Schemas

Binary files to upload. File Blobs. Not 1s and 0s, but actual files

Display names for each uploaded file, parallel to the files array.

JSON object with customer reference data for validation columns

Binary files to upload. Every file referenced by any manifest item must be included here.

JSON array of upload groups, one item per person or group. Item fields: files (required, list of uploaded filenames or zero-based indexes into files[]), collection_name or collection_id (optional, mutually exclusive; several files, or naming a collection, make the item a collection), validation_data (optional object with reference data for that item), file_names (optional display names parallel to the item files). An item with more than one file must set collection_name or collection_id. validation_data is capped at 50 keys and 4096 bytes. Omit validation_data to leave the reference data on an existing collection untouched. Sending it with a collection_id replaces the stored object wholesale, so send the complete object rather than a partial update. Every uploaded file must be referenced by exactly one item; use indexes when filenames repeat or when a file was uploaded without a name. Example: [{"files": ["claim1.pdf", "claim1-photos.pdf"], "collection_name": "Claim 1001", "validation_data": {"policy_number": "POL-1"}}, {"files": ["claim2.pdf"], "validation_data": {"policy_number": "POL-2"}}]

Start extraction automatically once every document in this batch has converted or failed conversion. A document counts as ready either way, so one bad file cannot stall the batch; documents from multiple uploads that become ready together are extracted in a single run. If the account has insufficient credits, extraction does not start and the reason is recorded on the affected documents' error_message.

URL of the source document

Extracted value; shape depends on column type

Extraction cell status

Total number of extraction results

Current page number

Items per page

Specific documents to extract. Should be used for filtering

Collection ID to extract. Should be used for filtering

Specific columns to extract. Should be used for filtering

Force re-extraction of already extracted columns

Document labels in desired merge order.

File IDs to confirm

Add documents to existing collection

Create new collection and add documents

Customer reference JSON to validate extractions against

File IDs to confirm, from the upload-url response

Add documents to an existing collection

Create a new collection and add documents to it

Customer reference data to validate extractions for this item against

One entry per person or group being confirmed

Start extraction automatically once every document in this batch has converted or failed conversion. A document counts as ready either way, so one bad file cannot stall the batch; documents from multiple uploads that become ready together are extracted in a single run. If the account has insufficient credits, extraction does not start and the reason is recorded on the affected documents' error_message.

test → alk_test.* token; live → alk_live.* (default)

Restrict key to these workflow IDs; omit for all entity workflows

Override default RPM cap

Operation status

Deleted workflow ID

Deleted column ID

Deleted document ID

Full Bearer token (shown once). Format: {alk_live|alk_test}..

Original filename

MIME type of the file

File size in bytes

Document processing pipeline status

Uploaded documents for this item. Empty when error is set.

Failure reason for this item. When set, documents is empty and any files uploaded for this item were rolled back.

Documents created across all items; an item with three files adds three

Number of items that failed; see error on each item

Total count

Current page

Items per page

Total pages

List of items

Name of the source document

Page number where the value was found

Source documents for collection extractions

Display name of the column

Extraction result status

Signed URL for the source document file

Whether the extracted value matches the customer reference data

Extracted value

Confidence level of the extraction

LLM reasoning

File record ID

GCS v4 signed PUT URL

Sanitized filename

File metadata for each upload

Signed URL items

Column name

Column data type

Natural language extraction prompt

Doc generation settings when task_type is generate

How option fields handle values outside their declared options

Value to use when no option matches and default_value is selected

Display order

Column width in UI

Tag options [{id, label, color, stageId}]

List of column IDs this column depends on

Task type for this column

Model ID for this column

Sub-field definitions for table columns [{name, type, prompt}]

Workflow name

Application ID (e.g., 'deal-screening')

LLM model (empty = use provider default)

LLM provider name

Workflow description

Extensible metadata

Workflow lifecycle status

Overall extraction status: not_started, pending, complete, has_failures

Documents whose extraction has fully resolved, including ones whose cells failed

Documents in the workflow

Document name

MIME type of document content

Signed GCS GET URL for document content

TTL for content_url in seconds

ID-enriched LLM context

First 1000 characters of the LLM context, for thumbnails

Per-page dimensions the segment bboxes are expressed in

Document segments with bboxes

Customer reference data for validation, stored on standalone documents

Associated file ID

Current workflow stage

Is this a collection

Parent document ID for collections

Collection ID this document belongs to

Order within collection

User-defined label for the document

Document processing pipeline status

File URL from associated file

Stage ID (e.g. 'stage_approved')

Display label

System stage cannot be deleted

Stage color (e.g. 'emerald', 'blue')

How many approvers need to sign off before a stage's gate is satisfied.

ANY: one approval advances the doc. ALL: every approver listed on the stage must approve.