Image Understanding

Output Schema by Type

Pet Analysis:

{
  "species": "dog",
  "breed_guess": "golden retriever",
  "physical_features": {
    "coat_color": "golden",
    "eye_color": "brown",
    "distinctive_features": ["fluffy ears"]
  },
  "personality_traits": ["playful", "friendly"],
  "suggested_character_archetype": "loyal companion"
}

Person Analysis:

{
  "hair": {"color": "black", "style": "short"},
  "fashion_aesthetic": "casual modern",
  "expression_mood": "confident",
  "suggested_anime_traits": ["protagonist energy"],
  "suggested_character_archetype": "determined hero"
}

World Analysis:

{
  "setting_type": "cyberpunk city",
  "visual_style": {
    "color_palette": "neon and dark",
    "lighting": "dramatic",
    "evoked_style": "blade runner"
  },
  "atmosphere": "mysterious",
  "mood_keywords": ["futuristic", "lonely", "electric"]
}

Implementation Contract

class ImageUnderstanding:
    async def execute(
        self,
        image_path: Path,
        analysis_type: str = "auto"
    ) -> dict:
        """
        Analyze an image and return structured information.

        Raises:
            FileNotFoundError: If image_path doesn't exist
            ValueError: If analysis_type is invalid
            APIError: If Gemini API call fails
        """
        ...

Example Usage

from skills.understand_image import ImageUnderstanding

skill = ImageUnderstanding()

# Auto-detect type
result = await skill.execute(image_path=Path("photo.jpg"))

# Specify type for better results
result = await skill.execute(
    image_path=Path("my_dog.jpg"),
    analysis_type="pet"
)

Dependencies

• Gemini 3 Flash/Pro API (multimodal)
• Pillow (image loading)

Error Handling