Recover the unpublished denoise work against current upstream. Preserve later opt-in model setup, shared Rust HTTPS downloads, pending-photo resume and an Amount-independent switch. Keep original RAWs unchanged and share the current color finish between ordinary and denoised sources.
Bound public tiling inputs, refuse corrupt-cache exports, and retain synthetic/real-model coverage. Include two Windows path spelling fixes in existing tests. No model weights, face code, media or Tract runtime are included.
Validation: cargo xtask ci (all seven gates); installed RawNIND CPU/GPU agreement 1.38e-6; native CPU preview/export on a public 24 MP Sony RAW, unchanged original SHA-256.
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Generated with Codex.
Decode lossless CRX and version 0x100/0x200 C-RAW in portable Rust,
including sensor geometry, timed white balance and black/white levels.
Verify all sensor samples against six independent M50/R100/R8 references.
Any CR3 decode error falls back to the embedded JPEG, as CR3 files did
before this decoder. The full sensor buffer is allocated only after the
first plane decodes, and plane buffers grow row by row. Tile planes
decode in parallel with a windowed bit reader (24 MP: 424 -> 89 ms
lossless, 519 -> 138 ms C-RAW on an M4).
Add guarded CR3 camera-JPEG colour fitting with independent framing
validation, relative white balance and cache invalidation.
Validation: cargo xtask ci.
Installed RAM overcommits a busy machine: on a 32 GB desktop with browsers,
a VM and editors open, about 7 GB was free and the installed-RAM rule chose
16 jobs. A fresh build then pushed free RAM below 1 GB.
`cargo xtask ci` now uses one build job per 1.5 GB of RAM available when it
starts (Linux MemAvailable, Windows AvailableMBytes, macOS free + inactive +
speculative pages), capped at the CPU count. With line-table debuginfo,
4 jobs measured about 3 GB in use, so this leaves about half the available
RAM spare. If available RAM can't be read, half the installed RAM is
assumed; if neither can, 4 jobs. Test threads follow, capped at 4.
Explicit CARGO_BUILD_JOBS / RUST_TEST_THREADS still win.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
#243 dropped every result whose request id wasn't the newest for its slot. The loupe requests a new
draft every frame while a slider moves, so when a draft took longer than a frame nothing landed
until the drag paused. Main, Before and Hover now take a superseded draft that is newer than what
they show (never one from before a library or preview-cache reset, never an error or its source).
Also saturate the xtask RAM rounding.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The fixed four jobs were safe on 8 GB but too many for 4 GB, and slower than
needed on large machines. `cargo xtask ci` now uses one build job per 2 GB of
installed RAM, capped at the CPU count, and 4 if the RAM can't be read
(/proc/meminfo, Win32_ComputerSystem, hw.memsize). With line-table debuginfo,
4 jobs measured about 3 GB of RAM in use and 2 jobs about 1.5 GB.
Test threads follow but stop at 4: more barely shortens CI, and at 16 threads
on a loaded machine the wall-clock frame-budget test
offline_originals_never_block_a_frame failed (one 303 ms frame). Explicit
CARGO_BUILD_JOBS / RUST_TEST_THREADS still win.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A full-debuginfo workspace test build with one rustc/link.exe per CPU peaked
at 14-20 GB commit on a 32-thread Windows machine before any test ran; the
test binaries themselves peak at about 3.3 GB each. `cargo xtask ci` now sets
CARGO_PROFILE_DEV_DEBUG=line-tables-only, CARGO_BUILD_JOBS=4 and
RUST_TEST_THREADS=4 unless they are already set. Ordinary cargo commands are
unchanged; GPU coverage is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Add pt-br as a fifth interface language beside English, Simplified
Chinese, Traditional Chinese (Taiwan) and Japanese.
`crates/ui-egui/locales/pt-br.json` and `pt-br-formats.json` carry the
messages and the format strings; the language table gains the entry
(`PtBr, "pt-br", "Português (Brasil)", "Latn", ...`), and Edit > Language
plus `app.language.portuguese` select it. Locale tags pt-BR, pt_BR.UTF-8
and pt all normalize to the same language.
Translation follows pt-BR UI conventions (Adobe terms: revelação,
balanço de branco, máscara, cópia virtual…); proper nouns and file
formats stay as written (LightCraft, ArtCraft, GitHub, DNG, TIFF, XMP).
Placeholders, format specs and punctuation match the English source; the
English plural suffix is consumed with `{:.0}` where Portuguese does not
need it.
Tests: every_language_has_a_catalog_covering_the_source,
catalogs_agree_on_placeholders_and_report_gaps, locale_tags_normalize
and the painted-window test now cover pt-br. Docs: localization.md, a
new localization-pt-br.md and the parity rows.
Co-Authored-By: DeepSeek V4 Flash (opencode) <noreply@opencode.ai>
The downloader in lightcraft-segment (mirrors, resume, pinned SHA-256, pure-Rust
HTTPS) was already generic apart from four SAM 3 constants. Move it to
crates/fetch (L0, native only) so other model features can use it without
depending on the SAM 3 inference code. lightcraft-segment keeps SAM3_FILES,
DEFAULT_MIRRORS, MIRRORS_ENV and wraps mirrors(); no behaviour change. The
downloader's 13 tests moved with it.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Decode every raw encoding Panasonic and Leica bodies write, from the DMC-LX1 (2005) to the DC-S1RM2 (2025):
compressed formats 4 (most bodies, 2008-2022) and 6 (S1 / S1R / S1H / S5, GH5M2), the prefix-coded strips of
format 8 (GH6 and later), packed formats 2, 5 and 7, and the 16-bit words of the 2005-2007 .RAW files. Until now
only format 5 decoded; every other Panasonic file opened as preview only.
Clean-room: no raw-decoder source was read. Tag names come from tag documentation; every encoding was recovered
by black-box analysis of 178 CC0 files from raw.pixls.us (118 bodies), and the module docs note the measurement
that settled each rule. Format-8 tags of unknown meaning must carry the values every sample has, otherwise the file
is reported as unsupported rather than guessed at.
Levels and geometry: formats 4 and 6 store samples 15 above the recorded black level (files without one use 15);
white is the active area's saturation plateau (the linearity limits are wrong on extended low-ISO shots, the GF1
and the Digilux 3); the in-camera aspect ratio is the default crop; mapped-out defect pixels are filled from
same-colour neighbours; the CFA pattern applies at sample (0, 0).
.rwl and .raw files are imported (engine, MCP folder expansion, file dialog, Linux MIME types). One file per
encoding joins the never-panic proptest; 11 CC0 files join `cargo xtask corpus`, and corpus_panasonic_encodings
checks each encoding's bit depth, black level, crop and CFA layout against the camera's own JPEG.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QtQmZ6peZXGn3wiqeowV2v
- the CJK fallback order follows the UI language's script, but the fonts were
only installed on the first frame: switching 日本語 ↔ 简体中文 at runtime kept
the previous language's Han forms. The app remembers the language its fonts
were built for and reinstalls them when it changes.
- the native menu bar rebuilds on any language change (keyed on the language
code, not on "uses CJK fonts", which is true for both CJK languages)
- restore the tests the language-table change dropped, for every language:
core catalog entries, user-named menu items stay untranslated, the window
paints each language and menu ids don't change with it, CJK glyphs in every
UI family, and the UI runs without craft-fonts
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- UI_COMMANDS still listed app.language.english/japanese next to the new
LANGUAGE_COMMANDS table, so Edit > Language showed English and 日本語 twice
- LANGUAGE_COMMANDS is `pub const`, the form xtask parity scans for; a test
reads the real menus.rs and checks every language id is found once
- the Simplified Chinese command is app.language.simplifiedChinese, in line
with the other camelCase command ids (the settings code stays zh-hans)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The localization layer was hardcoded to English and Japanese: an enum with two
variants, an `is_japanese()` gate compiled into the format macro, and one
catalog pair. Adding a language meant editing the menus, the settings row, the
checkmark logic, the format macro and the font stack — each in a different place.
A language is now one entry in `i18n::language_table!`: BCP-47 code, endonym,
ISO 15924 script and catalog. The Language menu, the settings row, the shortcut
sheet and the control channel are all built from that table, and `build.rs`
picks up `locales/<code>-formats.json` by name, so a further language is a table
entry plus two catalogs (docs/localization.md).
Simplified Chinese (`zh-hans`) ships as the second translation: `zh-hans.json`
(1019 entries, key-set and placeholder-identical to `ja.json`) and
`zh-hans-formats.json` (102 entries). Its glyphs come from craft-fonts' Noto
Sans CJK SC, whose `Hans` faces the UI now installs *before* the Japanese ones
for a Chinese reader, so shared Han characters keep the language's forms
(`fonts::cjk_fallback`); the export watermark accepts every CJK face.
`tr_format!` keeps its compile-time check. It formats the English source at the
call site — a runtime template cannot be handed to `format!` — and names those
values so a translation can place them by name (`{n}`) or position (`{_1}`).
Every translation is still checked by `format!`, so a translation with an
unknown placeholder fails the build. The Japanese catalog was rewritten from the
old `{:.0}` plural-discard trick to that by-name form; behaviour is unchanged.
Also: settings persist the BCP-47 code (never the Rust variant name),
`LIGHTCRAFT_LANGUAGE` accepts system-style tags (`zh_CN`, `ja_JP.UTF-8`),
`xtask parity` reads the second command table, and the glyph tests cover every
catalog character plus the interface symbols (`▸`/`▾` turned out to have no
glyph in any installed font and are gone from the Chinese text).
Chinese and Japanese both need `CRAFT_FONTS_DIR` as before; the web build still
embeds only the Japanese face (Noto Sans CJK SC alone would push the wasm past
Cloudflare's 25 MiB per-file cap), documented in docs/localization-zh-hans.md.
Co-Authored-By: DeepSeek V3.2 <noreply@deepseek.com>
LightCraft has no colour matrix for Nikon cameras. Because of this, NEF
files showed colours that were too gray and too green.
- Colour: LightCraft now calculates a colour matrix and a tone curve from
the JPEG preview in each NEF file. Sony ARW files use the same
procedure and the same quality checks. White balance is relative to
the as-shot look, as for ARW files.
- One function, camera_preview::file_local_look, selects the formats for
the fit and for the relative white balance. Photo::relative_wb in the
catalog uses the same formats.
- Black level: Nikon stores the black level (maker note 0x003d) in
14-bit units, also in 12-bit files. The decoder read it as a 12-bit
value. Because of this, 12-bit files from the D750, D780, D850, D7500
and Z 50 were too dark. Now the decoder divides the value by 4 for
12-bit files.
- The corpus gets the D7500 12-bit and 14-bit lossless samples. A test
compares their levels. The render cache version increases to 9.
Results on 13 NEF files that LightCraft can decode: LightCraft accepted
the fit for all 13 files. The mean colour difference from the camera
JPEG (CIE76 dE) decreased from 13-46 to 2.9-7.8. One file with mixed
light improved by a small quantity only (12.7 to 12.0).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
AI masks from Meta's SAM 3, run in LightCraft itself: a new
`lightcraft-segment` crate ports the Hugging Face transformers
implementation (Apache-2.0) to candle, on the GPU through Metal on macOS
(CPU elsewhere), with no C/C++ and no `unsafe`.
- Object: click to include, ⌥-click to leave a part out (SAM 3's
interactive prompt encoder + two-way mask decoder).
- Describe: a text prompt selects every instance ("sky", "the red car";
"car, road" merges several) via the CLIP text encoder and DETR detector.
- Both as new masks and as Add / Subtract / Intersect components; + / −
buttons under the selected mask combine selections.
- The segmentation is stored with the mask (288² quantized logits, ~10-30
KB), so renders, exports and the web build never need the model.
- A zoomed-in detail pass (background, ~5 s on an M4 Pro) re-segments the
area around a selection from a sharper source: 5-10x finer edges on
small objects.
- Per-selection Edge slider (hard ↔ soft); hovering a mask in the list
shows it in red (the always-on overlay is now off by default, O toggles).
- The engine feature `sam` (on in the desktop app) gates the model; the
image is encoded once per photo in the background, clicks take ~35 ms.
- Weights are not bundled (SAM License): tools/install-sam3.sh / .ps1
download and verify facebook/sam3; docs/ai-masks.md.
- Clearer errors (missing file, missing model) and a stderr logger in the
app (LIGHTCRAFT_LOG=info).
candle is pinned at 0.9.2: later releases make candle-core depend on
tokenizers with the Oniguruma C library. Every stage is checked against
the Python reference (tests/reference.rs, tools/sam3_reference.py): on
Metal the backbone matches to 7e-5 relative error, masks to ~5e-5.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01C8BkS7JDqa1EMcjW6spDGA
Sony bodies from before ~2017 (RX100–RX100M5, RX10, NEX, SLT, ILCE-6000,
A7/A7II/A7RII, ...) write no plain 0x7313/0x7310/0x74c7 tags in the raw IFD,
so the decoder used unit WB multipliers (strong green cast, e.g. the CC0
DSC-RX100M3 sample), black 512 (800 on 1-inch sensors), and kept padding
columns at the right edge.
- White balance: WB_RGBLevels from the maker note's enciphered Tag2010
block (layouts per ExifTool's Tag2010a..i tables; byte substitution
p^3 mod 249 checked black-box on 15 CC0 files: SonyISO matches Exif ISO,
ILCE-7M3 levels match its plain 0x7313).
- Black level: recovered from the encrypted SR2SubIFD by known-plaintext
analysis (constant key in every file; RX100M6/ILCE-7M3 plain 0x7310 as
plaintext; 8 keystream bytes at offset 2510, plausibility-checked).
- Crop: DNG DefaultCrop or the maker note's FullImageSize.
- "12-bit uncompressed" ARWs hold 14-bit-scale data: black/white follow it.
- Render cache version bumped; corpus manifest + corpus test, unit tests.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Bayer layout of Canon CR2 files differs by model (GBRG on the 7D, 50D,
60D, 550D, 600D, 1200D, 1300D, 5D Mark II, 1D Mark IV; RGGB on the 40D,
5D Mark III, 6D, 5DS R and others). Read it from raw IFD tag 0xc5e0
(CR2CFAPattern: 1 RGGB, 2 BGGR, 3 GBRG, 4 GRBG), anchored at the full
decoded sensor; fall back to RGGB when the tag is missing. No model table.
Checked on 35 decoded CC0 raw.pixls.us CR2s (35 models): the tag agrees
with the mosaic's green-diagonal statistic on every file; the SensorInfo
border parity does not predict the layout. Corpus manifest gains the
reporter's samples; a corpus test asserts each layout.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The macOS malloc_zone_pressure_relief call (returns freed allocator pages
after raw decodes) lived in identical copies inside the desktop app and the
CLI. It now lives in lightcraft-sysmem, an isolated L0 helper crate and the
workspace's only crate allowed unsafe: one #[allow(unsafe_code)] module, a
SAFETY comment, a safe API that returns bytes released (0 off macOS), and
tests (repeat calls, release after large frees, concurrent calls). Both apps'
alloc_release.rs now just install the safe hook, and their crate roots gain
#![forbid(unsafe_code)], so every production crate except sysmem forbids it.
AGENTS.md/CLAUDE.md name the exception, per craftrules standards/never-crash.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- tiffraw: Mode::Header parses the first LJ92 frame header (Compression 7),
so probe_info reports Unsupported for subsampled frames (Sony lossless
M/S) as the full decode does, instead of claiming the file decodes.
- corpus: CC0 ILCE-7M4 lossless L/M/S samples from raw.pixls.us in
`cargo xtask corpus`; M/S listed as known unsupported.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- ↑/↓ over any slider nudge it by ≈ 1/200 of its range (exposure 0.05, most sliders 1,
temperature 50 K; ⇧ ×5), one undo step each; the key doesn't also move the selection.
- ⇧Z picks and moves to the next photo (with ⇧X / ⇧U).
- library.source {kind: missing}: photos whose file can't be found; a "Missing Photos" row
appears in the left panel when there are any (checked every 5 s).
- Loupe context menu: Zoom submenu. Parity: stale rows fixed (compare/survey, per-camera
defaults, paste from previous, capture-time edit).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- `cargo xtask parity` adds a weighted-completion line (✅ 1, 🟡 ½, out of scope left out) per
tier to the generated summary: 65.7% of 500 in-scope rows (P0 98.5%, P1 79.9%, P2 11.7%).
- ROADMAP: remaining work by package in agent-hours (≈ 295–540; ≈ 75–130 h wall clock with
4–6 agents), effort-weighted completion ≈ 20–30% for full parity, milestone statuses
refreshed, Totals replaced.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Resize modes: long edge, short edge, width, height, W x H (either orientation), megapixels,
percent; "don't enlarge" (default on); ppi written to JFIF density / PNG pHYs / TIFF tags.
- Full-size export of a cropped photo now renders the crop at native resolution (it was
upscaled to the uncropped long edge).
- Formats: Original (file copied byte for byte + XMP sidecar with the edits, named after the
output) and DNG (raw photos re-encoded losslessly, edits in the embedded XMP).
- Export dialog: Resize-to dropdown, values, Don't enlarge, Resolution; DNG/Original hide
pixel options.
- lightcraft-cli render goes through the shared exporter (colour space, metadata, ppi) and
takes any app.export option via --opt KEY=VALUE; MCP export schema lists the new params.
- xtask resolves the workspace root at run time (a moved checkout broke ci).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
cargo xtask parity verifies every cmd:/ctl: id and repository path cited in
docs/parity.md exists (engine commands and controls via lightcraft-cli, UI
commands from UI_COMMANDS), that row ids are unique and statuses valid, and prints
the per-section summary; --write refreshes the summary table. ci runs the
reference check after the tests. Parser, checker and summary are unit-tested.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Renders run in dedicated module Web Workers, each a second instance of the app's wasm module
(index.html compiles it once and posts the WebAssembly.Module to worker.js). A job crosses as
JSON (WireJob: source, settings, request); the worker reads originals and cached thumbnails from
browser storage itself, keeps decoded sources and per-view stage caches, and transfers RGBA back.
Jobs follow the worker that decoded the photo; the loupe waits for it. No wasm threads or
SharedArrayBuffer needed; falls back to inline rendering if no worker starts (?workers=0).
Thumbnail disk cache on the web: thumbs/<key>.jpg in OPFS/IndexedDB with an index on the main
thread, pruned to the desktop budget (2 GB → 80 %, LRU).
ui-egui: RenderOffload hook in the Renderer (queue/priorities/dedup stay in the UI).
engine: RenderJob.origin + MediaCache::origin_ref; preview: encode_jpeg/decode_jpeg public.
xtask web: copies worker.js; --serve sends COOP/COEP/CORP (documented in docs/web.md).
Headless Chrome, cold load + grid scroll, 24 demo photos: main-thread busy 12.4 s → 0.7–1.1 s,
long tasks 23 (max 1.1 s) → 0; grid filled in 6.3–8.4 s vs 13–16 s; slider job 4 ms.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- embedded_preview: CR3 PRVW/THMB boxes (plus RAF, RW2, ORF, PEF from the
format commits); stops at the first valid box of each kind
- engine: raw variants that report Unsupported are probed and shown from
their embedded JPEG (orientation from the JPEG or the raw), not an error
- crates/raw/tests/corpus.rs: every corpus/raw file is recognised, has a
JPEG preview (DNG optional) and decodes or is a known-unsupported
variant; prints decode timings
- xtask corpus --download: 37 CC0 raw.pixls.us samples (matched by SHA-256)
- ROADMAP/crate docs: format coverage, NEF Huffman / RW2 format 4 /
compressed ORF+RAF gaps, neutral colour-matrix fallback
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
apps/lightcraft-web runs the egui UI in the browser (eframe web, glow/WebGL2):
procedural demo library on load; in-memory imports via file picker
(Cmd+Shift+I) and drag-and-drop (MemStore + bytes-based engine hooks);
export goes through app.export and downloads in the browser; `?bench` logs
first-frame and slider latency. `cargo xtask web [--serve [port]] [--dev]`
builds with the matching wasm-bindgen CLI into <target>/web/; the web app is
part of the CI wasm check. Wasm fixes: lazy profiling Instant in pipeline,
web-time clocks in ui-egui (render timings were 0 on wasm). Docs: docs/web.md.
No persistence on web yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every image/icon/font/media/profile file must match an assets/ATTRIBUTION.md path pattern;
Adobe profile/template formats and Adobe-named assets are rejected.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Daniel Voshart
Claude
Иван Лапочкин
sebastian pfluegelmeier
radialmonster
Brandon Thomas
RicSchonfelder
goujicheng
Trey Reynolds
Guinther Kovalski