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 <[email protected]>
- The crate ports Apache-2.0-only Hugging Face transformers code: it is
now `license = "Apache-2.0"`, and every ported file carries the
modification notice Apache-2.0 §4(b) asks for.
- Weights::open checks every tensor's byte range (and the header)
against the file's length: a truncated or damaged download is an
error at load, never a short read later; regression test with
hostile offsets.
- fetch/: the model download, only when asked for. An ordered mirror
list (env LIGHTCRAFT_SAM3_MIRRORS, a mirrors file, DEFAULT_MIRRORS —
empty until LightCraft's CDN exists, TODO for the maintainer),
fallback to the next mirror, resume via Range, redirects, connect and
stall timeouts, cancel, size caps, exact size + SHA-256 (pinned from
tools/install-sam3.sh) before an atomic rename; damaged files are
deleted, HTTP error bodies never written.
- HTTP/1.1 client over std sockets + rustls with the RustCrypto provider
(rustls-rustcrypto) and webpki-roots: no ring / aws-lc-rs, nothing
compiled from C (no new `cc` users in the app's build graph).
- Tests against a local TcpListener server (no internet): fallback,
redirects, chunked bodies, resume, 416, hash mismatch, too much data,
stalls, cancel.
- *.safetensors are gitignored.
Co-Authored-By: Claude Opus 5.5 <[email protected]>
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) <[email protected]>
Claude-Session: https://claude.ai/code/session_01C8BkS7JDqa1EMcjW6spDGA
