Skip to main content

2026-08-17

API

Billing Enabled for the Extract Specific Fields API

  • The Extract Specific Fields API (POST /api/app-api/sip/platform/v2/file/extract_fields) is now billed per page: each call is billed independently based on the number of pages of the task’s document, at the “other file types” rate (0.3 RMB/page) regardless of file type.
  • Billing occurs only when the extraction result is returned successfully; failed requests, failed extractions, or requests without any fields to extract are not billed.
  • When the account balance is insufficient, the API returns error code 1005025075 (insufficient quota or billing check failed).
  • See Product Pricing for details.

File Results and Translation Enhancements

  • Added /api/app-api/sip/platform/v2/file/translate to translate recognized regular fields, tables, stamps, and handwriting by task. Languages use ISO 639-1 codes; omitting source_language or sending an empty string enables automatic source-language detection.
  • Added the optional task_id query parameter to /api/app-api/sip/platform/v2/file/fetch for exact lookup using a string task ID.
  • Added the optional with_image_url parameter to /api/app-api/sip/platform/v2/file/fetch. When set to true, page data for regular, document-split, and image-crop tasks includes the corresponding image_url together with angle, width, and height. Image URLs are valid for 30 days.
  • When translations are available, /api/app-api/sip/platform/v2/file/fetch returns the corresponding translated_* fields in field, table, stamp, and handwriting objects. Existing response fields remain compatible.
  • Tokens returned by /api/app-api/sip/platform/v2/auth/token are valid for 1 day.

Category Keyword Rules

  • Create File Category and Update File Category now support category_keyword_rules. Positive rules match a category, while negative rules filter it out.
  • List File Categories now returns category_keyword_rules when configured.
  • A rule group matches when at least min_hit keywords match. Positive groups are combined with OR, and a matching negative group excludes the category. Each category supports up to 10 positive groups and 1 negative group, with up to 20 keywords per group.
  • On creation, omit the parameter or pass two empty arrays to leave rules unconfigured. On update, omit it or pass null for no change, pass two empty arrays to clear the rules, or pass another valid object to replace all rules.
  • Keywords preserve their original case and whitespace. Duplicates within a group, positive/negative conflicts, and cross-category same-type rule-group conflicts are compared using the original keyword strings; whitespace is no longer removed and case is no longer normalized.

SDK

  • Python SDK 1.1.2 adds category keyword-rule support for create, update, response parsing, and conflict prechecks. Keywords are validated and sent unchanged.

2026-07-08

API

Intelligent Review Model Update

  • Submit Review Task
    • Body: The available values of the model parameter are updated to deepseek-v4-pro, qwen3.6-plus, qwen3.7-max, and ORM-O1. Starting July 8, 2026, the following legacy models were deprecated. Use the corresponding replacement model:
    • Response: Added the model field, which returns the model actually used for the review task.
  • Get Review Task Result statistics now include not_triggered_count, which returns the number of review rules that were not triggered.

2026-07-05

API

Extraction Model Upgrade

The extraction models have been uniformly renamed, and a new Auto intelligent routing option has been added. This affects all endpoints that contain extract_model (Create File Category, Update File Category, adding and updating file category fields/tables, Extract Fields, Get File Results).
  • The request parameter extract_model only accepts the new model names (Auto/Acgpt/Acgpt-VL/DF-M1); the old values Model 1/Model 2/Model 3 and the legacy code values llm/vlm are no longer supported.
  • In the Get File Results and Extract Fields responses, fields and tables now include configModel (the configured model, which may be Auto; null for legacy data that was not configured); the original extractModel semantics have been adjusted to “actual hit model” (values Acgpt/Acgpt-VL/DF-M1).
  • A new extraction failure reason “Exceeded Acgpt-VL processing limit” has been added (Acgpt-VL has a document-level processing cap, surfaced when e.g. the page count exceeds the limit).
  • The extraction model Acgpt-VL does not currently support table multi-table merging (collect_from_multi_table).

2026-06-02

API

Primary Endpoints Upgrade

Process multiple objects in a single request. All endpoints use all-or-nothing semantics — if any item fails validation, the entire batch is rejected.

Deprecated Endpoints

The following single-item endpoints are now marked as deprecated. They remain functional but will no longer be maintained. Please migrate to the primary endpoints above:

One-Step Category Creation

Create File Category now accepts a tables parameter — configure tables and their fields in a single call when creating a category.

Field Advanced Config

Update File Category Field (single + batch) now support these properties:

Enriched Responses

Get full config without extra queries:
  • Get File Category Field List — added enabled, prompt, use_prompt, extract_model, alias, identity, multi_value, duplicate_value_distinct, transform_settings
  • Get File Category Table List — added prompt, collect_from_multi_table, extract_model
  • List File Categories — added category_prompt, extract_model
  • with_detail flag — pass with_detail=true on create/update to get the complete object in the response

SDK

  • Python SDK fields.update() and fields.batch_update() — added advanced config parameters and clear semantics examples

2026-05-27

API

2026-05-06

Guides

2026-04-30

Guides

2026-04-16

API

2026-04-15

API

2026-04-14

Guides API

2026-04-12

Guides

2026-04-08

SDK
  • Added: Python SDK documentation

2026-04-07

Guides
  • Updated: All 6 scenario guides (Expense Reimbursement, AP Review, Contract Review and their pre-configured versions) with improved titles, descriptions, and sample file references
  • Added: Java example code (ApReview.java, ContractReview.java, etc. — 6 files) and corresponding sample documents

2026-04-03

Guides API

2026-04-01

API

2026-03-24

Guides

2026-03-12

Guides API

2026-03-02

API

2026-02-03

Guides

2026-01-28

API

2026-01-27

Guides

2026-01-21

API

2025-12-25

Guides API

2025-12-24

Guides API

2025-12-19

Guides API

2025-12-17

Guides API

2025-12-11

Guides

2025-12-08

Guides API

2025-11-12

API

2025-11-04

Guides API

2025-10-27

Guides API

2025-10-16

Guides API

2025-09-01 ~ 2025-09-05

Guides API
  • Added: English OpenAPI specification

2025-08-26 ~ 2025-08-28

Guides

2025-08-14

Guides

2025-08-13

Guides API

2025-07-09

API
  • Fixed: Table and stamp result field definitions

2025-05-20 ~ 2025-05-21

API
  • Updated: Upload File now supports split_flag (file split) and crop_flag (multi-image crop) parameters
  • Updated: Upload File now supports auto_verify_vat (automatic VAT invoice verification) parameter
  • Updated: Upload File now supports target_process parameter for pipeline control (classify-only / classify+extract)
  • Updated: Get File Processing Result List response now supports sub-tasks and sub-image coordinates

2025-05-07 ~ 2025-05-12

API

2025-04-15

API

2025-02-18

API

2025-01-02 ~ 2025-01-03

API

2024-12-20 ~ 2024-12-23

API
  • DocFlow API initial release
  • Upload File endpoint uses multipart/form-data format with multi-file upload support
  • Authentication signature examples (Python)