> ## Documentation Index > Fetch the complete documentation index at: https://docs.dumplingai.com/llms.txt > Use this file to discover all available pages before exploring further. # Extract Document > Extract structured text and metadata from uploaded documents or document URLs. ## Description This endpoint extracts structured data from document files based on a user-defined prompt. It supports input via URL or base64-encoded file content and uses vision-capable Large Language Models (LLMs) to interpret and extract relevant information from the PDFs. The following file extensions are supported: `.123` `.602` `.abw` `.bib` `.bmp` `.cdr` `.cgm` `.cmx` `.csv` `.cwk` `.dbf` `.dif` `.doc` `.docm` `.docx` `.dot` `.dotm` `.dotx` `.dxf` `.emf` `.eps` `.epub` `.fodg` `.fodp` `.fods` `.fodt` `.fopd` `.gif` `.htm` `.html` `.hwp` `.jpeg` `.jpg` `.key` `.ltx` `.lwp` `.mcw` `.met` `.mml` `.mw` `.numbers` `.odd` `.odg` `.odm` `.odp` `.ods` `.odt` `.otg` `.oth` `.otp` `.ots` `.ott` `.pages` `.pbm` `.pcd` `.pct` `.pcx` `.pdb` `.pdf` `.pgm` `.png` `.pot` `.potm` `.potx` `.ppm` `.pps` `.ppt` `.pptm` `.pptx` `.psd` `.psw` `.pub` `.pwp` `.pxl` `.ras` `.rtf` `.sda` `.sdc` `.sdd` `.sdp` `.sdw` `.sgl` `.slk` `.smf` `.stc` `.std` `.sti` `.stw` `.svg` `.svm` `.swf` `.sxc` `.sxd` `.sxg` `.sxi` `.sxm` `.sxw` `.tga` `.tif` `.tiff` `.txt` `.uof` `.uop` `.uos` `.uot` `.vdx` `.vor` `.vsd` `.vsdm` `.vsdx` `.webp` `.wb2` `.wk1` `.wks` `.wmf` `.wpd` `.wpg` `.wps` `.xbm` `.xhtml` `.xls` `.xlsb` `.xlsm` `.xlsx` `.xlt` `.xltm` `.xltx` `.xlw` `.xml` `.xpm` `.zabw` ## Endpoint ``` POST https://app.dumplingai.com/api/v1/extract-document ``` ## Headers * **Content-Type:** `application/json` * **Authorization:** Bearer `` (required) ## Request Body ```json theme={null} { "inputMethod": "string", // Required. Either "url" or "base64". "files": ["string"], // Required. Array of URLs or base64-encoded file contents. "fileExtension": "string", // Optional. The file extension of the input files (e.g., ".pdf", ".docx"). Default: ".pdf". "prompt": "string", // Required. The prompt describing the data to extract. "jsonMode": boolean // Optional. Whether to return the result in JSON format. Default: false. } ``` ## Responses ### Success (200) Returns the extracted data based on the provided prompt, along with additional information. ```json theme={null} { "results": "string", // Extracted data based on the prompt "prompt": "string", // The original prompt used for extraction "pages": number, // Total number of pages processed "fileCount": number, // Number of files processed "creditUsage": number // Total credits used for this request } ``` * **Content-Type:** application/json * **X-RateLimit-Limit:** The rate limit for the user. * **X-RateLimit-Remaining:** The remaining number of requests for the user. ### Bad Request (400) Returned if the request is invalid or the total file size exceeds the limit. ```json theme={null} { "error": "Error message describing the issue" } ``` ### Unauthorized (401) Returned if the API key is invalid or missing. ```json theme={null} { "error": "Invalid or missing Authorization header" } ``` ### Internal Server Error (500) Returned if there's an error during the document extraction process. ```json theme={null} { "error": "Failed to extract document: [error details]" } ``` ## Example Request ### Example with PDF (default) ```bash theme={null} curl -X POST https://app.dumplingai.com/api/v1/extract-document \ -H "Content-Type: application/json" \ -H "Authorization: Bearer YOUR_API_KEY" \ -d '{ "inputMethod": "url", "files": ["https://example.com/sample.pdf"], "prompt": "Extract the main topics and their descriptions from this PDF.", "jsonMode": false }' ``` ### Example with Word document ```bash theme={null} curl -X POST https://app.dumplingai.com/api/v1/extract-document \ -H "Content-Type: application/json" \ -H "Authorization: Bearer YOUR_API_KEY" \ -d '{ "inputMethod": "url", "files": ["https://example.com/report.docx"], "fileExtension": ".docx", "prompt": "Extract all tables and their data from this Word document.", "jsonMode": true }' ``` ## Notes * The maximum total file size for all documents combined is 100MB. * The maximum file size for a single document is 30MB. * The maximum number of pages that can be processed in a single request is \~3000 pages. * fileExtension can be set to "autodetect" if not sure. * The maximum output is 8,192 tokens. * Credit usage: * Base cost: 100 credits * Additional 10 credits per page processed * The total credit usage is returned in the response as `creditUsage`. * If using the URL method, ensure the file is publicly accessible. * The `jsonMode` parameter determines whether the output is formatted as JSON (true) or plain text (false). * We apply vision-based LLMs to all pages of the document for extraction. * The endpoint can process multiple document files in a single request. * Temporary files are created during processing and are deleted after use. * You can get a list of supported file extensions by calling: ``` GET /api/v1/extract-document ``` ## Rate Limiting Rate limit headers (`X-RateLimit-Limit` and `X-RateLimit-Remaining`) are included in the response to indicate the user's current rate limit status. ## Error Handling * If the required parameters (`files` or `prompt`) are missing, a 400 Bad Request error is returned. * If the total file size exceeds 100MB, a 400 Bad Request error is returned. * If there's an error during extraction, a 500 Internal Server Error is returned with details about the failure. ## Security and Privacy * Uploaded files are temporarily stored and then deleted after processing. ## OpenAPI ````yaml POST /api/v1/extract-document openapi: 3.0.3 info: title: DumplingAI API version: 1.0.0 description: > REST API for DumplingAI's content intelligence and automation platform. All endpoints are grouped under `/api/v1`; most are secured via Bearer API keys unless an operation explicitly sets `security: []`. servers: - url: https://app.dumplingai.com description: Production security: - bearerAuth: [] tags: - name: YouTube description: Access metadata, search results, and transcripts from YouTube. - name: TikTok description: Retrieve TikTok profile, video, follower, and transcript data. - name: LinkedIn description: Programmatically fetch LinkedIn company and profile data. - name: Search description: Search-orientated endpoints spanning web, news, maps, and autocomplete. - name: Google description: Integrations with Google business listings and location data. - name: Scraping description: Webpage capture, crawling, and structured content extraction utilities. - name: Documents description: Document processing, conversion, and metadata utilities. - name: AI description: DumplingAI agent and knowledge base endpoints. - name: Developer Tools description: Utilities for executing sandboxed code via API. paths: /api/v1/extract-document: post: tags: - Documents summary: Extract document to structured JSON description: >- Extract structured text and metadata from uploaded documents or document URLs. operationId: extractDocument requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/ExtractDocumentRequest' examples: default: value: inputMethod: url files: - https://example.com/sample.pdf prompt: >- Extract the main topics and their summaries from this document. jsonMode: false responses: '200': description: Document extraction returned. content: application/json: schema: $ref: '#/components/schemas/ExtractDocumentResponse' '400': description: Invalid request payload. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' '401': description: Missing or invalid API key. '500': description: Unexpected server error. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' components: schemas: ExtractDocumentRequest: type: object required: - inputMethod - files - prompt properties: inputMethod: $ref: '#/components/schemas/FileInputMethod' files: type: array minItems: 1 description: >- Array of publicly accessible URLs or base64-encoded document contents. items: type: string fileExtension: type: string description: >- File extension for the provided documents (e.g., '.pdf', '.docx', or 'autodetect'). Defaults to '.pdf'. prompt: type: string description: >- Instructions that describe the structured data to extract from the documents. jsonMode: type: boolean description: When true, requests the model to respond with JSON-formatted output. default: false requestSource: $ref: '#/components/schemas/RequestSource' additionalProperties: false ExtractDocumentResponse: type: object required: - results - prompt - pages - fileCount - creditUsage properties: results: type: string description: Model output returned from the extraction prompt. prompt: type: string description: Prompt that was sent to the extraction model. pages: type: integer description: Total number of document pages processed. fileCount: type: integer description: Number of documents included in the request. creditUsage: type: integer description: Credits consumed while processing the request. additionalProperties: false ErrorResponse: type: object properties: error: type: string description: Human-readable description of what went wrong. required: - error FileInputMethod: type: string description: >- Indicates whether binary content is supplied via URL or base64-encoded string. enum: - url - base64 RequestSource: type: string description: Optional identifier describing where the API request originated. enum: - API - WEB - MAKE_DOT_COM - ZAPIER - N8N - PLAYGROUND - DEFAULT_AUTOMATION - AGENT_PREVIEW - AGENT_LIVE - AUTOPILOT - STUDIO securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: API Key ````