Full deployment history for this project.
Merge pull request #3 from mbarbine/bolt-coord-precalc-14380507663043584904 ⚡ Bolt: Precalculate pixel-cell coordinate grids in deASCII reconstruction
⚡ Bolt: Precalculate pixel-cell coordinate grids in deASCII reconstruction Optimized the reconstruction hot loops inside lib/deascii.ts by precalculating row and column coordinate mapping, fractional offsets, and mask indices outside the double-nested rendering loop. Previously, these geometry and scaling values were recalculated for every single pixel ($O(W \times H)$ complexity), leading to millions of redundant Math.floor, division, multiplication, and clamping calls. By computing them once in $O(W + H)$ arrays (TypedArrays) outside of the pixel loops, we avoid redundant CPU math and branching on every pixel and significantly speed up image reconstruction. All tests are passing and typechecks are clean. Co-authored-by: mbarbine <3211492+mbarbine@users.noreply.github.com>
Merge pull request #2 from mbarbine/jules-14942818252241281163-ecd2c7d5 🎨 Palette: UX and accessibility enhancement for quality presets and forms
palette: UX and accessibility enhancement for quality presets and forms - Derived and visually highlighted active quality presets (Photo, Crisp, Text, Soft) - Added 'aria-pressed' attributes for screen reader awareness of active presets - Associated sliders and select dropdown labels with matching htmlFor and element IDs for screen-reader and keyboard accessibility - Ensured zero cascading renders by computing activePreset state dynamically - Verified via Playwright screenshots, TypeScript compilation, and ESLint checks Co-authored-by: mbarbine <3211492+mbarbine@users.noreply.github.com>
Merge pull request #1 from mbarbine/bolt-optimize-bilinear-interpolation-7560987143831203076 ⚡ Bolt: Optimize bilinear interpolation in deASCII reconstruction engine
⚡ Bolt: Optimize bilinear interpolation in deASCII reconstruction engine This commit optimizes the deASCII reconstruction engine in `lib/deascii.ts` by: 1. Inlining and combining the bilinear grid sampling parameters and calculation steps (replacing separate `sampleLumGrid` and `sampleRgbGrid` calls with a unified inline computation). 2. Avoiding millions of redundant `clamp` and `clamp01` function calls within the hot inner rendering loop by using direct comparison/assignment operators. 3. Reusing the precomputed grid coordinates and interpolation weights (`w00`, `w10`, `w01`, `w11`) for both the luminance and source color grids. These improvements make the reconstruction loop measurably faster and yield a rendering performance speedup of up to ~22% for hybrid/tonal modes. Co-authored-by: mbarbine <3211492+mbarbine@users.noreply.github.com>
fix: keep MCP tool identity on the tool span
feat: emit canonical deASCII execution traces