Schemas

Unique model identifier. Use this value in job requests.

Human-readable model name

Model description and recommended use case

Model provider

Unique file identifier. Use this value in job requests.

Original filename

File size in bytes

Files are automatically deleted after a period of time

The filename of the file to upload

Base64-encoded chunk of the file. Each chunk must not exceed 5MB of the original (pre-encoded) data.

Total number of input characters submitted

Number of input characters actually processed

Number of input characters skipped (e.g. numeric segments)

Total number of output characters generated

Vector data bytes (reserved for future use)

Number of pages in the input PDF. Present for jobs whose input is a PDF file.

Unique job identifier. Use this to poll for results.

Model ID obtained from GET /structflow/models

Instruction describing what structured data to extract and how

A JSON object showing the desired output structure with example values

List of input records to process. Mutually exclusive with 'file_id'.

File ID of a pre-uploaded JSONL file (one JSON record per line), obtained from POST /files. Mutually exclusive with 'inputs'. When used, results are returned as an 'output_file_id' instead of inline 'results'.

File ID of the output JSONL file. Present only for jobs submitted with 'file_id' input and status is completed.

Result summary. Present only when status is completed AND the job was submitted with 'inputs' (inline).

Extracted data per input record. Present only when status is completed AND the job was submitted with 'inputs' (inline). Each record may independently succeed or fail. For jobs submitted with 'file_id', use 'output_file_id' to download the results as a JSONL file.

Job-level error. Present only when job status is failed.

File ID of the output JSONL file. Present only when status is completed AND the job was submitted with 'file_id' input. Each line contains either the extracted JSON for a successful record, or an object with a reserved '$error' key (e.g. {"$error":{"code":"...","message":"..."}}) for a failed record. Download via GET /files/{output_file_id}/content.

Model ID obtained from GET /refineloop/models

File ID of the XLIFF file to process, obtained from POST /files

If true, segments containing only numbers are skipped

If true, hyphenation in source text is removed before processing

Maximum number of revision iterations per segment (1-10)

Subject domain to guide translation review (e.g. "IR", "medical", "legal")

Language code for revision notes written by the AI (e.g. "ja", "en", "ko"). Defaults to the source language of the XLIFF file if not specified.

Controls what is included in the output XLIFF

Additional instructions appended to the AI's base prompt

Input file ID

Input file ID

File ID of the refined XLIFF output. Present only when status is completed. Download via GET /files/{output_file_id}/content.

Job-level error. Present only when job status is failed.

Unique engine identifier. Use this value in job requests.

Human-readable engine name

Engine provider

Engine description and recommended use case

List of supported input/output format combinations

List of supported OCR languages

Engine ID obtained from GET /renderocr/engines

File ID of the input file to process, obtained from POST /files

List of language codes for OCR recognition. Currently only the first language is used.

Output document format

Input file ID

Input file ID

File ID of the converted document output. Present only when status is completed. Download via GET /files/{output_file_id}/content.

Job-level error. Present only when job status is failed.

Engine ID obtained from GET /castdoc/engines

File ID of the input PDF file to process, obtained from POST /files

Output document format

Input file ID

Input file ID

File ID of the converted document output. Present only when status is completed. Download via GET /files/{output_file_id}/content.

Job-level error. Present only when job status is failed.

Engine ID obtained from GET /extractdoc/engines

File ID of the input file to process, obtained from POST /files. Supported input formats: PDF, DOCX, XLSX, PPTX. The input format is automatically detected from the uploaded file.

Output format. Use 'text' for plain text, or 'jsonl' for a single-line JSON object compatible with StructFlow input.

Input file ID

Input file ID

File ID of the extracted output. Present only when status is completed. Download via GET /files/{output_file_id}/content.

Job-level error. Present only when job status is failed.