# Claude Vision Auto Configuration

# Ollama Settings
ollama:
  url: "http://localhost:11434/api/generate"
  model: "minicpm-v:latest"  # Options: minicpm-v:latest, llama3.2-vision:latest, llava:latest
  timeout: 30

# Timing Settings
timing:
  idle_threshold: 3.0      # Seconds of no output before taking screenshot
  response_delay: 1.0      # Seconds to wait before sending response
  screenshot_timeout: 5    # Screenshot capture timeout

# Vision Analysis Prompt
vision_prompt: |
  You are analyzing a terminal screenshot showing a Claude Code approval prompt.

  Look for:
  - Numbered menu options like "1. Yes", "2. Yes, allow all", "3. No"
  - Questions asking for approval (create/edit/delete files)
  - Yes/No questions with (y/n) or [y/n] format

  RESPONSE RULES:
  - If you see numbered options with "Yes" as option 1: respond with ONLY "1"
  - If you see a yes/no question with (y/n) or [y/n]: respond with ONLY "y"
  - If you don't see any prompt requiring input: respond with ONLY "WAIT"
  - NEVER provide explanations, ONLY the single response character/number

  Your response (one character/number only):

# Response Mapping
# What the vision model should output for different scenarios
response_mapping:
  approval_prompt: "1"      # Response for numbered "Yes" option
  yes_no_question: "y"      # Response for y/n questions
  no_action: "WAIT"         # When no input is needed

# Approval Detection Keywords
# Keywords that indicate we might be waiting for approval
approval_keywords:
  - "Yes"
  - "No"
  - "(y/n)"
  - "[y/n]"
  - "Approve"
  - "Do you want to"
  - "create"
  - "edit"
  - "delete"
  - "Allow"

# Buffer Settings
buffer:
  size: 4096  # Output buffer size in bytes

# Debug Mode
debug: false  # Set to true for verbose logging

# Screenshot Settings
screenshot:
  tools:
    - "scrot"
    - "gnome-screenshot"
    - "import"
    - "maim"
  cache_cleanup_seconds: 3600  # Clean up screenshots older than 1 hour