createImageDocumentParser
@forge/monorepo / backend/src / createImageDocumentParser
Function: createImageDocumentParser()
createImageDocumentParser(
deps):DocumentParser
Defined in: backend/src/documents/vision.ts:173
The image parser: OCR when there is text to transcribe, vision to describe, or both.
Both is the useful default for a screenshot. A screenshot of a dashboard has text and a chart, and either alone loses half of it — the numbers without the layout, or the layout without the numbers.
Parameters
deps
ocr?
onVisionUsage?
(usage) => void
Observed for tests and hosts that want the raw result.
Not the billing path. Billing goes through the returned document's usage, because only the pipeline
knows which file a parse belonged to — a callback held by the parser is shared across every file it ever
parses, so correlating one would mean trusting that parses never interleave.