Constellation Composition MCP Server

Instructions

Implementation Reference

• src/constellation_composition_mcp/server.py:1856-1954 (handler)

  The handler function 'generate_constellation_attractor_prompt' which takes AttractorPromptInput to generate image prompts. It uses visual vocabulary types and parameter coordinates.

  async def generate_constellation_attractor_prompt(params: AttractorPromptInput) -> str:
      """
      Generate image-generation-ready prompts from 5D constellation coordinates.
  
      Layer 2: Deterministic vocabulary extraction (0 tokens).
  
      Translates abstract parameter coordinates into concrete visual vocabulary
      suitable for Stable Diffusion, ComfyUI, DALL-E, or Midjourney. Three
      output modes are available:
  
      - **composite**: Single prompt string combining keywords and geometric specs.
        Best for direct image generation.
      - **split_keywords**: Categorized keyword lists (visual type, specifications,
        parameter descriptors). Best for ComfyUI prompt engineering.
      - **descriptive**: Narrative paragraph prompt. Best for DALL-E / Midjourney.
  
      Coordinates can be provided directly, or derived from a canonical state name.
      """
      # Resolve coordinates
      if params.coordinates:
          coords = params.coordinates
          # Fill missing params with 0.5 default
          for p in CONSTELLATION_PARAMETER_NAMES:
              if p not in coords:
                  coords[p] = 0.5
      elif params.canonical_state:
          state = CONSTELLATION_CANONICAL_STATES[params.canonical_state]
          coords = {p: state[p] for p in CONSTELLATION_PARAMETER_NAMES}
      else:
          return json.dumps({
              "error": "Provide either 'coordinates' dict or 'canonical_state' name"
          })
  
      vocab = _extract_visual_vocabulary(coords, params.strength)
  
      if params.mode == "composite":
          prompt = _generate_composite_prompt(coords, params.strength)
          return json.dumps({
              "mode": "composite",
              "prompt": prompt,
              "nearest_visual_type": vocab["nearest_type"],
              "type_distance": vocab["distance"],
              "coordinates": coords
          }, indent=2)
  
      elif params.mode == "split_keywords":
          # Build categorized keyword sets
          specs = {}
          sd = coords.get("stellar_density", 0.5)
          specs["density"] = (
              "dense rich starfield" if sd > 0.7
              else "sparse isolated stars" if sd < 0.3
              else "moderate star density"
          )
          gr = coords.get("geometric_regularity", 0.5)
          specs["geometry"] = (
              "precise geometric arrangement" if gr > 0.7
              else "organic flowing forms" if gr < 0.3
              else "naturally balanced arrangement"
          )
          lc = coords.get("luminance_contrast", 0.5)
          specs["luminance"] = (
              "extreme brightness contrast" if lc > 0.7
              else "soft even luminosity" if lc < 0.3
              else "moderate contrast"
          )
          ni = coords.get("narrative_intensity", 0.5)
          specs["mood"] = (
              "dramatic mythological atmosphere" if ni > 0.7
              else "contemplative quiet mood" if ni < 0.3
              else "balanced narrative tone"
          )
          se = coords.get("spatial_extent", 0.5)
          specs["scale"] = (
              "wide panoramic field" if se > 0.7
              else "intimate compact framing" if se < 0.3
              else "moderate field of view"
          )
  
          return json.dumps({
              "mode": "split_keywords",
              "visual_type_keywords": vocab["keywords"],
              "parameter_descriptors": specs,
              "nearest_visual_type": vocab["nearest_type"],
              "type_distance": vocab["distance"],
              "coordinates": coords
          }, indent=2)
  
      else:  # descriptive
          prompt = _generate_descriptive_prompt(coords, params.strength)
          return json.dumps({
              "mode": "descriptive",
              "prompt": prompt,
              "nearest_visual_type": vocab["nearest_type"],
              "type_distance": vocab["distance"],
              "coordinates": coords
          }, indent=2)