Find a workspace by name, or create one if it does not exist.

Returns: workspace_id (str): ID of the workspace. created (bool): True if a new workspace was created.

Use this as Step 1 of every workflow to ensure a workspace is ready.

List all enabled categories in a workspace, including their field configurations.

Returns a list of category objects with name, category_id, and fields. Use this to check which categories are already configured before creating new ones.

Create a file category with field configuration and a sample file.

Upload files and wait for classification and extraction to complete.

Provide either file_paths (explicit list) or directory (auto-scans supported files). Supported formats: PDF, JPG, PNG, DOC, DOCX, XLS, XLSX, OFD, TXT.

Create a review rule repository (idempotent — reuses existing repo if name matches).

Submit a review task and wait for results.

Query processed files in a workspace (GET /file/fetch).

Useful for checking status of previously uploaded files or retrieving results later. recognition_status: "0"=pending, "1"=success, "2"=failed, "3"=processing.

Update metadata or verification status of a processed file (POST /file/update).

Delete processed files from a workspace (POST /file/delete). Provide either task_ids or batch_numbers. Deletion is permanent.

Re-extract specific fields from an already-processed file (POST /file/extract_fields).

Use when you need to extract additional fields without re-uploading the file, or when the original extraction missed certain fields.

Retry processing for files that failed (POST /file/retry). Use when recognition_status=2 (failed).

Correct a misclassified file's category and trigger re-extraction (POST /file/amend_category).

Use when a file was classified into the wrong category. The file will be re-processed using the correct category's field configuration.

List all workspaces accessible to the current account (GET /workspace/list). Returns workspace_id, name, description, and auth_scope for each.

Get detailed information about a specific workspace (GET /workspace/get). Returns workspace metadata including name, description, auth_scope, and statistics.

Update workspace settings (POST /workspace/update).

Delete a workspace and all its contents (POST /workspace/delete). WARNING: This permanently deletes all files, categories, and review rules in the workspace.

Update an existing file category's settings (POST /category/update).

Delete a file category and its field configuration (POST /category/delete). WARNING: Existing files classified under this category retain their extracted data, but future uploads will not match this category.

List all table configurations in a category (GET /category/tables/list).

Returns table_id, name, and fields for each table. Use this to get table_id before adding table fields or building review rules.

Add a table field group to a category (POST /category/tables/add).

Use for extracting structured table data (e.g. invoice line items, hotel charges).

Update a table configuration in a category (POST /category/tables/update).

Delete table configurations from a category (POST /category/tables/delete).

List all fields configured in a category (GET /category/fields/list).

Returns field_id, name, description, and configuration for each field. Use this to get field_ids needed when building review rules with referenced_fields.

Add fields to an existing category (POST /category/fields/add).

Update a field's name, description, or extraction prompt (POST /category/fields/update).

Updating the prompt can improve extraction accuracy for specific field types.

Delete fields from a category (POST /category/fields/delete).

Add sample files to a category to improve classification accuracy (POST /category/sample/upload).

More samples (3–5 recommended, max 10 total) improve classification reliability.

List all sample files configured for a category (GET /category/sample/list). Returns sample_id, file name, and upload time for each sample.

Delete sample files from a category (POST /category/sample/delete). Use to remove low-quality or irrelevant samples that hurt classification accuracy.

List all review rule repositories in a workspace (GET /review/rule_repo/list). Returns repo_id, name, and creation time for each.

Get a review rule repository with all its groups and rules (GET /review/rule_repo/get). Use this to inspect existing rule configurations before making changes.

Rename a review rule repository (POST /review/rule_repo/update).

Delete a review rule repository and all its groups and rules (POST /review/rule_repo/delete). WARNING: This is permanent. Existing review task results are not affected.

Rename a review rule group (POST /review/rule_group/update).

Delete a review rule group and all rules within it (POST /review/rule_group/delete).

Update an existing review rule's prompt or configuration (POST /review/rule/update).

Use to refine rule prompts based on review results, or adjust risk levels. risk_level: 10=high, 20=medium, 30=low.

Delete a review rule (POST /review/rule/delete).

Get the current result of a review task without waiting (POST /review/task/result).

Use to check status of a previously submitted review, or for manual polling. status: 0=pending, 1=pass, 2=failed, 3=reviewing, 4=not_pass, 7=recognition_failed.

Delete review tasks (POST /review/task/delete).

Retry an entire review task that failed or produced unexpected results (POST /review/task/retry). All rules in the task will be re-evaluated.

Retry evaluation of a single rule within a review task (POST /review/task/rule/retry).

Use when one specific rule failed or needs re-evaluation after updating its prompt, without re-running all other rules in the task.