gemini-vision

Image analysis with Gemini Vision generates descriptive captions and classifications by sending images to Google's Gemini API. Provide an image file and a query; the Skill returns text captions, structured metadata, and content classifications for automated tagging and organization.

Yes, multi-image analysis supports up to 3,600 images per request, enabling trend detection and change comparison across image sets. The Skill processes batches and returns aggregated insights alongside individual image results.

Document understanding for PDFs processes files up to 1,000 pages, extracting text, layout structure, and semantic content. This automates document analysis and information retrieval workflows without manual transcription.

You need a valid Gemini API key or Vertex AI configuration to authenticate requests. The Skill supports model selection across gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite, gemini-2.0-flash, and related variants depending on your setup.

Object detection and segmentation identify objects and return bounding boxes plus pixel-level masks for precise spatial understanding. This enables scene analysis, inventory tracking, and automated visual inspection at object granularity.

Images and PDFs must stay within defined size limits enforced by the Gemini API. Check your Gemini or Vertex AI documentation for exact thresholds; the Skill validates inputs before processing and returns structured metadata on processing success.