diff --git a/packages/service/common/file/read/utils.ts b/packages/service/common/file/read/utils.ts
index 00335e9e40b1..a1a336d0d00b 100644
--- a/packages/service/common/file/read/utils.ts
+++ b/packages/service/common/file/read/utils.ts
@@ -6,6 +6,7 @@ import { addHours } from 'date-fns';
 import { WorkerNameEnum, runWorker } from '../../../worker/utils';
 import { ReadFileResponse } from '../../../worker/file/type';
 import {DatasetSchemaType} from "@fastgpt/global/core/dataset/type";
+import {initPdfText} from "../../../worker/file/extension/unstructured";
 
 export const initMarkdownText = ({
   teamId,
@@ -55,6 +56,16 @@ export const readFileRawContent = async ({
     dataset
   });
 
+  // pdf image query
+  if (['pdf'].includes(extension)) {
+    result.rawText = await initPdfText({
+      teamId: teamId,
+      metadata: metadata,
+      dataset: dataset,
+      pageElements: result?.metadata? result?.metadata["elements"] : []
+    });
+  }
+
   // markdown data format
   if (['md', 'html', 'docx'].includes(extension)) {
     result.rawText = await initMarkdownText({
diff --git a/packages/service/core/ai/functions/queryImageDescription.ts b/packages/service/core/ai/functions/queryImageDescription.ts
index 3aa460d2da8d..f280275a120e 100644
--- a/packages/service/core/ai/functions/queryImageDescription.ts
+++ b/packages/service/core/ai/functions/queryImageDescription.ts
@@ -20,11 +20,13 @@ export async function queryImageDescription({
         rawTex,
         image_base64,
         model,
+        ai,
         language="eng"
     }: {
     rawTex: string;
     image_base64: string;
     model: string;
+    ai: any
     language?: string;
 }) {
     // 无"data:image/jpeg;base64,"开头的base64结构
@@ -49,9 +51,6 @@ export async function queryImageDescription({
         }
     ];
 
-    const ai = getAIApi({
-        timeout: 480000
-    });
     const data = await ai.chat.completions.create({
         model: getLLMModel(model).model,
         temperature: 0.1,
diff --git a/packages/service/core/dataset/unstructured/config.ts b/packages/service/core/dataset/unstructured/config.ts
index d3b5e7121599..d9c1ee448be2 100644
--- a/packages/service/core/dataset/unstructured/config.ts
+++ b/packages/service/core/dataset/unstructured/config.ts
@@ -16,29 +16,32 @@ export type UnstructuredEnvType = {
 
 let client: UnstructuredClient | null = null;
 
-function initClient(){
+function initClient(config?: UnstructuredEnvType){
+    if (!config){
+        config = global.unstructuredConfigs;
+    }
     const httpClient = axios.create({
-        timeout: global.unstructuredConfigs.timeout,
+        timeout: config.timeout,
     })
 
     // httpClient.interceptors.request.use((config) => {
     //     return config;
     // })
     client = new UnstructuredClient({
-        serverURL: global.unstructuredConfigs.baseUrl || 'http://localhost:8000',
+        serverURL: config.baseUrl || 'http://localhost:8000',
         security: {
             apiKeyAuth: ""
         },
         defaultClient: httpClient,
         retryConfig: {
             logger: addLog,
-            strategy: global.unstructuredConfigs.retryConfig.strategy,
+            strategy: config.retryConfig.strategy,
             retryConnectionErrors: true,
             backoff: {
-                initialInterval: global.unstructuredConfigs.retryConfig.initialInterval,
-                maxInterval: global.unstructuredConfigs.retryConfig.maxInterval,
-                maxElapsedTime: global.unstructuredConfigs.retryConfig.maxElapsedTime,
-                exponent: global.unstructuredConfigs.retryConfig.exponent,
+                initialInterval: config.retryConfig.initialInterval,
+                maxInterval: config.retryConfig.maxInterval,
+                maxElapsedTime: config.retryConfig.maxElapsedTime,
+                exponent: config.retryConfig.exponent,
             }
         }
     });
@@ -46,11 +49,9 @@ function initClient(){
 
 
 
-export const getClient = ({
-
-}) => {
+export const getClient = (config?: UnstructuredEnvType) => {
     if (!client) {
-        initClient();
+        initClient(config);
     }
     return client;
 }
\ No newline at end of file
diff --git a/packages/service/worker/file/extension/unstructured.ts b/packages/service/worker/file/extension/unstructured.ts
index a076f6309088..89bf37790328 100644
--- a/packages/service/worker/file/extension/unstructured.ts
+++ b/packages/service/worker/file/extension/unstructured.ts
@@ -1,23 +1,14 @@
-import { ReadFileByBufferParams, ReadFileResponse } from "../../../common/file/read/type";
-import { initMarkdownText } from '../../../common/file/read/utils';
-import { getDownloadStream, getFileById } from "../../../common/file/gridfs/controller";
-import { queryImageDescription } from "../../../core/ai/functions/queryImageDescription";
-import { getLLMModel } from "../../../core/ai/model";
-import { getClient } from "../../../core/dataset/unstructured/config";
-import { addLog } from "../../../common/system/log";
-import { PDFDocument } from "pdf-lib"
+import {ReadFileResponse} from "../../../common/file/read/type";
+import {initMarkdownText} from '../../../common/file/read/utils';
+import {queryImageDescription} from "../../../core/ai/functions/queryImageDescription";
+import {getClient} from "../../../core/dataset/unstructured/config";
+import {addLog} from "../../../common/system/log";
+import {PDFDocument} from "pdf-lib"
 import pLimit from "p-limit";
 import {ReadRawTextByBuffer} from "../type";
-
-type TokenType = {
-    str: string;
-    dir: string;
-    width: number;
-    height: number;
-    transform: number[];
-    fontName: string;
-    hasEOL: boolean;
-};
+import {workerData} from "worker_threads"
+import {DatasetSchemaType} from "@fastgpt/global/core/dataset/type";
+import {getAIApi} from "../../../core/ai/config";
 
 type UnstructuredElementType = {
     type: string;
@@ -32,7 +23,7 @@ type UnstructuredElementType = {
 const limit = pLimit(3);
 
 // 解构文件，目前接收pdf、word
-export const readUnFile = async ({ buffer, preview, metadata, teamId, dataset }: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
+export const readUnFile = async ({ buffer, preview, metadata }: ReadRawTextByBuffer): Promise<ReadFileResponse> => {
 
     if (preview) {
         const pdfDoc = await PDFDocument.load(buffer);
@@ -47,7 +38,7 @@ export const readUnFile = async ({ buffer, preview, metadata, teamId, dataset }:
 
     //1. 请求分割pdf
     addLog.info(`File ${metadata?.relatedId} partition started.`);
-    const client = getClient({})
+    const client = getClient(workerData.globalConfig.unstructuredConfigs)
     const res = await client?.general.partition({
         files: {
             content: buffer,
@@ -67,7 +58,23 @@ export const readUnFile = async ({ buffer, preview, metadata, teamId, dataset }:
     if (!pageElements || pageElements.length == 0) {
         pageElements = []
     }
+    if (metadata) {
+        metadata["elements"] = pageElements;
+    }
+    return {
+        formatText: "", metadata: metadata, rawText: ""
+    }
+}
 
+export const initPdfText = async ({ metadata, teamId, dataset, pageElements }: {
+    metadata: any;
+    teamId: string;
+    dataset: DatasetSchemaType|undefined;
+    pageElements: any[];
+}): Promise<string> => {
+    const ai = getAIApi({
+        timeout: 480000
+    })
     //3. 请求llm-v对图片（图片和表格）进行描述 4. 将图片、表格插入mongodb
     const asyncOperation = async (element: UnstructuredElementType) => {
         if (["Image", "Table"].includes(element.type) && element.text.length >= 2 && element.metadata.image_base64) {
@@ -77,6 +84,7 @@ export const readUnFile = async ({ buffer, preview, metadata, teamId, dataset }:
                     rawTex: element.text,
                     image_base64: "data:image/jpeg;base64," + element.metadata.image_base64,
                     model: (dataset?.agentModel || "gemini-pro-vision"),
+                    ai: ai,
                     language: element.metadata.languages[0],
                 }).catch(error => {
                     addLog.error(`Llm image ${element.element_id} error:`, error)
@@ -105,7 +113,5 @@ export const readUnFile = async ({ buffer, preview, metadata, teamId, dataset }:
     }).join('');
     addLog.info(`Join ${metadata?.relatedId} pdf text end.`);
 
-    return {
-        formatText: "", metadata: metadata, rawText: finalText
-    }
+    return finalText
 }
\ No newline at end of file
diff --git a/packages/service/worker/file/type.d.ts b/packages/service/worker/file/type.d.ts
index 1992e3395af4..2b0563de862a 100644
--- a/packages/service/worker/file/type.d.ts
+++ b/packages/service/worker/file/type.d.ts
@@ -8,7 +8,6 @@ export type ReadRawTextProps<T> = {
   encoding: string;
   preview?: boolean;
   teamId: string;
-  dataset?: DatasetSchemaType;
   metadata?: Record<string, any>;
 };
 
@@ -17,4 +16,5 @@ export type ReadRawTextByBuffer = ReadRawTextProps<Buffer>;
 export type ReadFileResponse = {
   rawText: string;
   formatText?: string;
+  metadata?: Record<string, any>;
 };
diff --git a/packages/service/worker/utils.ts b/packages/service/worker/utils.ts
index f9ab4be72f9b..8390ccd2b0e7 100644
--- a/packages/service/worker/utils.ts
+++ b/packages/service/worker/utils.ts
@@ -9,7 +9,11 @@ export enum WorkerNameEnum {
 
 export const getWorker = (name: WorkerNameEnum) => {
   const workerPath = path.join(process.cwd(), '.next', 'server', 'worker', `${name}.js`);
-  return new Worker(workerPath);
+  return new Worker(workerPath, {workerData: {
+    globalConfig: {
+      unstructuredConfigs: global.unstructuredConfigs
+    }
+    }});
 };
 
 export const runWorker = <T = any>(name: WorkerNameEnum, params?: Record<string, any>) => {
diff --git a/projects/app/src/global/common/api/systemRes.d.ts b/projects/app/src/global/common/api/systemRes.d.ts
index 8ddf3369766c..f224615cc868 100644
--- a/projects/app/src/global/common/api/systemRes.d.ts
+++ b/projects/app/src/global/common/api/systemRes.d.ts
@@ -8,6 +8,7 @@ import type {
 
 import type { FastGPTFeConfigsType } from '@fastgpt/global/common/system/types/index.d';
 import { SubPlanType } from '@fastgpt/global/support/wallet/sub/type';
+import {UnstructuredEnvType} from "@fastgpt/service/core/dataset/unstructured/config";
 
 export type InitDateResponse = {
   llmModels: LLMModelItemType[];
@@ -15,6 +16,7 @@ export type InitDateResponse = {
   audioSpeechModels: AudioSpeechModels[];
   reRankModels: ReRankModelItemType[];
   whisperModel: WhisperModelType;
+  unstructuredConfigs: UnstructuredEnvType;
   feConfigs: FastGPTFeConfigsType;
   subPlans?: SubPlanType;
   systemVersion: string;
diff --git a/projects/app/src/pages/api/common/system/getInitData.ts b/projects/app/src/pages/api/common/system/getInitData.ts
index 9bd032964b7e..6b0e45f93fd6 100644
--- a/projects/app/src/pages/api/common/system/getInitData.ts
+++ b/projects/app/src/pages/api/common/system/getInitData.ts
@@ -19,6 +19,7 @@ export default async function handler(req: NextApiRequest, res: NextApiResponse)
   jsonRes<InitDateResponse>(res, {
     data: {
       feConfigs: global.feConfigs,
+      unstructuredConfigs: global.unstructuredConfigs,
       subPlans: global.subPlans,
       llmModels: global.llmModels,
       vectorModels: global.vectorModels,
diff --git a/projects/app/src/service/events/generateFileChunk.ts b/projects/app/src/service/events/generateFileChunk.ts
index dd8528af00cc..2e0455a3438f 100644
--- a/projects/app/src/service/events/generateFileChunk.ts
+++ b/projects/app/src/service/events/generateFileChunk.ts
@@ -1,7 +1,7 @@
 import {readFileContentFromMongo} from "@fastgpt/service/common/file/gridfs/controller";
 import {BucketNameEnum} from "@fastgpt/global/common/file/constants";
 import {splitText2Chunks} from "@fastgpt/global/common/string/textSplitter";
-import {DatasetCollectionTypeEnum, TrainingModeEnum} from "@fastgpt/global/core/dataset/constants";
+import {TrainingModeEnum} from "@fastgpt/global/core/dataset/constants";
 import {checkDatasetLimit} from "@fastgpt/service/support/permission/teamLimit";
 import {predictDataLimitLength} from "@fastgpt/global/core/dataset/utils";
 import {createTrainingUsage} from "@fastgpt/service/support/wallet/usage/controller";
@@ -9,17 +9,12 @@ import {UsageSourceEnum} from "@fastgpt/global/support/wallet/usage/constants";
 import {getLLMModel, getVectorModel} from "@fastgpt/service/core/ai/model";
 import {pushDataListToTrainingQueue} from "@fastgpt/service/core/dataset/training/controller";
 import {MongoImage} from "@fastgpt/service/common/file/image/schema";
-import {jsonRes} from "@fastgpt/service/common/response";
 import {addLog} from "@fastgpt/service/common/system/log";
 import {startTrainingQueue} from "@/service/core/dataset/training/utils";
 import {DatasetSchemaType} from "@fastgpt/global/core/dataset/type";
-import {ClientSession} from "@fastgpt/service/common/mongo";
 import {mongoSessionRun} from "@fastgpt/service/common/mongo/sessionRun";
-import {createOneCollection} from "@fastgpt/service/core/dataset/collection/controller";
-import {putDatasetCollectionById} from "@/web/core/dataset/api";
 import {hashStr} from "@fastgpt/global/common/string/tools";
 import {MongoDatasetCollection} from "@fastgpt/service/core/dataset/collection/schema";
-import {getCollectionUpdateTime} from "@fastgpt/service/core/dataset/collection/utils";
 
 export const generateFileChunk = async ({
                                             teamId,