Submit an extraction task

Submit a structured field extraction task and get a biz_id in return. The input can be a file URL or the parsed_file_biz_id returned by a previous parsing task. Field extraction is an async task: after submitting, use that ID to call Query the extraction result to get the processing status and structured result.

Make sure you have obtained an API Key before calling. See Authentication.

Request parameters

Parameter	Type	Required	Description
`file_url`	string	No	An accessible URL of the document to extract from. Mutually exclusive with `parsed_file_biz_id`.
`parsed_file_biz_id`	string	No	The business task ID of a completed parsing task. Mutually exclusive with `file_url`.
`file_name`	string	No	The file name.
`file_name_extension`	string	No	The file type.
`config_id`	string	No	A saved Extract configuration ID, for example `config_xxx`.
`processing`	object	No	Inline processing configuration; takes precedence over `config_id`.
`processing.user_prompt`	string	No	User prompt.
`processing.extract_processing_config`	object	No	Inline extraction processing configuration. Required when using `processing`.
`processing.extract_processing_config.citation_required`	boolean	No	Whether to return citation information; default `true`.
`processing.extract_processing_config.allow_inference`	boolean	No	Whether to allow inference; default `false`.
`processing.extract_processing_config.extract_schema`	object	No	The extraction schema, as a JSON object. Required when using inline processing configuration.
`output`	object	No	Output configuration.
`output.oss_config`	object	No	OSS output configuration.
`output.oss_config.bucket`	string	No	OSS bucket name.
`output.oss_config.endpoint`	string	No	OSS endpoint.
`output.oss_config.access_key_id`	string	No	OSS AccessKey ID.
`output.oss_config.access_key_secret`	string	No	OSS AccessKey Secret.
`output.oss_config.security_token`	string	No	OSS Security Token.

The request requires two pairs of mutually exclusive choices: use file_url or parsed_file_biz_id as the file source, and use config_id or inline processing as the processing definition. When reusing a parsing result, the result must still be within the 7-day retention period, and the input type must support reuse.

Response

Field	Type	Description
`request_id`	string	The request ID
`data.biz_id`	string	The business task ID, used to call `/extract/result`

Code examples

The following examples show the two ways of defining processing. When config_id and processing are both present, processing takes precedence. The URLs, task IDs, and configuration IDs in the examples are desensitized placeholder values.

Use a saved configuration
Use inline configuration

Submit

curl -X POST 'https://{workspaceId}.cn-beijing.maas.aliyuncs.com/api/v2/apps/parse-x/extract/submit' \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "file_url": "https://example.com/contracts/sample.pdf",
    "file_name": "sample.pdf",
    "config_id": "config_xxx"
  }'

extract_schema is a string, so the JSON it contains must be escaped.

Submit

curl -X POST 'https://{workspaceId}.cn-beijing.maas.aliyuncs.com/api/v2/apps/parse-x/extract/submit' \
  -H "Authorization: Bearer $DASHSCOPE_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "parsed_file_biz_id": "biz_xxx",
    "processing": {
        "user_prompt": "按 Schema 抽取合同信息",
        "extract_processing_config": {
            "citation_required": true,
            "allow_inference": false,
            "extract_schema": {
                "type": "object",
                "properties": {
                    "contract_number": {
                        "type": "string"
                    },
                    "total_amount": {
                        "type": "number"
                    }
                }
            }
        }
    }
}'

Extraction

{
  "request_id": "request_xxx",
  "data": {
    "biz_id": "biz_xxx"
  }
}

Extract does not support audio/video input. If a reused parsing result is beyond the 7-day retention period or its type does not match, the endpoint returns the corresponding error; see Error codes. The schema size and long-document page count are planned only within engineering design boundaries; public quotas are subject to the official page. See Supported files and limits.

After submitting, use Query the extraction result to poll the task status and get the structured result.

​Request parameters

​Response

​Code examples

Request parameters

Response

Code examples