PlatPhorm ASCII final polish adds perceptual math, compact color, and better deASCII handoff

The final polish upgrades the converter where quality actually starts: luminance math, local detail, and output payload discipline. The server conversion path now uses linear-light perceptual luminance, local edge-aware tone mapping, and serpentine error diffusion before characters are selected. That produces clearer lines and more readable detail without changing the public response contract.

Better math, same practical workflow

The new perceptual_edge_serpentine_v1 model maps source pixels through a perceptual luminance curve, boosts meaningful local contrast, and diffuses quantization error in alternating scan directions. The compact rgb24-b64-v1 color grid remains the preferred color payload, so deASCII receives exact source color without forcing giant JSON color arrays into every response.

Handoff quality matters downstream

ASCII is now tuned for deASCII reconstruction rather than plain terminal display alone. The handoff keeps line breaks, compact chroma, source dimensions, and quality metadata intact. deASCII can then use glyph masks, Unicode Braille dot-cell reconstruction, and restoration-only AI enhancement when a pasted source needs readability instead of reinvention.

What stays unchanged

Browser conversion, copy, download, Markdown export, PNG export, REST, MCP, OpenAPI, LLMS files, feeds, sitemap, trust metadata, and protected action boundaries remain in place. The converter stays an image to ASCII tool; the platform layer makes the output more discoverable, testable, and reusable.