coremlit 0.1.2

[package]
name = "coremlit"
description = "Safe, synchronous CoreML runtime for macOS (CPU/GPU/Neural Engine) with opt-in on-device multimodal pipelines: speech (Whisper STT, forced alignment, speaker diarization, Silero VAD), AudioSet sound-event tagging, and audio/text/image embeddings (CLAP, granite, SigLIP)"
version.workspace = true
edition.workspace = true
publish = true
rust-version.workspace = true
license.workspace = true
repository.workspace = true
readme = "../README.md"
keywords = ["coreml", "macos", "speech", "embeddings", "multimodal"]
categories = ["multimedia::audio", "os::macos-apis", "artificial-intelligence", "computer-vision", "api-bindings"]
exclude = [
  "conversion/",
  "tests/align/fixtures/",
  "tests/ced/fixtures/",
  "tests/clap/fixtures/",
  "tests/face/fixtures/",
  "tests/identity/fixtures/",
  "tests/lid/fixtures/",
  "tests/siglip/fixtures/",
  "tests/speaker/fixtures/",
  "tests/vad/fixtures/",
  "tests/whisper/fixtures/",
  "tests/whisper_swift_probes/",
]

[package.metadata.docs.rs]
all-features = true
default-target = "aarch64-apple-darwin"
targets = ["aarch64-apple-darwin"]

[features]
default = []
# `windit?/serde` is a WEAK dep-feature: it turns on windit's own serde impls
# only when windit is already in the graph (via `clap`/`granite`/`ced`/`lid`),
# never pulling windit in on its own. It is what lets
# `embeddings::granite::LongTextOptions` derive — it composes windit's
# `WindowOptions`, so it cannot be serialized without it — and it is the
# step that module's doc named as the prerequisite.
serde = ["dep:serde", "windit?/serde"]
tracing = ["dep:tracing"]

whisper = [
  "dep:libc",
  "dep:mach2",
  "dep:rand",
  "dep:serde_json",
  "dep:tokenizers",
  "dep:unicode_categories",
]
# Former whisperkit `nl-recognizer` (Apple `NLLanguageRecognizer` redetection).
# Whisper-only code, so it implies `whisper`.
nl-recognizer = ["whisper", "dep:objc2-natural-language"]

# `dep:humantime` is `AlignerOptions`'s canonical `Display` rendering
# `max_intra_silent_run` as humantime text ("80ms"), the SAME grammar
# `humantime::format_duration` already writes elsewhere in this workspace
# (ingraph's own `[registry]` seat; asry/diaric/zuoer instead use
# `humantime-serde`, a document-only bridge that does not implement `Display`
# itself, so it cannot fill this seam).
#
# `dep:serde_json` reads a model's own `{token: id}` vocabulary
# (`audio::align::vocab::Vocabulary`) and writes, for the model's contract, the
# tokenizer document asry's seam builder parses, its specials declared special
# added tokens; `dep:serde` is the entry-by-entry visitor that reader
# refuses a repeated token with. `dep:tokenizers` parses that document again at
# load, as asry's builder does, so the aligner reads back the columns the seam
# reserves: asry keeps its tokenizer off its public API. asry's `emissions`
# feature already pulls `tokenizers` 0.23, which depends on serde and
# serde_json, so these pull no new version into the lockfile, and the parse is
# asry's own crate (and `dep:serde` does not turn on this crate's own `serde`
# feature). The workspace row's `unstable_wasm` adds `fancy-regex` (with
# `bit-set` and `bit-vec`, all three already locked) to an `align`-only build,
# compiled and unused: the `onig` backend asry enables takes precedence.
align = ["dep:asry", "dep:humantime", "dep:serde_json", "dep:serde", "dep:tokenizers"]
# asry's ONNX aligner (pulls ort + whisper.cpp)
# as the word-timing parity oracle. DEV/TEST only — adds nothing to the library.
align-oracle = ["align", "asry/alignment"]

speaker = ["dep:diaric"]
vad = ["dep:zuoer"]
# `dep:serde_json` is the shared `embeddings::tokenizer_guard`: the tokenizers
# crate depends on serde_json itself, so this pulls nothing new into the graph.
clap = ["dep:rustfft", "dep:tokenizers", "dep:windit", "dep:serde_json"]

# `dep:serde_json` reads the BPE merge table back out of the loaded model for
# `embeddings::granite::token_index`'s separatorless fast lane (#72); tokenizers
# itself depends on serde_json, so this pulls nothing new into the graph.
granite = ["dep:tokenizers", "dep:windit", "windit/text", "dep:sha2", "dep:serde_json"]

# LICENSE PREREQUISITE (owner-gated, NOT resolved by this change): `pixon` is
# currently licensed GPL-3.0-or-later; the owner will relicense it permissive
# (MIT OR Apache-2.0) before this branch merges. It is a flagged merge blocker
# tracked outside this change, not something wired here.
siglip = ["dep:tokenizers", "dep:pixon", "dep:sha2", "dep:serde_json"]

ced = ["dep:rustfft", "dep:soundevents-dataset", "dep:windit"]

lid = ["dep:rustfft", "dep:windit"]

# `audio::identity` — one fixed 6 s window in, one raw 192-d speaker embedding
# out. Only the mel front end needs a dependency; the windowing and scoring of
# several windows are the caller's, so there is no `windit` here.
identity = ["dep:rustfft"]

# The face-embedding surface (`embeddings::face`): the ArcFace 5-point
# similarity alignment plus a manifest-driven CoreML embedder. Its ONE
# dependency is `sha2`, already an optional workspace dep that `granite` and
# `siglip` pull: `FaceEmbedder::load` hashes the artifact directory it loads so
# an embedding carries the identity of the weights that produced it, not just
# their schema (issue #138 §3). The alignment is `f64` arithmetic and the
# embedder builds a `MultiArray` the always-compiled runtime core already owns.
#
# PLAIN `face`, and that is deliberate rather than a rename left undone. Two
# thirds of this feature is encumbered by nothing: the alignment is `f64`
# arithmetic over a synthetic golden, and the embedder loads a CALLER-SUPPLIED
# path — including an artifact whose licence permits a product. The ONE staged
# artifact (InsightFace's `w600k_r50`, research-only at both the weights and the
# corpus layer) rides `commercial-face-arcface` below, which gates its manifest
# module and nothing else. Gating this feature instead would make the register
# say "requires a commercial licence" about code that requires nothing.
face = ["dep:sha2"]

# Requires a commercial licence for the ArcFace `w600k_r50` weights and their WebFace600K
# corpus. InsightFace publishes both for non-commercial research only, and offers no
# commercial grant over them, so what this feature adds is the manifest for an artifact that
# may be used to develop, evaluate and test — never to ship. It is never in `default`, no
# other feature may pull it in, and this repository still redistributes no weight bytes: the
# artifact lives in a PRIVATE Hugging Face repository that CI fetches with a read token, the
# same posture `identity` already has (see NOTICE, "CI DOWNLOADS; IT DOES NOT
# REDISTRIBUTE"). The `commercial-` prefix reads backwards on purpose — it looks like
# "cleared for commercial use" — which is why the sentence you just read has to be the first
# one.
#
# WHAT ENABLING IT DOES, exactly: it WIRES coremlit's own registered copy of that artifact.
# It compiles the `embeddings::face::arcface` manifest module (four constants naming the
# bundle, the path `MODELS_LOCK` stages it to, its I/O contract and its measured compute
# placement), it turns on the `arcface` `MODELS_LOCK` table that stages those bytes for CI,
# and it builds the four gated suites in `tests/face/`. Manifested, staged, tested — that
# wiring, and only that wiring, is what the licence register in `tests/model_licences.rs`
# governs.
#
# WHAT IT DOES NOT DO: it does not stop a caller loading their own copy of these — or of any
# other — weights through the plain `face` feature. `FaceEmbedder::load` takes a
# caller-supplied path and a caller-written `FaceModel`, so a product with a commercially
# licensed ArcFace-shaped model of its own must be able to write that manifest and load it
# under `face`, and nothing but a digest would tell those bytes from InsightFace's. The
# licence of bytes a caller supplies is between them and whoever published them — for these
# weights, InsightFace. That residual is issue #138 §8's, stated in the module doc of
# `tests/model_licences.rs`; what this feature's NAME and first sentence are for is to tell
# you that coremlit's registered artifact needs a licence, and coremlit hands nobody the
# bytes.
commercial-face-arcface = ["face"]

[dependencies]
objc2.workspace = true
objc2-core-ml.workspace = true
objc2-foundation.workspace = true
objc2-core-video.workspace = true
objc2-core-foundation.workspace = true
block2.workspace = true
half.workspace = true
thiserror.workspace = true
derive_more.workspace = true

# --- cross-cutting optionals ---
serde = { workspace = true, optional = true }
tracing = { workspace = true, optional = true }

# --- audio::whisper optionals ---
libc = { workspace = true, optional = true }
mach2 = { workspace = true, optional = true }
rand = { workspace = true, optional = true }
serde_json = { workspace = true, optional = true }
tokenizers = { workspace = true, optional = true }
unicode_categories = { workspace = true, optional = true }
objc2-natural-language = { workspace = true, optional = true }

# --- audio::align optional dep (the `align` feature) ---
asry = { version = "0.3", default-features = false, features = ["emissions"], optional = true }
humantime = { version = "2", optional = true }

# --- audio::speaker optional dep (the `speaker` feature) ---
diaric = { version = "0.2", optional = true }

# --- audio::vad optional dep (the `vad` feature) ---
zuoer = { version = "0.3", optional = true }

# --- embeddings::clap optionals (the `clap` feature) ---
rustfft = { workspace = true, optional = true }

# audio::ced's 527-class rated AudioSet label set (ort-free data crate; see the
# workspace table for the feature pins).
soundevents-dataset = { workspace = true, optional = true }

# --- shared windowing engine (the `clap`, `granite` + `lid` features) ---
windit = { version = "0.5", optional = true }

# --- embeddings::granite runtime tokenizer-identity digest (the `granite` feature) ---
sha2 = { workspace = true, optional = true }

# --- embeddings::siglip image resampler (the `siglip` feature) ---
# LICENSE: pixon is GPL-3.0-or-later today; its owner-gated relicense to
# MIT OR Apache-2.0 is a merge blocker for this branch (see the `siglip` feature
# comment above).
pixon = { version = "0.6", default-features = false, features = ["std", "rgb"], optional = true }

[dev-dependencies]
half.workspace = true
cpal.workspace = true
criterion.workspace = true
hound.workspace = true
rubato.workspace = true
serde = { workspace = true, features = ["derive"] }
serde_json.workspace = true
postcard.workspace = true
sha2.workspace = true
tempfile.workspace = true
toml.workspace = true
syn.workspace = true
proc-macro2 = { workspace = true, features = ["span-locations"] }
tokenizers.workspace = true
npyz.workspace = true
png.workspace = true
libc.workspace = true
mach2.workspace = true
asry = { version = "0.3", default-features = false, features = ["emissions"] }

[[test]]
name = "face_align_golden"
path = "tests/face/align_golden.rs"
required-features = ["face"]

# The two-sided gate assertion. `face`, NOT `commercial-face-arcface`: half of
# what it pins is what a plain-`face` caller can still do, and a suite that
# needed the gate could not say that.
[[test]]
name = "face_gate"
path = "tests/face/gate.rs"
required-features = ["face"]

# The four `commercial-face-arcface` gate suites. Their model gates need the
# staged w600k_r50 bundle and are `#[ignore]`d; the eight hermetic tests beside
# them read only committed bytes — the fixtures, the ONNX reference, and the
# vendored silero bundle the manifest refuses — so the modelless `features` job
# runs those on every PR.
[[test]]
name = "face_model_io"
path = "tests/face/model_io.rs"
required-features = ["commercial-face-arcface"]

[[test]]
name = "face_parity"
path = "tests/face/parity.rs"
required-features = ["commercial-face-arcface"]

[[test]]
name = "face_known_pairs"
path = "tests/face/known_pairs.rs"
required-features = ["commercial-face-arcface"]

[[test]]
name = "face_placement"
path = "tests/face/placement.rs"
required-features = ["commercial-face-arcface"]

[[test]]
name = "whisper_golden_provenance"
path = "tests/whisper/golden_provenance.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_longform"
path = "tests/whisper/longform.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_model_io"
path = "tests/whisper/model_io.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_models_lock"
path = "tests/whisper/models_lock.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_parity_es"
path = "tests/whisper/parity_es.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_parity_jfk"
path = "tests/whisper/parity_jfk.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_pipeline"
path = "tests/whisper/pipeline.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_streaming"
path = "tests/whisper/streaming.rs"
required-features = ["whisper"]

[[test]]
name = "whisper_vad_source"
path = "tests/whisper/vad_source.rs"
required-features = ["whisper", "vad"]

[[test]]
name = "whisper_rtf_gate"
path = "benches/whisper/rtf_gate.rs"
required-features = ["whisper"]

[[bench]]
name = "whisper_stages"
path = "benches/whisper/stages.rs"
harness = false
required-features = ["whisper"]

[[bench]]
name = "whisper_rtf"
path = "benches/whisper/rtf.rs"
harness = false
required-features = ["whisper"]

[[example]]
name = "whisper_mic_stream"
path = "examples/whisper/mic_stream.rs"
test = true
required-features = ["whisper"]

[[example]]
name = "whisper_transcribe_wav"
path = "examples/whisper/transcribe_wav.rs"
required-features = ["whisper"]

# --- audio::align ---
[[test]]
name = "align_align_chunk"
path = "tests/align/align_chunk.rs"
required-features = ["align"]

[[test]]
name = "align_model_io"
path = "tests/align/model_io.rs"
required-features = ["align"]

[[test]]
name = "align_registry_handle"
path = "tests/align/registry_handle.rs"
required-features = ["align"]

# The word-timing parity gate needs asry's ONNX aligner oracle (ort +
# whisper.cpp) — the former alignkit `parity-oracle`, now `align-oracle`.
[[test]]
name = "align_parity_words"
path = "tests/align/parity_words.rs"
required-features = ["align-oracle"]

[[bench]]
name = "align_align"
path = "benches/align/align.rs"
harness = false
required-features = ["align"]

# --- audio::speaker ---
[[test]]
name = "speaker_argmax_model_io"
path = "tests/speaker/argmax_model_io.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_golden_metadata"
path = "tests/speaker/golden_metadata.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_model_io"
path = "tests/speaker/model_io.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_argmax_accuracy"
path = "tests/speaker/parity_argmax_accuracy.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_argmax_swift"
path = "tests/speaker/parity_argmax_swift.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_embed"
path = "tests/speaker/parity_embed.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_seg"
path = "tests/speaker/parity_seg.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_online_swift"
path = "tests/speaker/parity_online_swift.rs"
required-features = ["speaker"]

[[test]]
name = "speaker_parity_diarize_wiring"
path = "tests/speaker/parity_diarize_wiring.rs"
required-features = ["speaker"]

# --- audio::vad ---
[[test]]
name = "vad_model_io"
path = "tests/vad/model_io.rs"
required-features = ["vad"]

[[test]]
name = "vad_model_state"
path = "tests/vad/model_state.rs"
required-features = ["vad"]

[[test]]
name = "vad_parity_swift"
path = "tests/vad/parity_swift.rs"
required-features = ["vad"]

[[test]]
name = "vad_reexport"
path = "tests/vad/reexport.rs"
required-features = ["vad"]

[[test]]
name = "clap_e2e"
path = "tests/clap/e2e.rs"
required-features = ["clap"]

[[test]]
name = "clap_model_io"
path = "tests/clap/model_io.rs"
required-features = ["clap"]

[[test]]
name = "clap_text_model_io"
path = "tests/clap/text_model_io.rs"
required-features = ["clap"]

[[test]]
name = "clap_placement"
path = "tests/clap/placement.rs"
required-features = ["clap"]

[[test]]
name = "clap_tokenizer_identity"
path = "tests/clap/tokenizer_identity.rs"
required-features = ["clap"]

[[test]]
name = "granite_tokenizer_identity"
path = "tests/granite/tokenizer_identity.rs"
required-features = ["granite"]

[[test]]
name = "granite_driver_crosscheck"
path = "tests/granite/driver_crosscheck.rs"
required-features = ["granite"]

[[test]]
name = "granite_model_io"
path = "tests/granite/model_io.rs"
required-features = ["granite"]

[[test]]
name = "granite_parity_embed"
path = "tests/granite/parity_embed.rs"
required-features = ["granite"]

[[test]]
name = "granite_placement"
path = "tests/granite/placement.rs"
required-features = ["granite"]

[[test]]
name = "granite_embed_long"
path = "tests/granite/embed_long.rs"
required-features = ["granite"]

[[test]]
name = "granite_embed_windows"
path = "tests/granite/embed_windows.rs"
required-features = ["granite"]

[[test]]
name = "siglip_tokenizer_identity"
path = "tests/siglip/tokenizer_identity.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_preprocess"
path = "tests/siglip/preprocess.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_model_io"
path = "tests/siglip/model_io.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_text_model_io"
path = "tests/siglip/text_model_io.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_parity_embed"
path = "tests/siglip/parity_embed.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_placement"
path = "tests/siglip/placement.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_e2e"
path = "tests/siglip/e2e.rs"
required-features = ["siglip"]

[[test]]
name = "siglip_band_provenance"
path = "tests/siglip/band_provenance.rs"
required-features = ["siglip"]

[[test]]
name = "ced_model_io"
path = "tests/ced/model_io.rs"
required-features = ["ced"]

[[test]]
name = "ced_parity_logits"
path = "tests/ced/parity_logits.rs"
required-features = ["ced"]

[[test]]
name = "ced_placement"
path = "tests/ced/placement.rs"
required-features = ["ced"]

[[test]]
name = "ced_e2e"
path = "tests/ced/e2e.rs"
required-features = ["ced"]

[[test]]
name = "lid_model_io"
path = "tests/lid/model_io.rs"
required-features = ["lid"]

[[test]]
name = "lid_e2e"
path = "tests/lid/e2e.rs"
required-features = ["lid"]

[[test]]
name = "lid_long_clip"
path = "tests/lid/long_clip.rs"
required-features = ["lid"]

[[test]]
name = "identity_model_io"
path = "tests/identity/model_io.rs"
required-features = ["identity"]

[[test]]
name = "identity_parity_embed"
path = "tests/identity/parity_embed.rs"
required-features = ["identity"]

[[bench]]
name = "clap_encode"
path = "benches/clap/encode.rs"
harness = false
required-features = ["clap"]

[lints]
workspace = true