ai_renderer.py

from image_backend import ImageBackend

class AIRenderer:
    """
    Converts structured scene graphs into:
    - image prompts
    - animation frame descriptions
    - future diffusion model inputs
    """

    def __init__(self, llm=None):

        self.llm = llm
        self.backend = ImageBackend(mode="ollama")

    # -----------------------------
    def render_frame(self, scene, frame_index):

        objects = scene.get("objects", [])

        description = "\n".join([
            f"{o['id']} pos={o['position']} vel={o['velocity']}"
            for o in objects
        ])

        prompt = f"""
Generate visual frame description for rendering pipeline.

Scene:
{description}

Return structured visual prompt.
"""

        result = self.backend.generate(prompt, output_path=f"frames/frame_{frame_index:04d}")

        return result

    # =====================================================
    # SCENE → ANIMATION FRAME DESCRIPTION
    # =====================================================
    def render_animation_frame(self, scene, next_scene=None):

        prompt = f"""
Describe a single animation frame transition.

Current scene:
{scene}

Next scene:
{next_scene}

Focus on:
- motion changes
- object displacement
- physics behavior
"""

        if self.llm:
            return self.llm.call(prompt)

        return prompt