Document parsing (large model)

Updated at:

The Document Analysis (Large Model Edition) API provides general-purpose document extraction and understanding. It supports various common document formats, extracts rich layout information, and outputs content in Markdown format. This document explains how to call the API. Before you start, please read the API user guide.

Invocation

The Document Parser (Large Model Version) is an asynchronous interface. To use this service, submit an asynchronous task by calling SubmitDocParserJobAdvance or SubmitDocParserJob. Then, query the processing status by calling QueryDocParserStatus. Finally, call GetDocParserResult to retrieve the result. The following workflow diagram illustrates the process:

image
Note
  • Document Parsing (Large Model Edition) includes a free monthly quota of 3,000 pages. If your free quota or resource package is exhausted, the service defaults to Pay-As-You-Go. Once an asynchronous task completes, you can query its results for up to 24 hours. After this period, the results are no longer available.

  • Supported formats include documents (PDF, Word, PPT, PPTX, XLS, XLSX, and XLSM), images (JPG, JPEG, PNG, BMP, and GIF), and other file types (Markdown, HTML, EPUB, MOBI, RTF, and TXT).

  • Supported audio and video formats include MP4, MKV, AVI, MOV, WMV, MP3, WAV, and AAC.

  • Document Parsing (Large Model Edition) can parse images and tables in documents. The chart2table component processes images containing charts, converting the chart data into a table.

  • For table parsing, Document Parsing (Large Model Edition) is the most accurate, followed by Intelligent Document Parsing and Electronic Document Parsing. Therefore, we recommend Document Parsing (Large Model Edition) for document analysis. In terms of overall processing speed, Electronic Document Parsing is the fastest, followed by Document Parsing (Large Model Edition) and Intelligent Document Parsing.

Step 1: Submit an asynchronous parsing job

The asynchronous submission service supports two submission methods: by local file or by file URL.

  • To submit a local file, call the SubmitDocParserJobAdvance API.

  • To submit a file by URL, call the SubmitDocParserJob API.

Important
  • Asynchronous processing times can vary. After activating the Electronic Document Parsing service, you will receive a free quota for testing.

  • The service selects a parser based on the file extension in FileName or FileNameExtension. If you are unsure of the document type, you can omit the extension, and the service will route the file to a default parser to prevent unexpected results.

Request parameters

Parameter

Type

Required

Description

Example

FileUrl

string

No

The URL of the document to submit.

Supports single documents, such as PDF and Word, up to 15,000 pages and 150 MB, or single images up to 20 MB.

https://example.com/example.pdf

FileUrlObject

stream

No

The stream of the local file to submit.

Supports single documents, such as PDF and Word, up to 15,000 pages and 150 MB, or single images up to 20 MB.

A FileInputStream generated from a local file.

FileName

string

No

The filename, which must include the file extension. You must specify either FileName or FileNameExtension.

example.pdf

FileNameExtension

string

No

The file extension. You must specify either FileName or FileNameExtension.

pdf

FormulaEnhancement

bool

No

Specifies whether to enable enhanced formula recognition. The default value is false.

True

LlmEnhancement

bool

No

Specifies whether to enable enhancement with a large language model. The default value is false. Important: Enabling this option may significantly increase the API response time. Use it only when necessary. When enabled, the service uses the large language model to process the following content:

  • PPT files (including PDF files converted from PPT) when the layout confidence (layoutConf) is 0.8 or lower.

  • Charts within the document. Currently, three chart types are supported: value charts, valuation charts, and other charts.

False

EnhancementMode

string

No

The large model enhancement mode is empty (disabled) by default. This parameter takes effect only when LlmEnhancement=true is set.

The following option is supported:

  • VLM: Uses a multimodal large model for layout analysis and optical character recognition (OCR). This can improve parsing accuracy, including more precise layout order, type, and coordinates; more accurate text content; and better multilingual support.

Important: Enabling this option may significantly increase the API response time and will incur additional charges per page. Use it only when necessary.

VLM

Option

string

No

Specifies the processing option for audio and video files. The default value is Base.

base: Performs basic recognition. (Default)

advance: Performs enhanced recognition and returns plot analysis results in addition to the basic results.

Base

OssBucket

string

No

The name of your Object Storage Service (OSS) bucket. For more information, see OSS hosting support.

docmind-trust

OssEndpoint

string

No

The endpoint of your OSS bucket. For more information, see OSS hosting support.

oss-cn-hangzhou.aliyuncs.com

PageIndex

string

No

The page range to parse. Note: The format must be 1-n.

1-5

OutputHtmlTable

bool

No

Specifies whether to return table content in HTML format. The default value is False. If this option is enabled, for a layout block of the table type, the parsed HTML content of the table is returned and stored in the llmResult field of the block.

⚠️ Important: If you enable this parameter, you must also set the LlmEnhancement parameter to true. This may increase the API response time. Use it only when necessary.

false

EnableEventCallback

bool

No

Specifies whether to enable event callbacks. The default value is false. For more information, see EventBridge support.

false

OutputFormat

List<String>

No

The output format for the parsing results.

markdown

MultimediaParameters

object

No

For audio or video files, use this field to configure optional features. If set, this parameter overrides the Option parameter.

  • VlParsePrompt

string

No

The prompt for frame parsing. (Supported since July 25).

##Constraints:- Do not output long sentences. Adhere to the original image and capture only key information.

Note
  • The PageIndex parameter is supported only for PDF, Word, and PPT files.

  • The OutputFormat parameter supports the following values:

    • markdown: The response from the status query service includes the document content in Markdown format.

    • visualLayoutInfo: The QueryDocParserStatus response includes page image information, and the GetDocParserResult response includes layout coordinate information.

  • When VLM enhancement is enabled, 33 languages are supported:

    • Chinese, Japanese, Korean, Indonesian, Vietnamese, Thai, English, French, German, Russian, Portuguese, Spanish, Italian, Swedish, Danish, Czech, Norwegian, Dutch, Finnish, Turkish, Polish, Swahili, Romanian, Serbian, Greek, Kazakh, Uzbek, Cebuano, Arabic, Urdu, Persian, Hindi/Devanagari, and Hebrew.

Response parameters

Parameter

Type

Description

Example

RequestId

string

The unique ID for the request.

43A29C77-405E-4DC0-BC55-EE694AD0****

Data

object

The returned data.

{"Id": "docmind-20240712-b15f****"}

Id

string

The job ID. Use this ID to query the job status.

doc-mind-20220712-b15f****

Code

string

The status code.

200

Message

string

Detailed status information.

Message

Examples

You can call this API in two ways: by uploading a local file or by providing a file URL.

  • Upload a local file: The following Java SDK example shows how to upload a local file by calling the SubmitDocParserJobAdvance API with the fileUrlObject parameter.

    Note

    For information about how to obtain and use an AccessKey, see the SDK guide for your preferred language in SDK Overview.

    import com.aliyun.docmind_api20220711.models.*;
    import com.aliyun.teaopenapi.models.Config;
    import com.aliyun.docmind_api20220711.Client;
    import com.aliyun.teautil.models.RuntimeOptions;
    import java.io.File;
    import java.io.FileInputStream;
    
    public static void main(String[] args) throws Exception {
            submit();
        }
    public static void submit() throws Exception {
        // Initialize the Credentials client by using the default credential provider.
        com.aliyun.credentials.Client credentialClient = new com.aliyun.credentials.Client();
        Config config = new Config()
            // Obtain the AccessKey ID from the credentials.
            .setAccessKeyId(credentialClient.getAccessKeyId())
            // Obtain the AccessKey secret from the credentials.
            .setAccessKeySecret(credentialClient.getAccessKeySecret());
        // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
        config.endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
        Client client = new Client(config);
        // Create a RuntimeOptions instance and configure runtime parameters.
        RuntimeOptions runtime = new RuntimeOptions();
        SubmitDocParserJobAdvanceRequest advanceRequest = new SubmitDocParserJobAdvanceRequest();
        File file = new File("D:\\example.pdf");
        advanceRequest.fileUrlObject = new FileInputStream(file);
        advanceRequest.fileName = "example.pdf";
        // Send the request and handle the response or exceptions.
        SubmitDocParserJobResponse response = client.submitDocParserJobAdvance(advanceRequest, runtime);
        System.out.println(com.alibaba.fastjson.JSON.toJSON(response.getBody()));
    
    }
    const Client = require('@alicloud/docmind-api20220711');
    const Credential = require('@alicloud/credentials');
    const Util = require('@alicloud/tea-util');
    const fs = require('fs');
    
    const getResult = async () => {
    	// Initialize the Credentials client by using the default credential provider.
      const cred = new Credential.default();
      const client = new Client.default({
        // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
        endpoint: 'docmind-api.cn-hangzhou.aliyuncs.com',
        // Obtain the AccessKey ID from the credentials.
        accessKeyId: cred.credential.accessKeyId,
        // Obtain the AccessKey secret from the credentials.
        accessKeySecret: cred.credential.accessKeySecret,
        type: 'access_key',
        regionId: 'cn-hangzhou'
      }); 
      const advanceRequest = new Client.SubmitDocParserJobAdvanceRequest();
      const file = fs.createReadStream('./example.pdf');
      advanceRequest.fileUrlObject = file;
      advanceRequest.fileName = 'example.pdf';
      const runtimeObject = new Util.RuntimeOptions({});
      const response = await client.submitDocParserJobAdvance(advanceRequest, runtimeObject);
      return response.body;
    };
    
    from alibabacloud_docmind_api20220711.client import Client as docmind_api20220711Client
    from alibabacloud_tea_openapi import models as open_api_models
    from alibabacloud_docmind_api20220711 import models as docmind_api20220711_models
    from alibabacloud_tea_util.client import Client as UtilClient
    from alibabacloud_credentials.client import Client as CredClient
    
    if __name__ == '__main__':
      	# Initialize the Credentials client by using the default credential provider.
        cred=CredClient()
        config = open_api_models.Config(
            # Obtain the AccessKey ID from the credentials.
            access_key_id=cred.get_credential().get_access_key_id(),
            # Obtain the AccessKey secret from the credentials.
            access_key_secret=cred.get_credential().get_access_key_secret()
        )
        # The endpoint of the service.
        config.endpoint = f'docmind-api.cn-hangzhou.aliyuncs.com'
        client = docmind_api20220711Client(config)
        request = docmind_api20220711_models.SubmitDocParserJobAdvanceRequest(
            # file_url_object: The local file stream.
            file_url_object=open("./example.pdf", "rb"),
            # file_name: The name of the file. The name must include the file extension.
            file_name='123.pdf',
            # file_name_extension: The file extension. Specify either FileName or FileNameExtension.
            file_name_extension='pdf'
        )
        runtime = util_models.RuntimeOptions()
        try:
            # Process the API response as needed.
            response = client.submit_doc_parser_job_advance(request, runtime)
            # The API response is structured as body -> data -> specific properties. You can print the results as needed. The following example shows how to print the returned job ID.
            # Property names of the returned values start with a lowercase letter.
            print(response.body)
        except Exception as error:
            # Print the error message if needed.
            UtilClient.assert_as_string(error.message)
    
    import (
    	"fmt"
    	"os"
      
    	openClient "github.com/alibabacloud-go/darabonba-openapi/v2/client"
    	"github.com/alibabacloud-go/docmind-api-20220711/client"
    	"github.com/alibabacloud-go/tea-utils/v2/service"
      "github.com/aliyun/credentials-go/credentials"
    )
    
    func submit(){
     // Initialize the Credentials client by using the default credential provider.
    	credential, err := credentials.NewCredential(nil)
    	// Obtain the AccessKey ID from the credentials.
    	accessKeyId, err := credential.GetAccessKeyId()
    	// Obtain the AccessKey secret from the credentials.
    	accessKeySecret, err := credential.GetAccessKeySecret()
      // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
      var endpoint string = "docmind-api.cn-hangzhou.aliyuncs.com"
    	config := openClient.Config{AccessKeyId: accessKeyId, AccessKeySecret: accessKeySecret, Endpoint: &endpoint}
    	// Initialize the client.
      cli, err := client.NewClient(&config)
    	if err != nil {
    		panic(err)
    	}
      // Call the API to upload a local document.
      filename := "D:\\example.pdf"    
      f, err := os.Open(filename)
    	if err != nil {
        panic(err)
    	}
      // Initialize the API request.
      request := client.SubmitDocParserJobAdvanceRequest{
    		FileName:      &filename,
    		FileUrlObject: f,
    	}
      // Create a RuntimeOptions instance and configure runtime parameters.
      options := service.RuntimeOptions{}
      response, err := cli.SubmitDocParserJobAdvance(&request, &options)
      if err != nil {
    		panic(err)
    	}
      // Print the result.
    	fmt.Println(response.Body.String())
    }
  • Upload a file by URL: The following code provides an example of how to submit a file by URL by using the Java SDK. Call the SubmitDocParserJob API and provide the document URL in the fileUrl parameter. The URL must be publicly accessible, free of cross-origin restrictions, and not contain any special escaped characters.

    Note

    For information about how to obtain and use an AccessKey, see the SDK guide for your preferred language in SDK Overview.

    import com.aliyun.docmind_api20220711.models.*;
    import com.aliyun.teaopenapi.models.Config;
    import com.aliyun.docmind_api20220711.Client;
    
    public static void main(String[] args) throws Exception {
            submit();
        }
    public static void submit() throws Exception {
        // Initialize the Credentials client by using the default credential provider.
        com.aliyun.credentials.Client credentialClient = new com.aliyun.credentials.Client();
        Config config = new Config()
            // Obtain the AccessKey ID from the credentials.
            .setAccessKeyId(credentialClient.getAccessKeyId())
            // Obtain the AccessKey secret from the credentials.
            .setAccessKeySecret(credentialClient.getAccessKeySecret());
        // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
        config.endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
        Client client = new Client(config);
        SubmitDocParserJobRequest request = new SubmitDocParserJobRequest();
        request.fileName = "example.pdf";
        request.fileUrl = "https://example.com/example.pdf";
        SubmitDocParserJobResponse response = client.submitDocParserJob(request);
        System.out.println(com.alibaba.fastjson.JSON.toJSON(response.getBody()));
    
    }
    const Client = require('@alicloud/docmind-api20220711');
    const Credential = require('@alicloud/credentials');
    
    const getResult = async () => {
    	// Initialize the Credentials client by using the default credential provider.
      const cred = new Credential.default();
      const client = new Client.default({
        // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
        endpoint: 'docmind-api.cn-hangzhou.aliyuncs.com',
        // Obtain the AccessKey ID from the credentials.
        accessKeyId: cred.credential.accessKeyId,
        // Obtain the AccessKey secret from the credentials.
        accessKeySecret: cred.credential.accessKeySecret,
        type: 'access_key',
        regionId: 'cn-hangzhou'
      });
      
      const request = new Client.SubmitDocParserJobRequest();
      request.fileName = 'example.pdf';
      request.fileUrl = 'https://example.com/example.pdf';
      const response = await client.submitDocParserJob(request);
      
      return response.body;
    }
    from alibabacloud_docmind_api20220711.client import Client as docmind_api20220711Client
    from alibabacloud_tea_openapi import models as open_api_models
    from alibabacloud_docmind_api20220711 import models as docmind_api20220711_models
    from alibabacloud_tea_util.client import Client as UtilClient
    from alibabacloud_credentials.client import Client as CredClient
    
    if __name__ == '__main__':
      	# Initialize the Credentials client by using the default credential provider.
        cred=CredClient()
        config = open_api_models.Config(
            # Obtain the AccessKey ID from the credentials.
            access_key_id=cred.get_credential().get_access_key_id(),
            # Obtain the AccessKey secret from the credentials.
            access_key_secret=cred.get_credential().get_access_key_secret()
        )
        # The endpoint of the service.
        config.endpoint = f'docmind-api.cn-hangzhou.aliyuncs.com'
        client = docmind_api20220711Client(config)
        request = docmind_api20220711_models.SubmitDocParserJobRequest(
            # file_url: The URL of the file.
            file_url='https://example.com/example.pdf',
            # file_name: The name of the file. The name must include the file extension.
            file_name='123.pdf',
            # file_name_extension: The file extension. Specify either FileName or FileNameExtension.
            file_name_extension='pdf'
        )
        try:
            # Process the API response as needed.
            response = client.submit_doc_parser_job(request)
            # The API response is structured as body -> data -> specific properties. You can print the results as needed. The following example shows how to print the returned job ID.
            # Property names of the returned values start with a lowercase letter.
            print(response.body)     
        except Exception as error:
            # Print the error message if needed.
            UtilClient.assert_as_string(error.message)
    
    import (
    	"fmt"
    
    	openClient "github.com/alibabacloud-go/darabonba-openapi/v2/client"
      "github.com/alibabacloud-go/docmind-api-20220711/client"
      "github.com/aliyun/credentials-go/credentials"
    )
    
    func submit(){
      // Initialize the Credentials client by using the default credential provider.
    	credential, err := credentials.NewCredential(nil)
    	// Obtain the AccessKey ID from the credentials.
    	accessKeyId, err := credential.GetAccessKeyId()
    	// Obtain the AccessKey secret from the credentials.
    	accessKeySecret, err := credential.GetAccessKeySecret()
      // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
      var endpoint string = "docmind-api.cn-hangzhou.aliyuncs.com"
    	config := openClient.Config{AccessKeyId: accessKeyId, AccessKeySecret: accessKeySecret, Endpoint: &endpoint}
    	// Initialize the client.
      cli, err := client.NewClient(&config)
    	if err != nil {
    		panic(err)
    	}
      // The file URL.
      fileURL := "https://example.com/example.pdf"
      // The filename.
      fileName := "example.pdf"
      // Initialize the API request.
      request := client.SubmitDocParserJobRequest{
    		FileUrl:  &fileURL,
    		FileName: &fileName,
    	}
      response, err := cli.SubmitDocParserJob(&request)
      if err != nil {
    		panic(err)
    	}
      // Print the result.
    	fmt.Println(response.Body.String())
    }
    using Newtonsoft.Json;
    using System;
    using System.Collections;
    using System.Collections.Generic;
    using System.IO;
    using System.Threading.Tasks;
    
    using Tea;
    using Tea.Utils;
    
    public static void SubmitUrl()
            {
                // Initialize the Credentials client by using the default credential provider.
              	var akCredential = new Aliyun.Credentials.Client(null);
                AlibabaCloud.OpenApiClient.Models.Config config = new AlibabaCloud.OpenApiClient.Models.Config
                {
                    // Obtain the AccessKey ID from the credentials.
                    AccessKeyId = akCredential.GetAccessKeyId(),
                    // Obtain the AccessKey secret from the credentials.
                    AccessKeySecret = akCredential.GetAccessKeySecret(),
                };
                // The endpoint of the service.
                config.Endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
                AlibabaCloud.SDK.Docmind_api20220711.Client client = new AlibabaCloud.SDK.Docmind_api20220711.Client(config);
                AlibabaCloud.SDK.Docmind_api20220711.Models.SubmitDocParserJobRequest request = new AlibabaCloud.SDK.Docmind_api20220711.Models.SubmitDocParserJobRequest
                {
                    FileUrl = "https://example.pdf",
                    FileNameExtension = "pdf"
                };
                try
                {
                    // Process the API response as needed.
                    client.SubmitDocParserJob(request);
                }
                catch (TeaException error)
                {
                    // Print the error message if needed.
                    AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
                }
                catch (Exception _error)
                {
                    TeaException error = new TeaException(new Dictionary<string, object>
                    {
                        { "message", _error.Message }
                    });
                    // Print the error message if needed.
                    AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
                }
            }
    use AlibabaCloud\SDK\Docmindapi\V20220711\Docmindapi;
    use AlibabaCloud\SDK\Docmindapi\V20220711\Models\SubmitDocStructureJobRequest;
    use Darabonba\OpenApi\Models\Config;
    use AlibabaCloud\Tea\Utils\Utils\RuntimeOptions;
    use AlibabaCloud\Tea\Exception\TeaUnableRetryError;
    use AlibabaCloud\Credentials\Credential;
    
    // Initialize the Credentials client by using the default credential provider.
    $bearerToken = new Credential();    
    $config = new Config();
    // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for IPv6.
    $config->endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
    // Obtain the AccessKey ID from the credentials.
    $config->accessKeyId = $bearerToken->getCredential()->getAccessKeyId();
    // Obtain the AccessKey secret from the credentials.
    $config->accessKeySecret = $bearerToken->getCredential()->getAccessKeySecret();
    $config->type = "access_key";
    $config->regionId = "cn-hangzhou";
    $client = new Docmindapi($config);
    $request = new SubmitDocParserJobRequest();
    
    $runtime = new RuntimeOptions();
    $runtime->maxIdleConns = 3;
    $runtime->connectTimeout = 10000;
    $runtime->readTimeout = 10000;
    
    $request->fileName = "example.pdf";
    $request->fileUrl = "https://example.com/example.pdf";
    
    try {
      $response = $client->submitDocParserJob($request, $runtime);
      var_dump($response->toMap());
    } catch (TeaUnableRetryError $e) {
      var_dump($e->getMessage());
      var_dump($e->getErrorInfo());
      var_dump($e->getLastException());
      var_dump($e->getLastRequest());
    }

Example response

{
  "RequestId": "43A29C77-405E-4DC0-BC55-EE694AD0****",
  "Data": {
    "Id": "docmind-20240712-b15f****"
  }  
}

Step 2: Call the QueryDocParserStatus API

To call the query API, use the Id returned by the asynchronous task submission API. The query result includes the Status of the job and NumberOfSuccessfulParsing, which is the number of successfully parsed modules. The Status can be Processing, success, or Fail.

Request parameters

Parameter

Type

Required

Description

Example

Id

string

Yes

The ID of the job to query. This ID is returned in the response of the asynchronous task submission API.

docmind-20220712-b15f****

Response parameters

Parameter

Type

Description

Example

RequestId

string

The unique identifier for the request.

43A29C77-405E-4CC0-BC55-EE694AD0****

Data

string

The object that contains the detailed response data.

-

Status

string

The processing status of the job. For a list of possible values, see the note below.

success

NumberOfSuccessfulParsing

integer

The number of successfully parsed modules.

166

Tokens

long

The total number of tokens. A token is equivalent to one English word or one Chinese character.

4429

ParagraphCount

integer

The number of paragraphs.

91

TableCount

integer

The number of tables. This parameter is returned only if the uploaded document contains tables.

2

ImageCount

integer

The number of images. This parameter is returned only if the uploaded document contains images.

0

PageCountEstimate

integer

The estimated page count.

3

Processing

float

The processing progress of the file as a percentage.

100.0

OutputFormatResult

array

The parsing results in the specified format. This parameter is returned only if the OutputFormat parameter was set in the submission request. View outputFormatResult attribute details

Code

string

The status code.

200

Message

string

A detailed message about the response.

Message

Note

The Status parameter can have the following values:

  • Init: The job is queued for processing.

  • Processing: The job is being parsed.

  • success: The job completed successfully. The value of NumberOfSuccessfulParsing is now final.

  • Fail: The job failed to process.

Examples

The following examples show how to query a document parsing result by using various SDKs. In each example, the job ID is passed to the Id parameter.

Note

To learn how to obtain and use an AccessKey, see the SDK overview.

import com.aliyun.docmind_api20220711.models.*;
import com.aliyun.teaopenapi.models.Config;
import com.aliyun.docmind_api20220711.Client;

public static void main(String[] args) throws Exception {
        submit();
    }
public static void submit() throws Exception {
    // Initialize the Credentials client with default credentials.
    com.aliyun.credentials.Client credentialClient = new com.aliyun.credentials.Client();
    Config config = new Config()
        // Obtain the AccessKey ID from the credentials.
        .setAccessKeyId(credentialClient.getAccessKeyId())
        // Obtain the AccessKey secret from the credentials.
        .setAccessKeySecret(credentialClient.getAccessKeySecret());
    // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for both IPv4 and IPv6.
    config.endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
    Client client = new Client(config);
    QueryDocParserStatusRequest resultRequest = new QueryDocParserStatusRequest();
    resultRequest.id = "docmind-20220902-824b****";
    QueryDocParserStatusResponse response = client.queryDocParserStatus(resultRequest);
    System.out.println(com.alibaba.fastjson.JSON.toJSON(response.getBody()));
}
const Client = require('@alicloud/docmind-api20220711');
const Credential = require('@alicloud/credentials');

const getResult = async () => {
	// Initialize the Credentials client with default credentials.
  const cred = new Credential.default();
  const client = new Client.default({
    // The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for both IPv4 and IPv6.
    endpoint: 'docmind-api.cn-hangzhou.aliyuncs.com',
    // Obtain the AccessKey ID from the credentials.
    accessKeyId: cred.credential.accessKeyId,
    // Obtain the AccessKey secret from the credentials.
    accessKeySecret: cred.credential.accessKeySecret,
    type: 'access_key',
    regionId: 'cn-hangzhou'
  });
  
  const resultRequest = new Client.QueryDocParserStatusRequest();
  resultRequest.id = "docmind-20220902-824b****";
  const response = await client.queryDocParserStatus(resultRequest);
  
  return response.body;
}
from typing import List
from alibabacloud_docmind_api20220711.client import Client as docmind_api20220711Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_docmind_api20220711 import models as docmind_api20220711_models
from alibabacloud_tea_util.client import Client as UtilClient
from alibabacloud_credentials.client import Client as CredClient

if __name__ == '__main__':
  	# Initialize the Credentials client with default credentials.
    cred=CredClient()
    config = open_api_models.Config(
        # Obtain the AccessKey ID from the credentials.
        access_key_id=cred.get_credential().get_access_key_id(),
        # Obtain the AccessKey secret from the credentials.
        access_key_secret=cred.get_credential().get_access_key_secret()
    )
    # The endpoint of the service.
    config.endpoint = f'docmind-api.cn-hangzhou.aliyuncs.com'
    client = docmind_api20220711Client(config)
    request = docmind_api20220711_models.QueryDocParserStatusRequest(
        # id: The job ID, which is returned by the asynchronous task submission API.
        id='docmind-20220902-824b****'
    )
    try:
        # If you copy this code, print the API response.
        response = client.query_doc_parser_status(request)
        # The API response is structured as body -> data -> specific properties. Property names start with a lowercase letter. You can print the results as needed.
        # We recommend that you convert response.body.data to JSON and then retrieve the required values.
        print(response.body)
    except Exception as error:
        # If needed, print the error.
        UtilClient.assert_as_string(error.message)  
import (
	"fmt"

	openClient "github.com/alibabacloud-go/darabonba-openapi/v2/client"
  "github.com/alibabacloud-go/docmind-api-20220711/client"
  "github.com/aliyun/credentials-go/credentials"
)

func submit(){
    // Initialize the Credentials client with default credentials.
		credential, err := credentials.NewCredential(nil)
		// Obtain the AccessKey ID from the credentials.
		accessKeyId, err := credential.GetAccessKeyId()
		// Obtain the AccessKey secret from the credentials.
		accessKeySecret, err := credential.GetAccessKeySecret()
  	// The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for both IPv4 and IPv6.
  	var endpoint string = "docmind-api.cn-hangzhou.aliyuncs.com"
		config := openClient.Config{AccessKeyId: accessKeyId, AccessKeySecret: accessKeySecret, Endpoint: &endpoint}
    // Initialize the client.
    cli, err := client.NewClient(&config)
    if err != nil {
      panic(err)
    }
    id := "docmind-20220925-76b1****"
    // Call the query API.
    request := client.QueryDocParserStatusRequest{Id: &id}
    response, err := cli.QueryDocParserStatus(&request)
    if err != nil {
      panic(err)
    }
    // Print the query result.
    fmt.Println(response.Body.String())
}
using Newtonsoft.Json;
using System;
using System.Collections;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;

using Tea;
using Tea.Utils;

 public static void GetResult() 
        {
            // Initialize the Credentials client with default credentials.
          	var akCredential = new Aliyun.Credentials.Client(null);
            AlibabaCloud.OpenApiClient.Models.Config config = new AlibabaCloud.OpenApiClient.Models.Config
            {
                // Obtain the AccessKey ID from the credentials.
                AccessKeyId = akCredential.GetAccessKeyId(),
                // Obtain the AccessKey secret from the credentials.
                AccessKeySecret = akCredential.GetAccessKeySecret(),
            };
            // The endpoint of the service.
            config.Endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
            AlibabaCloud.SDK.Docmind_api20220711.Client client = new AlibabaCloud.SDK.Docmind_api20220711.Client(config);
            AlibabaCloud.SDK.Docmind_api20220711.Models.QueryDocParserStatusRequest request = new AlibabaCloud.SDK.Docmind_api20220711.Models.QueryDocParserStatusRequest
            {
                Id = "docmind-20240902-824b****"
            };
            AlibabaCloud.TeaUtil.Models.RuntimeOptions runtime = new AlibabaCloud.TeaUtil.Models.RuntimeOptions();
            try
            {
                // If you copy this code, print the API response.
                client.QueryDocParserStatus(request);
            }
            catch (TeaException error)
            {
                // If needed, print the error.
                AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
            }
            catch (Exception _error)
            {
                TeaException error = new TeaException(new Dictionary<string, object>
                {
                    { "message", _error.Message }
                });
                // If needed, print the error.
                AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
            }
        }
use AlibabaCloud\SDK\Docmindapi\V20220711\Docmindapi;
use AlibabaCloud\SDK\Docmindapi\V20220711\Models\GetDocStructureResultRequest;
use Darabonba\OpenApi\Models\Config;
use AlibabaCloud\Tea\Utils\Utils\RuntimeOptions;
use AlibabaCloud\Tea\Exception\TeaUnableRetryError;
use AlibabaCloud\Credentials\Credential;

// Initialize the Credentials client with default credentials.
$bearerToken = new Credential();    
$config = new Config();
// The endpoint of the service. You can use docmind-api-dualstack.cn-hangzhou.aliyuncs.com for both IPv4 and IPv6.
$config->endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
// Obtain the AccessKey ID from the credentials.
$config->accessKeyId = $bearerToken->getCredential()->getAccessKeyId();
// Obtain the AccessKey secret from the credentials.
$config->accessKeySecret = $bearerToken->getCredential()->getAccessKeySecret();
$config->type = "access_key";
$config->regionId = "cn-hangzhou";
$client = new Docmindapi($config);
$request = new QueryDocParserStatusRequest();   
$request->id = "docmind-20220902-824b****";

$runtime = new RuntimeOptions();
$runtime->maxIdleConns = 3;
$runtime->connectTimeout = 10000;
$runtime->readTimeout = 10000; 

try {
  $response = $client->queryDocParserStatus($request, $runtime);
  var_dump($response->toMap());
} catch (TeaUnableRetryError $e) {
  var_dump($e->getMessage());
  var_dump($e->getErrorInfo());
  var_dump($e->getLastException());
  var_dump($e->getLastRequest());
}

Sample success response:

{
    "RequestId": "43A29C77-405E-4DC0-BC55-EE694AD0****",
    "Data": {
        "Status": "success",
        "NumberOfSuccessfulParsing": 93,
        "ImageCount": 0,
        "PageCountEstimate": 8,
        "ParagraphCount": 91,
        "TableCount": 2,
        "Tokens": 4429
    }
}

Sample failure response:

{
    "code": "FileUrlLegal",
    "message": "File url is not legal.",
    "RequestId": "E26AC4F5-F633-5468-8F6F-708D2B3FXXXX"
}

Step 3: Call the GetDocParserResult API

Request parameters

Parameter

Type

Required

Description

Example

Id

string

Yes

The business order ID to query. You can get this ID from the response of the submission API.

docmind-20220712-b15f****

LayoutStepSize

integer

Yes

  • The number of layout blocks to retrieve in a single request.

  • Minimum value: 1

  • Maximum value: 3000

100

LayoutNum

integer

Yes

  • The index of the first layout block to retrieve.

  • Minimum value: 0

0

ExcludeFields

List<String>

No

Fields to exclude from the response. Currently, only top-level fields can be filtered. For a list of excludable fields, refer to the top-level fields of the layout object described in Response for successful file format processing.

"text","firstLinesChars"

Important
  • You can page through results from the GetDocParserResult API by making multiple calls. For example, use a range like [LayoutNum, LayoutNum + LayoutStepSize] to fetch blocks in batches. This method lets you retrieve parsed content even while a long document parsing job is still running.

  • If the value of LayoutStepSize is too large, an error may occur. Reduce the value if necessary.

Note

image

Response parameters

Parameter

Type

Description

Example

RequestId

string

The unique request ID.

43A29C77-405E-4CC0-BC55-EE694AD0****

Data

string

The parsing result from Document Parsing (Large Model Edition).

-

Code

string

The status code.

200

Message

string

The response message.

Message

Examples

The following examples show how to use various SDKs to query the result of a document parsing job.

import com.aliyun.docmind_api20220711.models.*;
import com.aliyun.teaopenapi.models.Config;
import com.aliyun.docmind_api20220711.Client;

public static void main(String[] args) throws Exception {
        submit();
    }
public static void submit() throws Exception {
    // Initialize the credential client by using the default credential provider.
    com.aliyun.credentials.Client credentialClient = new com.aliyun.credentials.Client();
    Config config = new Config()
        // Obtain the AccessKey ID from the credential.
        .setAccessKeyId(credentialClient.getAccessKeyId())
        // Obtain the AccessKey Secret from the credential.
        .setAccessKeySecret(credentialClient.getAccessKeySecret());
    // The service endpoint. IPv4 and IPv6 are supported. For IPv6, use docmind-api-dualstack.cn-hangzhou.aliyuncs.com.
    config.endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
    Client client = new Client(config);
    GetDocParserResultRequest resultRequest = new GetDocParserResultRequest();
    // The job ID. This is the ID returned by the asynchronous SubmitDocParserJob API.
    resultRequest.id = "docmind-20220902-824b****";
    /**
     * When you use the GetDocParserResult operation, you may need to make multiple queries.
     * For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
     * This allows you to retrieve the parsed content of a long document in advance.
    */
    // The step size for querying layout blocks. Valid values: 1 to 3000.
    resultRequest.layoutStepSize = 10;
    // The starting position of the query. Minimum: 0.
    resultRequest.layoutNum = 0;
    GetDocParserResultResponse response = client.getDocParserResult(resultRequest);
    System.out.println(com.alibaba.fastjson.JSON.toJSON(response.getBody()));
}
const Client = require('@alicloud/docmind-api20220711');
const Credential = require('@alicloud/credentials');

const getResult = async () => {
	// Initialize the credential client by using the default credential provider.
  const cred = new Credential.default();
  const client = new Client.default({
    // The service endpoint. IPv4 and IPv6 are supported. For IPv6, use docmind-api-dualstack.cn-hangzhou.aliyuncs.com.
    endpoint: 'docmind-api.cn-hangzhou.aliyuncs.com',
    // Obtain the AccessKey ID from the credential.
    accessKeyId: cred.credential.accessKeyId,
    // Obtain the AccessKey Secret from the credential.
    accessKeySecret: cred.credential.accessKeySecret,
    type: 'access_key',
    regionId: 'cn-hangzhou'
  });
  
  const resultRequest = new Client.GetDocParserResultRequest();
  resultRequest.id = "docmind-20220902-824b****";
  /**
   * When you use the GetDocParserResult operation, you may need to make multiple queries.
   * For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
   * This allows you to retrieve the parsed content of a long document in advance.
   */
  // The step size for querying layout blocks. Valid values: 1 to 3000.
  resultRequest.layoutStepSize = 10;
  // The starting position of the query. Minimum: 0.
  resultRequest.layoutNum = 0;
  const response = await client.getDocParserResult(resultRequest);  
  return response.body;
}
from typing import List
from alibabacloud_docmind_api20220711.client import Client as docmind_api20220711Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_docmind_api20220711 import models as docmind_api20220711_models
from alibabacloud_tea_util.client import Client as UtilClient
from alibabacloud_credentials.client import Client as CredClient

if __name__ == '__main__':
    # Initialize the credential client by using the default credential provider.
    cred = CredClient()
    config = open_api_models.Config(
        # Obtain the AccessKey ID from the credential.
        access_key_id=cred.get_credential().get_access_key_id(),
        # Obtain the AccessKey Secret from the credential.
        access_key_secret=cred.get_credential().get_access_key_secret()
    )
    # The service endpoint.
    config.endpoint = f'docmind-api.cn-hangzhou.aliyuncs.com'
    client = docmind_api20220711Client(config)
    request = docmind_api20220711_models.GetDocParserResultRequest(
        # The job ID, returned by the SubmitDocParserJob operation.
        id='docmind-20220902-824b****',
        '''
          When you use the GetDocParserResult operation, you may need to make multiple queries.
          For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
          This allows you to retrieve the parsed content of a long document in advance.
        '''
        # LayoutStepSize: The number of layout blocks to query. This is the step size. Minimum value: 1. Maximum value: 3000.
        layout_step_size=10,
        # LayoutNum: The starting position of the query. Minimum value: 0.
        layout_num=0
    )
    try:
        # Make sure to print the API response to see the output.
        response = client.get_doc_parser_result(request)
        # The API response is structured as body -> data -> specific properties. You can print the results as needed. The property values start with a lowercase letter.
        # To retrieve specific values, convert response.body.data to JSON.
        print(response.body)
    except Exception as error:
        # Print the error if necessary.
        UtilClient.assert_as_string(error.message)
import (
	"fmt"

	openClient "github.com/alibabacloud-go/darabonba-openapi/v2/client"
  "github.com/alibabacloud-go/docmind-api-20220711/client"
  "github.com/aliyun/credentials-go/credentials"
)

func submit(){
    // Initialize the credential client by using the default credential provider.
		credential, err := credentials.NewCredential(nil)
		// Obtain the AccessKey ID from the credential.
		accessKeyId, err := credential.GetAccessKeyId()
		// Obtain the AccessKey Secret from the credential.
		accessKeySecret, err := credential.GetAccessKeySecret()
  	// The service endpoint. IPv4 and IPv6 are supported. For IPv6, use docmind-api-dualstack.cn-hangzhou.aliyuncs.com.
  	var endpoint string = "docmind-api.cn-hangzhou.aliyuncs.com"
		config := openClient.Config{AccessKeyId: accessKeyId, AccessKeySecret: accessKeySecret, Endpoint: &endpoint}
    // Initialize the client.
    cli, err := client.NewClient(&config)
    if err != nil {
      panic(err)
    }
    id := "docmind-20220925-76b1****"
    /**
     * When you use the GetDocParserResult operation, you may need to make multiple queries.
    * For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
    * This allows you to retrieve the parsed content of a long document in advance.
    */
    // LayoutStepSize: The number of layout blocks to query. This is the step size. Minimum value: 1. Maximum value: 3000.
    layoutStepSize := 10
    // LayoutNum: The starting position of the query. Minimum value: 0.
    layoutNum := 0
    // Call the operation.
    request := client.GetDocParserResultRequest{Id: &id}
    response, err := cli.GetDocParserResult(&request)
    if err != nil {
      panic(err)
    }
    // Print the result.
    fmt.Println(response.Body.String())
}
using Newtonsoft.Json;
using System;
using System.Collections;
using System.Collections.Generic;
using System.IO;
using System.Threading.Tasks;

using Tea;
using Tea.Utils;

 public static void GetResult() 
        {
            // Initialize the credential client by using the default credential provider.
          	var akCredential = new Aliyun.Credentials.Client(null);
            AlibabaCloud.OpenApiClient.Models.Config config = new AlibabaCloud.OpenApiClient.Models.Config
            {
                // Obtain the AccessKey ID from the credential.
                AccessKeyId = akCredential.GetAccessKeyId(),
                // Obtain the AccessKey Secret from the credential.
                AccessKeySecret = akCredential.GetAccessKeySecret(),
            };
            // The service endpoint.
            config.Endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
            AlibabaCloud.SDK.Docmind_api20220711.Client client = new AlibabaCloud.SDK.Docmind_api20220711.Client(config);
            AlibabaCloud.SDK.Docmind_api20220711.Models.GetDocParserResultRequest request = new AlibabaCloud.SDK.Docmind_api20220711.Models.GetDocParserResultRequest
            {
                /**
                 * When you use the GetDocParserResult operation, you may need to make multiple queries.
                 * For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
                 * This allows you to retrieve the parsed content of a long document in advance.
                 */
                // The job ID. This is the ID returned by the asynchronous SubmitDocParserJob API.
                Id = "docmind-20240902-824b****",
                // The step size for querying layout blocks. Valid values: 1 to 3000.
                LayoutStepSize = 10,
                // The starting position of the query. Minimum: 0.
                LayoutNum = 0
            };
            AlibabaCloud.TeaUtil.Models.RuntimeOptions runtime = new AlibabaCloud.TeaUtil.Models.RuntimeOptions();
            try
            {
                // Make sure to print the API response to see the output.
                client.GetDocParserResult(request);
            }
            catch (TeaException error)
            {
                // Print the error if necessary.
                AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
            }
            catch (Exception _error)
            {
                TeaException error = new TeaException(new Dictionary<string, object>
                {
                    { "message", _error.Message }
                });
                // Print the error if necessary.
                AlibabaCloud.TeaUtil.Common.AssertAsString(error.Message);
            }
        }
use AlibabaCloud\SDK\Docmindapi\V20220711\Docmindapi;
use AlibabaCloud\SDK\Docmindapi\V20220711\Models\GetDocStructureResultRequest;
use Darabonba\OpenApi\Models\Config;
use AlibabaCloud\Tea\Utils\Utils\RuntimeOptions;
use AlibabaCloud\Tea\Exception\TeaUnableRetryError;
use AlibabaCloud\Credentials\Credential;

// Initialize the credential client by using the default credential provider.
$bearerToken = new Credential();    
$config = new Config();
// The service endpoint. IPv4 and IPv6 are supported. For IPv6, use docmind-api-dualstack.cn-hangzhou.aliyuncs.com.
$config->endpoint = "docmind-api.cn-hangzhou.aliyuncs.com";
// Obtain the AccessKey ID from the credential.
$config->accessKeyId = $bearerToken->getCredential()->getAccessKeyId();
// Obtain the AccessKey Secret from the credential.
$config->accessKeySecret = $bearerToken->getCredential()->getAccessKeySecret();
$config->type = "access_key";
$config->regionId = "cn-hangzhou";
$client = new Docmindapi($config);
$request = new GetDocParserResultRequest();   
$request->id = "docmind-20220902-824b****";
/**
 * When you use the GetDocParserResult operation, you may need to make multiple queries.
 * For example, use [LayoutNum, LayoutNum + LayoutStepSize] to retrieve content blocks 0-100, 100-200, and so on.
 * This allows you to retrieve the parsed content of a long document in advance.
 */
// The step size for querying layout blocks. Valid values: 1 to 3000. 
$request->layoutStepSize = 10;
// The starting position of the query. Minimum: 0.
$request->layoutNum = 0;

$runtime = new RuntimeOptions();
$runtime->maxIdleConns = 3;
$runtime->connectTimeout = 10000;
$runtime->readTimeout = 10000; 

try {
  $response = $client->getDocParserResult($request, $runtime);
  var_dump($response->toMap());
} catch (TeaUnableRetryError $e) {
  var_dump($e->getMessage());
  var_dump($e->getErrorInfo());
  var_dump($e->getLastException());
  var_dump($e->getLastRequest());
}
Note

For information about how to obtain and use an AccessKey, see the guides for different programming languages in SDK Overview.

Sample response

Failed response
{
  "RequestId": "A8EF3A36-1380-1116-A39E-B377BE27****",
  "Code": "UrlNotLegal",
  "Message": "Failed to process the document.  The document url you provided is not legal.",
  "HostId": "docmind-api.cn-hangzhou.aliyuncs.com",
  "Recommend": "https://next.api.aliyun.com/troubleshoot?q=IDP.UrlNotLegal&product=docmind-api"
}

If the request fails, the response includes an error code and a detailed message. For more information, see Error codes.

Successful response (file)
{
    "Data": {
        "layouts": [
            {
                "firstLinesChars": 0,
                "level": 0,
                "blocks": [
                    {
                        "text": "PRACTITIONERS’ SECTION"
                    }
                ],
                "markdownContent": "# PRACTITIONERS’ SECTION  \n\n",
                "index": 2,
                "subType": "doc_title",
                "lineHeight": 0,
                "text": "PRACTITIONERS’ SECTION",
                "alignment": "center",
                "type": "title",
                "pageNum": 0,
                "uniqueId": "3d8ded229371f879eac478dee2574a82"
            },
            {
                "firstLinesChars": 0,
                "level": 0,
                "blocks": [
                    {
                        "text": "RAGGING: A PUBLIC HEALTH PROBLEM IN INDIA"
                    }
                ],
                "markdownContent": "# RAGGING: A PUBLIC HEALTH PROBLEM IN INDIA  \n\n",
                "index": 3,
                "subType": "doc_subtitle",
                "lineHeight": 0,
                "text": "RAGGING: A PUBLIC HEALTH PROBLEM IN INDIA",
                "alignment": "center",
                "type": "title",
                "pageNum": 0,
                "uniqueId": "44256f2d41419956ff30c6590683217b"
            },
            {
                "firstLinesChars": 0,
                "level": 1,
                "blocks": [
                    {
                        "text": "RAJESH GARG"
                    }
                ],
                "markdownContent": "RAJESH GARG  \n\n",
                "index": 4,
                "subType": "para",
                "lineHeight": 0,
                "text": "RAJESH GARG",
                "alignment": "center",
                "type": "text",
                "pageNum": 0,
                "uniqueId": "d0d64505df8cb4d73f80e0f63007bf60"
            }
        ]
    },
    "RequestId": "7B8CC68D-D498-5EDA-8352-FAF41591D97A"
}

Parameters

Parameter

Type

Description

layouts

array

A list of layout blocks.

text

string

The text content.

markdownContent

string

The content in Markdown format.

index

int

The reading order of the layout block.

uniqueId

string

The unique ID of the layout block.

alignment

string

The alignment of the content.

pageNum

int

The page number of the layout block.

level

int

The level of the layout block. A value of 0 indicates the root node.

type

string

The type of the layout block. For more information, see the notes on layout types.

subType

string

The subtype of the layout block. For more information, see the notes on layout types.

llmResult

string

The result from the large language model. For details, see the llmResult examples below.

  • For charts in the document:

    • The results are classified as "value chart", "valuation chart", or "other chart". The category is specified on the line after <|Category|>.

    • The data table in Markdown format is specified after <|Value|>.

    • If the category is "other chart", no data table is generated.

  • For PPT files, the result is specified after markdown\n .

layoutConf

float

The confidence of the parsing result, ranging from 0 to 1.

pos

array

The coordinates.

x

int

The x-coordinate of the vertex.

y

int

The y-coordinate of the vertex.

Successful response (audio and video)
{
    "Data": {
        "segments": [
                        {
                "audio_frames": [
                    {
                        "ASR_info": "Ne Zha is just a self-indulgent hype for Chinese netizens. It only has box office success in China; no one watches it anywhere else.",
                        "end_time": 5019.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/audio/request_1111_1.mp3?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645100&Signature=ioyj0ogA2pXSod1gPlM4KMCXLr4%3D",
                        "start_time": 0.0
                    }
                ],
                "end_time": 5019.0,
                "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/video/request_1111_0.mp4?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645176&Signature=RoMsrYsA%2B4jkuJTQA6Pjpv2knBY%3D",
                "index": 0,
                "start_time": 0.0,
                "video_frames": [
                    {
                        "end_time": 4960.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/frame/request_1111-Scene-001-01.jpg?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645151&Signature=E8IsZesWrW%2B3i8VLs7ZXLqMFZYg%3D",
                        "start_time": 0.0,
                        "text_info": "A man wearing glasses holds a mobile phone in a circular frame. A bookshelf is in the background. A comment is displayed on the right: 'Ne Zha is just a self-indulgent hype for Chinese netizens. It only has box office success in China; no one watches it anywhere else.' Below are icons for like, dislike, share, and comment. The subtitle at the bottom reads: 'Only has box office success in China.'"
                    },
                    {
                        "end_time": 10600.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/frame/request_1111-Scene-002-01.jpg?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645151&Signature=ZDOgHjs0iKmYCPcmsxTPk6fhg64%3D",
                        "start_time": 4960.0,
                        "text_info": "A man wearing glasses stands in front of a bookshelf that holds books and two Minion figurines. He is wearing a brown shirt and speaking. The subtitle at the bottom reads: 'So why would anyone say that?'"
                    }
                ]
            },
            {
                "audio_frames": [
                    {
                        "ASR_info": "Because this statement hides a very deep chain of discrimination. Let's consider a few questions. First, is Ne Zha's low foreign box office because foreign audiences don't like it or because they can't see it?",
                        "end_time": 19900.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/audio/request_1111_3.mp3?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645101&Signature=7QwfDASrWywOA%2FmlZ64r%2FMcu0b4%3D",
                        "start_time": 10577.0
                    }
                ],
                "end_time": 19900.0,
                "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/video/request_1111_2.mp4?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645181&Signature=m03c5oHyqY88mmtPLdSgMpi8Ul4%3D",
                "index": 2,
                "start_time": 10577.0,
                "video_frames": [
                    {
                        "end_time": 14080.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/frame/request_1111-Scene-003-01.jpg?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645151&Signature=NCAM%2Bt1KLTNDt4jIFXOH%2B0m13Vs%3D",
                        "start_time": 10600.0,
                        "text_info": "A man wearing glasses stands in front of a bookshelf that holds books and two Minion figurines. He is wearing a brown shirt and has a serious expression. The subtitle at the bottom reads: 'hides a very deep chain of discrimination.'"
                    },
                    {
                        "end_time": 15600.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/frame/request_1111-Scene-004-01.jpg?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645151&Signature=%2FhmtuSNvtjwHI2Ho88tIv9keM3k%3D",
                        "start_time": 14080.0,
                        "text_info": "A man wearing glasses stands in front of a bookshelf that holds books and two Minion figurines. He is wearing a dark shirt and speaking. The text at the bottom of the screen reads: 'Let's consider a few questions.'"
                    },
                    {
                        "end_time": 52920.0,
                        "file_url": "http://doc-mind-video.oss-cn-hangzhou.aliyuncs.com/multimodal_data/20251125/frame/request_1111-Scene-005-01.jpg?OSSAccessKeyId=LTAI5tPtEwpyT4JR9XXXXX&Expires=1764645151&Signature=vfYhFendANuuzihVIlQAmq0YAuQ%3D",
                        "start_time": 15600.0,
                        "text_info": "A black background with a white box in the center containing the number '3'. Below it, white text reads: 'Is Captain America 4 'self-indulgent hype'?' The subtitle at the bottom reads: 'But it has such a high number of screenings in the US.'"
                    }
                ]
            }   
        ],
        "synopsis_result": "### 0s - 5s\n**Frame parsing result**: A man wearing glasses stands in front of a bookshelf that holds books and two Minion figurines. He is wearing a brown shirt and speaking. The subtitle at the bottom reads: 'So why would anyone say that?'"
    },
    "RequestId": "7B8CC68D-D498-5EDA-8352-FAF41591D97A"
}

Parameters

Parameter

Type

Description

segments

array

An array of parsed segments.

start_time

float

The start time of the segment, in milliseconds (ms).

end_time

float

The end time of the segment, in milliseconds (ms).

index

int

The index of the segment, from 0 to n-1.

file_url

string

The file URL of the segment.

audio_frames

array

The parsing information for the audio frames in the segment.

video_frames

array

The parsing information for the video frames in the segment.

synopsis_result

string

The plot summary. This field is returned only when option is set to advance.

In responses from Document Parsing (LLM Edition), the audio_frames object contains the following fields:

Parameter

Description

Type

start_time

The start time of the audio, in milliseconds (ms).

float

end_time

The end time of the audio, in milliseconds (ms).

float

file_url

The URL of the audio file.

string

ASR_info

The Automatic Speech Recognition (ASR) result.

string

In responses from Document Parsing (LLM Edition), the video_frames object contains the following fields:

Parameter

Description

Type

start_time

The start time of the video, in milliseconds (ms).

float

end_time

The end time of the video, in milliseconds (ms).

float

file_url

The URL of the video file.

string

text_info

The parsing result for the video frame from the large language model.

string

The following examples show the llmResult field in the response from Document Parsing (LLM Edition).

  1. For a PPT document:

{
  "llmResult": "```markdown\n# Welcome to Capital Markets Day London 2015\n\n## HEXAGON\n### Shaping Smart Change\n```",
  "layoutConf": 0.6
}
  1. For a chart in a document:

{
  "llmResult": "<|Category|>\nValuation chart\n<|Value|>\n| Protocol Followed | Time Required (Minutes) |\n|-------------------|-------------------------|\n| Standard          | 80                      |\n| Halifax             | 80                      |\n| Halifaster        | 40                      |",
  "layoutConf": 0.4
}

Layout type

The following tables list the layout types (type) and subtypes (subType) returned by the document analysis large model.

Type

Description

Subtype

Description

title

Title

doc_name

Document name

doc_title

Document title

doc_subtitle

Document subtitle

para_title

Paragraph title

contents_title

Table of contents title

cate_title

Table of contents title

contents

Table of contents

cate

Table of contents

text

Plain text

para

Paragraph

figure

Figure

picture

Picture

logo

Logo

figure_name

Figure title

pic_title

Picture title

figure_note

Figure caption

pic_caption

Picture caption

foot

Page footer

page_footer

Page footer

head

Page header

page_header

Page header

head_pagenum

Header page number

page

Page number

foot_pagenum

Footer page number

page

Page number

corner_note

Footnote

footer_note

Footnote

end_note

Endnote

endnode

Endnote

side

Sidebar

sidebar

Sidebar

The table below lists additional layout types.

Type

Description

table_name

Table title

table_note

Table note

formula

Formula

The following types do not have a subtype:

Type

Description

multicolumn

Multi-column text

table

Table

foot_image

Footer image

head_image

Header image

The following table describes the properties of outputFormatResult.

Parameter

Description

outputType

The output format of the result.

outputFileUrl

The URL of the output file.

pages

Contains details about the images on each page.

The following table describes the values for outputType.

Value

Description

markdown

Markdown format.

pageImage

Page image format.

The following table describes the properties of pages.

Parameter

Description

imageUrl

The URL of the image.

imageWidth

The width of the image.

imageHeight

The height of the image.

pageIdCurDoc

The current page number within the document.

Reference example (Python)

Parsing to markdown (local file upload)

To disable VLM enhancement, set llm_enhancement=False.

Usage:

python3 docmind_example_parser.py input_file output_file.md
#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import time
import json
from alibabacloud_docmind_api20220711.client import Client as docmind_api20220711Client
from alibabacloud_tea_openapi import models as open_api_models
from alibabacloud_docmind_api20220711 import models as docmind_api20220711_models
from alibabacloud_tea_util.client import Client as UtilClient
from alibabacloud_tea_util import models as util_models
from alibabacloud_credentials.client import Client as CredClient

class DocParser:
    def __init__(self, endpoint="docmind-api.cn-hangzhou.aliyuncs.com"):
        """
        Initializes the document parser.

        Args:
            endpoint (str): The API endpoint.
        """
        self.endpoint = endpoint
        self.client = self._init_client()

    def _init_client(self):
        """
        Initializes the API client.

        Returns:
            docmind_api20220711Client: The API client instance.
        """
        # Initialize the `CredClient` with the default credential.
        cred = CredClient()
        config = open_api_models.Config(
            # Get the AccessKey ID.
            access_key_id=cred.get_credential().get_access_key_id(),
            # Get the AccessKey Secret.
            access_key_secret=cred.get_credential().get_access_key_secret()
        )
        # Set the endpoint.
        config.endpoint = self.endpoint
        return docmind_api20220711Client(config)

    def submit_job(self, file_path, file_name=None):
        """
        Submits a document parsing task.

        Args:
            file_path (str): The path to the local file.
            file_name (str, optional): The name of the file.

        Returns:
            str: The task ID, or None if the submission fails.
        """
        try:
            # Extract the filename from the path if not provided.
            if not file_name:
                file_name = file_path.split('/')[-1]

            # Build the request object.
            # To disable VLM enhancement, set llm_enhancement=False.
            request = docmind_api20220711_models.SubmitDocParserJobAdvanceRequest(
                file_url_object=open(file_path, "rb"),
                file_name=file_name,
                file_name_extension=file_name.split('.')[-1] if '.' in file_name else None,
                llm_enhancement=True,
                enhancement_mode="VLM",
            )

            runtime = util_models.RuntimeOptions()

            # Submit the task.
            response = self.client.submit_doc_parser_job_advance(request, runtime)
            task_id = response.body.data.id

            print(f"Task submitted. Task ID: {task_id}")
            return task_id
        except Exception as error:
            print(f"Error submitting the task: {error}")
            return None

    def query_status(self, task_id):
        """
        Queries the task status.

        Args:
            task_id (str): The task ID.

        Returns:
            dict: The task status, or None if the query fails.
        """
        try:
            request = docmind_api20220711_models.QueryDocParserStatusRequest(id=task_id)
            response = self.client.query_doc_parser_status(request)
            return response.body.data.to_map() if response.body.data else None
        except Exception as error:
            print(f"Error querying task status: {error}")
            return None

    def get_result(self, task_id, layout_num=0, layout_step_size=10):
        """
        Fetches the document parsing result. Supports incremental fetch.

        Args:
            task_id (str): The task ID.
            layout_num (int): The starting layout number.
            layout_step_size (int): The step size.

        Returns:
            dict: The parsing result, or None if the retrieval fails.
        """
        try:
            request = docmind_api20220711_models.GetDocParserResultRequest(
                id=task_id,
                layout_step_size=layout_step_size,
                layout_num=layout_num
            )
            response = self.client.get_doc_parser_result(request)
            return response.body.data if response.body.data else None
        except Exception as error:
            print(f"Error fetching the parsing result: {error}")
            return None

    def wait_for_completion(self, task_id, poll_interval=5):
        """
        Waits for the task to complete by polling its status.

        Args:
            task_id (str): The task ID.
            poll_interval (int): The polling interval in seconds.

        Returns:
            bool: True if the task completes successfully, False otherwise.
        """
        print("Waiting for task completion...")
        while True:
            status_data = self.query_status(task_id)
            if not status_data:
                return False
            print(status_data)
            status = status_data.get('Status', '').lower()

            # Check for task completion.
            if status == 'success':
                print("Task completed successfully.")
                return True
            elif status == 'failed':
                print("Task failed.")
                return False
            else:
                # The task is still processing.
                print(".")
                time.sleep(poll_interval)

    def collect_results_incrementally(self, task_id, layout_step_size=10):
        """
        Fetches parsing results incrementally.

        Args:
            task_id (str): The task ID.
            layout_step_size (int): The step size.

        Yields:
            list: A list of layout blocks from each fetch.
        """
        layout_num = 0
        while True:
            result_data = self.get_result(task_id, layout_num, layout_step_size)
            if not result_data:
                break
            layouts = result_data.get('layouts', [])
            if not layouts:
                break

            yield layouts

            # Update the starting position for the next fetch.
            layout_num += len(layouts)

            # If the number of returned layouts is less than the step size, fetching is complete.
            if len(layouts) < layout_step_size:
                break

    def table_to_html(self, table_layout):
        """
        Converts a table layout to an HTML string.

        Args:
            table_layout (dict): The table layout data.

        Returns:
            str: The table in HTML format.
        """
        cells = table_layout.get('cells', [])
        if not cells:
            return ""

        # Create the HTML table.
        html_parts = ['<table border="1" cellspacing="0" cellpadding="2">']

        # Track processed cells to handle row and column spans.
        processed_cells = set()

        # Group cells by row.
        rows = {}
        for cell in cells:
            row_start = cell.get('ysc', 0)
            if row_start not in rows:
                rows[row_start] = []
            rows[row_start].append(cell)

        # Process rows in order.
        for row_idx in sorted(rows.keys()):
            html_parts.append('<tr>')
            # Sort cells in the row by column order.
            row_cells = sorted(rows[row_idx], key=lambda x: x.get('xsc', 0))

            for cell in row_cells:
                cell_key = (cell.get('ysc', 0), cell.get('xsc', 0))
                if cell_key in processed_cells:
                    continue

                # Calculate row and column spans.
                rowspan = cell.get('yec', 0) - cell.get('ysc', 0) + 1
                colspan = cell.get('xec', 0) - cell.get('xsc', 0) + 1

                # Mark cells as processed.
                for i in range(rowspan):
                    for j in range(colspan):
                        processed_cells.add((cell.get('ysc', 0) + i, cell.get('xsc', 0) + j))

                # Get the cell text.
                cell_text = ""
                cell_layouts = cell.get('layouts', [])
                for layout in cell_layouts:
                    if 'text' in layout:
                        cell_text += layout['text']

                # Clean up the text.
                cell_text = cell_text.strip().replace('\n', '<br>')

                # Add the cell HTML.
                cell_attrs = []
                if rowspan > 1:
                    cell_attrs.append(f'rowspan="{rowspan}"')
                if colspan > 1:
                    cell_attrs.append(f'colspan="{colspan}"')

                html_parts.append(f'<td {" ".join(cell_attrs)}>{cell_text}</td>')

            html_parts.append('</tr>')

        html_parts.append('</table>')
        return ''.join(html_parts)

    def generate_markdown(self, layouts):
        """
        Converts layout blocks to Markdown content.

        Args:
            layouts (list): A list of layout blocks.

        Returns:
            str: The content in Markdown format.
        """
        markdown_content = ""
        for layout in layouts:
            # Directly use the `markdownContent` field from the response.
            if layout.get('type')=="table":
                # For table types, convert to HTML format.
                table_html = self.table_to_html(layout)
                markdown_content += table_html + "\n\n"
            else:
                markdown_content += layout.get('markdownContent', '') + "\n"
        return markdown_content

    def process_document(self, file_path, output_path, layout_step_size=10, poll_interval=5):
        """
        Processes a document by submitting a task, polling for completion,
        fetching results incrementally, and generating a Markdown file.

        Args:
            file_path (str): The path to the local file.
            output_path (str): The path for the output Markdown file.
            layout_step_size (int): The step size for incremental fetch.
            poll_interval (int): The polling interval in seconds.

        Returns:
            bool: True on success, False otherwise.
        """
        # 1. Submit the task.
        print("Submitting the document parsing task...")
        task_id = self.submit_job(file_path)
        if not task_id:
            print("Failed to submit the task.")
            return False

        # 2. Wait for the task to complete.
        if not self.wait_for_completion(task_id, poll_interval):
            print("The task did not complete successfully.")
            return False

        # 3. Fetch results incrementally and write to the file.
        print("Task complete. Fetching parsing results...")
        with open(output_path, 'w', encoding='utf-8') as f:

            # Fetch and write content incrementally.
            for layouts in self.collect_results_incrementally(task_id, layout_step_size):
                markdown_content = self.generate_markdown(layouts)
                f.write(markdown_content)
                f.flush()  # Immediately flush to the file.
        print(f"Document parsing complete. Results saved to: {output_path}")
        return True

def main():
    import argparse

    parser = argparse.ArgumentParser(description='Document Parsing Tool')
    parser.add_argument('input_file', help='Path to the input document file.')
    parser.add_argument('output_file', help='Path for the output Markdown file.')
    parser.add_argument('--step-size', type=int, default=10, help='Step size for incremental fetch. Default: 10.')
    parser.add_argument('--poll-interval', type=int, default=5, help='Polling interval in seconds. Default: 5.')

    args = parser.parse_args()

    # Create a document parser instance.
    parser = DocParser()

    # Process the document.
    success = parser.process_document(
        args.input_file,
        args.output_file,
        layout_step_size=args.step_size,
        poll_interval=args.poll_interval
    )

    if success:
        print("Document processing completed.")
    else:
        print("Document processing failed.")
        exit(1)

if __name__ == "__main__":
    main()