@ -126,7 +127,8 @@ export async function dispatchContentExtract(props: Props): Promise<Response> {
       totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
       model: modelName,
       query: content,
-      tokens,
+      inputTokens,
+      outputTokens,
       extractDescription: description,
       extractResult: arg,
       contextTotalLen: chatHistories.length + 2
@@ -136,7 +138,8 @@ export async function dispatchContentExtract(props: Props): Promise<Response> {
         moduleName: name,
         totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
         model: modelName,
-        tokens
+        inputTokens,
+        outputTokens
       }
     ]
   };
@@ -249,15 +252,18 @@ const toolChoice = async (props: ActionProps) => {
     }
   })();
 
-  const completeMessages: ChatCompletionMessageParam[] = [
-    ...filterMessages,
+  const AIMessages: ChatCompletionMessageParam[] = [
     {
       role: ChatCompletionRequestMessageRoleEnum.Assistant,
       tool_calls: response.choices?.[0]?.message?.tool_calls
     }
   ];
+
+  const inputTokens = await countGptMessagesTokens(filterMessages, tools);
+  const outputTokens = await countGptMessagesTokens(AIMessages);
   return {
-    tokens: await countGptMessagesTokens(completeMessages, tools),
+    inputTokens,
+    outputTokens,
     arg
   };
 };
@@ -286,17 +292,21 @@ const functionCall = async (props: ActionProps) => {
 
   try {
     const arg = JSON.parse(response?.choices?.[0]?.message?.function_call?.arguments || '');
-    const completeMessages: ChatCompletionMessageParam[] = [
-      ...filterMessages,
+
+    const AIMessages: ChatCompletionMessageParam[] = [
       {
         role: ChatCompletionRequestMessageRoleEnum.Assistant,
         function_call: response.choices?.[0]?.message?.function_call
       }
     ];
 
+    const inputTokens = await countGptMessagesTokens(filterMessages, undefined, functions);
+    const outputTokens = await countGptMessagesTokens(AIMessages, undefined, functions);
+
     return {
       arg,
-      tokens: await countGptMessagesTokens(completeMessages, undefined, functions)
+      inputTokens,
+      outputTokens
     };
   } catch (error) {
     console.log(response.choices?.[0]?.message);
@@ -305,7 +315,8 @@ const functionCall = async (props: ActionProps) => {
 
     return {
       arg: {},
-      tokens: 0
+      inputTokens: 0,
+      outputTokens: 0
     };
   }
 };
@@ -370,23 +381,26 @@ Human: ${content}`
   if (!jsonStr) {
     return {
       rawResponse: answer,
-      tokens: await countMessagesTokens(messages),
+      inputTokens: await countMessagesTokens(messages),
+      outputTokens: 0,
       arg: {}
     };
   }
 
   try {
     return {
       rawResponse: answer,
-      tokens: await countMessagesTokens(messages),
+      inputTokens: data.usage?.prompt_tokens || 0,
+      outputTokens: data.usage?.completion_tokens || 0,
       arg: json5.parse(jsonStr) as Record<string, any>
     };
   } catch (error) {
     console.log('Extract error, ai answer:', answer);
     console.log(error);
     return {
       rawResponse: answer,
-      tokens: await countMessagesTokens(messages),
+      inputTokens: await countMessagesTokens(messages),
+      outputTokens: 0,
       arg: {}
     };
   }

diff --git a/packages/service/core/workflow/dispatch/chat/oneapi.ts b/packages/service/core/workflow/dispatch/chat/oneapi.ts
@@ -5,7 +5,11 @@ import { ChatRoleEnum } from '@fastgpt/global/core/chat/constants';
 import { SseResponseEventEnum } from '@fastgpt/global/core/workflow/runtime/constants';
 import { textAdaptGptResponse } from '@fastgpt/global/core/workflow/runtime/utils';
 import { createChatCompletion } from '../../../ai/config';
-import type { ChatCompletion, StreamChatType } from '@fastgpt/global/core/ai/type.d';
+import type {
+  ChatCompletion,
+  ChatCompletionMessageParam,
+  StreamChatType
+} from '@fastgpt/global/core/ai/type.d';
 import { formatModelChars2Points } from '../../../../support/wallet/usage/utils';
 import type { LLMModelItemType } from '@fastgpt/global/core/ai/model.d';
 import { postTextCensor } from '../../../../common/api/requestPlusApi';
@@ -214,16 +218,23 @@ export const dispatchChatCompletion = async (props: ChatProps): Promise<ChatResp
     return Promise.reject(getEmptyResponseTip());
   }
 
-  const completeMessages = requestMessages.concat({
-    role: ChatCompletionRequestMessageRoleEnum.Assistant,
-    content: answerText
-  });
+  const AIMessages: ChatCompletionMessageParam[] = [
+    {
+      role: ChatCompletionRequestMessageRoleEnum.Assistant,
+      content: answerText
+    }
+  ];
+
+  const completeMessages = [...requestMessages, ...AIMessages];
   const chatCompleteMessages = GPTMessages2Chats(completeMessages);
 
-  const tokens = await countMessagesTokens(chatCompleteMessages);
+  const inputTokens = await countMessagesTokens(GPTMessages2Chats(requestMessages));
+  const outputTokens = await countMessagesTokens(GPTMessages2Chats(AIMessages));
+
   const { totalPoints, modelName } = formatModelChars2Points({
     model,
-    tokens,
+    inputTokens,
+    outputTokens,
     modelType: ModelTypeEnum.llm
   });
 
@@ -232,7 +243,8 @@ export const dispatchChatCompletion = async (props: ChatProps): Promise<ChatResp
     [DispatchNodeResponseKeyEnum.nodeResponse]: {
       totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
       model: modelName,
-      tokens,
+      inputTokens: inputTokens,
+      outputTokens: outputTokens,
       query: `${userChatInput}`,
       maxToken: max_tokens,
       historyPreview: getHistoryPreview(
@@ -247,7 +259,8 @@ export const dispatchChatCompletion = async (props: ChatProps): Promise<ChatResp
         moduleName: name,
         totalPoints: externalProvider.openaiAccount?.key ? 0 : totalPoints,
         model: modelName,
-        tokens
+        inputTokens: inputTokens,
+        outputTokens: outputTokens
       }
     ],
     [DispatchNodeResponseKeyEnum.toolResponses]: answerText,

diff --git a/packages/service/support/wallet/usage/utils.ts b/packages/service/support/wallet/usage/utils.ts
@@ -1,25 +1,35 @@
-import { ModelTypeEnum, getModelMap } from '../../../core/ai/model';
+import { LLMModelItemType } from '@fastgpt/global/core/ai/model.d';
+import { ModelTypeEnum, getLLMModelPriceType, getModelMap } from '../../../core/ai/model';
 
 export const formatModelChars2Points = ({
   model,
   tokens = 0,
+  inputTokens = 0,
+  outputTokens = 0,
   modelType,
   multiple = 1000
 }: {
   model: string;
-  tokens: number;
+  tokens?: number;
+  inputTokens?: number;
+  outputTokens?: number;
   modelType: `${ModelTypeEnum}`;
   multiple?: number;
 }) => {
-  const modelData = getModelMap?.[modelType]?.(model);
+  const modelData = getModelMap?.[modelType]?.(model) as LLMModelItemType;
   if (!modelData) {
     return {
       totalPoints: 0,
       modelName: ''
     };
   }
 
-  const totalPoints = (modelData.charsPointsPrice || 0) * (tokens / multiple);
+  const isIOType = modelType === ModelTypeEnum.llm && getLLMModelPriceType();
+
+  const totalPoints = isIOType
+    ? ((modelData as LLMModelItemType).inputPrice || 0) * (inputTokens / multiple) +
+      ((modelData as LLMModelItemType).outputPrice || 0) * (outputTokens / multiple)
+    : (modelData.charsPointsPrice || 0) * ((tokens || inputTokens + outputTokens) / multiple);
 
   return {
     modelName: modelData.name,

diff --git a/packages/web/i18n/en/account_usage.json b/packages/web/i18n/en/account_usage.json
@@ -6,12 +6,14 @@
   "details": "Details",
   "duration_seconds": "Duration (seconds)",
   "generation_time": "Generation time",
+  "input_token_length": "input tokens",
   "member": "member",
   "member_name": "Member name",
   "module_name": "module name",
   "month": "moon",
   "no_usage_records": "No usage record yet",
   "order_number": "Order number",
+  "output_token_length": "output tokens",
   "project_name": "Project name",
   "source": "source",
   "text_length": "text length",
@@ -20,4 +22,4 @@
   "total_points_consumed": "AI points consumption",
   "usage_detail": "Usage details",
   "user_type": "type"
-}
+}
diff --git a/packages/web/i18n/en/common.json b/packages/web/i18n/en/common.json
@@ -460,10 +460,12 @@
   "core.chat.response.module historyPreview": "History Preview (Only Partial Content Displayed)",
   "core.chat.response.module http result": "Response Body",
   "core.chat.response.module if else Result": "Condition Result",
+  "core.chat.response.module input tokens": "input tokens",
   "core.chat.response.module limit": "Single Search Limit",
   "core.chat.response.module maxToken": "Max Response Tokens",
   "core.chat.response.module model": "Model",
   "core.chat.response.module name": "Model Name",
+  "core.chat.response.module output tokens": "output tokens",
   "core.chat.response.module query": "Question/Search Term",
   "core.chat.response.module quoteList": "Quote Content",
   "core.chat.response.module similarity": "Similarity",
@@ -1042,6 +1044,8 @@
   "support.user.team.Team Tags Async Success": "Sync Completed",
   "support.user.team.member": "Member",
   "support.wallet.Ai point every thousand tokens": "{{points}} Points/1K Tokens",
+  "support.wallet.Ai point every thousand tokens_input": "Input：{{points}} points/1K tokens",
+  "support.wallet.Ai point every thousand tokens_output": "Output：{{points}} points/1K tokens",
   "support.wallet.Amount": "Amount",
   "support.wallet.Buy": "Buy",
   "support.wallet.Not sufficient": "Insufficient AI Points, Please Upgrade Your Package or Purchase Additional AI Points to Continue Using.",