Skip to main content
Ground image answers in both OCR and model interpretation:
Arka reports what OCR says, what the visual/vLLM model says, and a combined answer that identifies agreement, conflicts, and uncertainty. It does not silently treat a model guess as text evidence. Natural language requests are supported.

Related topics

Image description, OCR, and screen capturevLLM fallbackVideo evidence for dev workArka — AI terminal agent documentationTesting Arka with the pytest suite