Which tier produced this result.
Provider name within the tier (e.g. 'paddle', 'tesseract', 'openai').
For the cloud tier, the provider that answered: cloudProvider, or a
cloudFallbackProviders entry after it failed.
Optional modelFor the cloud tier, the model that answered, as generateText reports it.
Extracted or generated text content.
Confidence score from 0 (no confidence) to 1 (certain).
Wall-clock processing time in milliseconds.
Optional regionsBounding boxes for detected text regions, when available.
Result from a single processing tier.
Each tier that runs produces one of these, regardless of whether the pipeline ultimately uses its output or prefers a higher-confidence alternative from another tier.