Skip to main content

Interface: OCRResult

Defined in: packages/agentos/src/api/performOCR.ts:106

Result returned by performOCR.

Properties​

category?​

optional category: string

Defined in: packages/agentos/src/api/performOCR.ts:124

Content category detected by the pipeline (e.g. 'printed-text').

confidence​

confidence: number

Defined in: packages/agentos/src/api/performOCR.ts:111

Overall confidence score (0 = no confidence, 1 = certain).

provider​

provider: string

Defined in: packages/agentos/src/api/performOCR.ts:127

Provider name that produced the winning result (e.g. 'paddle', 'openai').

regions?​

optional regions: object[]

Defined in: packages/agentos/src/api/performOCR.ts:133

Text regions with bounding boxes, when the winning tier provides spatial information. Not all tiers return region data.

bbox?​

optional bbox: object

bbox.height​

height: number

bbox.width​

width: number

bbox.x​

x: number

bbox.y​

y: number

confidence​

confidence: number

text​

text: string

text​

text: string

Defined in: packages/agentos/src/api/performOCR.ts:108

Extracted text content.

tier​

tier: VisionTier

Defined in: packages/agentos/src/api/performOCR.ts:121

Which processing tier produced the winning result.

  • 'ocr' — PaddleOCR or Tesseract.js (fast, local, free).
  • 'handwriting' — TrOCR handwriting recognition (local).
  • 'document-ai' — Florence-2 document understanding (local).
  • 'cloud-vision' — Cloud LLM (GPT-4o, Claude, Gemini).