ask.rb

# frozen_string_literal: true

# Visual Q&A — up to 5 questions about one file, priced like a single extraction.
#
#   ruby examples/ask.rb photo.jpg "Is there a dog in the image?" "How many people?"

require "vision_api"

abort("usage: ruby examples/ask.rb <file> [question…]") if ARGV.empty?

path, *questions = ARGV
vision = VisionAPI.new

res = vision.ask(file: path, questions: questions.empty? ? ["What is in this image?"] : questions)

res["answers"].each do |a|
  # `verdict` is the machine-readable half: yes | no | uncertain | n/a.
  # "uncertain" is a yes/no question the image does not settle — a real answer, not a
  # failure. "n/a" means it wasn't a yes/no question in the first place.
  puts "Q: #{a['question']}"
  puts "A: #{a['answer']}"
  puts "   verdict=#{a['verdict']} confidence=#{a['confidence']}\n\n"
end

puts "#{res['credits_used']} credit(s) — the questions themselves are free, " \
     "you pay per image or per selected page."