Skip to content
Documentation
Cookbook
Multimodal: Vision, Charts, and Transcription
Best practices for vision

Best practices for vision

Prompting patterns for more reliable multimodal outcomes (visual prompting, few-shot, multiple images, etc.).

  • Upstream notebook: multimodal/best_practices_for_vision.ipynb

What to focus on

  • Few-shot examples for consistent extraction
  • Multiple images as input (comparisons, sequences)
  • “Ask for what you need” (structure, constraints, refusal modes)

Run locally 

make test-notebooks NOTEBOOK=multimodal/best_practices_for_vision.ipynb
Getting started with visionCharts, graphs, and slide decks