Cargo.toml

[package]
name = "eaRS"
version = "0.6.0"
edition = "2024"
description = "Easy automatic speech recognition library using Kyutai's STT models"
license = "MIT"
repository = "https://github.com/tommyfalkowski/eaRS"
authors = ["Tommy Falkowski"]
homepage = "https://byteowlz.com"

[build-dependencies]
pkg-config = "0.3"

[lib]
name = "ears"
path = "src/lib.rs"

[[bin]]
name = "ears"
path = "src/main.rs"

[[bin]]
name = "ears-server"
path = "src/bin/ears-server.rs"

[[bin]]
name = "ears-dictation"
path = "src/bin/ears-dictation.rs"

[dependencies]
aho-corasick = "1"
anyhow = "1.0"
atty = "0.2"
candle = { version = "0.9", package = "candle-core" }
candle-nn = "0.9"
clap = { version = "4.4.12", features = ["derive"] }
cpal = "0.16"
crossbeam-channel = "0.5"
hf-hub = "0.4.3"
kaudio = "0.2.1"
moshi = "0.6.4"
# prkt must come before sentencepiece to ensure ort's protobuf symbols are used
# (both bundle protobuf-lite, and the linker uses the first definition)
prkt = { path = "prkt", optional = true }
# Pin ort to a version compatible with both prkt (caret ^2.0.0-rc.10) and
# parakeet-rs (caret ^2.0.0-rc.12). eaRS pins exact so both engines resolve to
# one ort; rc.12 satisfies both caret requirements. eaRS's own ort dep is only
# pulled by the prkt `parakeet` feature (it is optional below).
ort = { version = "=2.0.0-rc.12", optional = true, features = ["load-dynamic"] }
ort-sys = { version = "=2.0.0-rc.12", optional = true, features = ["download-binaries", "disable-linking"] }
serde = { version = "1.0.210", features = ["derive"] }
serde_json = "1.0.115"
toml = "0.9"
dirs = "6.0"
time = { version = "0.3", features = ["formatting"] }
tokio = { version = "1.0", features = ["full"] }
tokio-tungstenite = "0.28"
futures = "0.3"
indicatif = "0.17.5"
uuid = { version = "1.0", features = ["v4"] }
colored = "3.0"
libc = "0.2"
enigo = "0.6"
rdev = "0.5"
futures-util = "0.3"
ctrlc = { version = "3.4", features = ["termination"] }
# cudarc auto-detects CUDA version from the build system (nvcc)
# Users can override by setting CUDARC_CUDA_VERSION environment variable
cudarc = { version = "0.17", optional = true, features = ["cuda-version-from-build-system"] }
notifica = "3.0.2"
webrtc-vad = { version = "0.4", optional = true }
# parakeet-rs (Nemotron streaming). Mutually exclusive in practice with the
# `parakeet` (prkt) feature: prkt pins ort =2.0.0-rc.10 while parakeet-rs needs
# ort 2.0.0-rc.12, and eaRS's own `ort` dep is optional (only pulled by the
# prkt feature), so do not enable both `parakeet` and `parakeet-rs` together.
parakeet-rs = { package = "parakeet-rs", git = "https://github.com/altunenes/parakeet-rs.git", rev = "086e3bb734dd9a3f01e4e99ef6895da0db85b47a", optional = true, default-features = false }

# transcribe.cpp GGUF streaming engine. Pinned past the 0.1.3 crates.io
# release because multitalker-parakeet streaming landed after it. Builds the
# C++ library from source via cmake; backend features (metal/cuda) are
# forwarded from eaRS's `apple`/`nvidia` features.
transcribe-cpp = { git = "https://github.com/handy-computer/transcribe.cpp.git", rev = "5a5a49664a8ea1f0e5b3be1dfc544730d1b62561", optional = true, default-features = false }
# `ears models refresh` fetches the maintained catalog.json. Reuses hf-hub's TLS
# stack (native-tls); already in the tree, so this adds no new transitive cost.
ureq = { version = "2.12", optional = true }

# Platform-specific sentencepiece configuration
# All platforms: compile from source (static linking).
[target.'cfg(target_os = "linux")'.dependencies]
sentencepiece = { version = "0.12", features = ["static"] }
uinput = "0.1"
# hf-hub's sync (ureq) path uses native-tls -> openssl-sys on Linux. Build
# OpenSSL from source so cross-compiled release artifacts (cargo-zigbuild) do
# not depend on a system OpenSSL that isn't present in the cross sysroot.
openssl = { version = "0.10", features = ["vendored"] }

[target.'cfg(not(target_os = "linux"))'.dependencies]
sentencepiece = { version = "0.12", features = ["static"] }

[features]
default = []
cudnn = ["candle/cudnn", "candle-nn/cudnn"]
hooks = []

# Engines
parakeet = ["dep:prkt", "dep:ort", "dep:ort-sys", "dep:webrtc-vad"]
# parakeet-rs Nemotron streaming engine (ort 2.0.0-rc.12). Enables ort via
# parakeet-rs's own features. Do NOT combine with the `parakeet` (prkt) feature.
# Dynamic ORT avoids symbol collisions with eaRS's statically linked
# SentencePiece protobuf-lite. Set ORT_DYLIB_PATH to the ORT dylib at runtime.
parakeet-rs = ["dep:parakeet-rs", "parakeet-rs/cpu", "parakeet-rs/load-dynamic"]
# transcribe.cpp GGUF engine (streaming multitalker/nemotron models).
# CPU-only by default; combine with `metal`/`cuda` for GPU compute.
transcribe-cpp = ["dep:transcribe-cpp", "dep:ureq"]

# Acceleration (shared flags)
nvidia = ["dep:cudarc", "candle/cuda", "candle-nn/cuda", "transcribe-cpp?/cuda"]
apple = ["candle/metal", "candle-nn/metal", "transcribe-cpp?/metal"]
amd = []
directml = []

# Backwards-compat aliases
cuda = ["nvidia"]
metal = ["apple"]
parakeet-cuda = ["parakeet", "nvidia"]
parakeet-coreml = ["parakeet", "apple"]
parakeet-directml = ["parakeet", "directml"]
parakeet-rocm = ["parakeet", "amd"]

[profile.release]
debug = true

[profile.release-no-debug]
inherits = "release"
debug = false

# The profile that 'dist' will build with
[profile.dist]
inherits = "release"
lto = "thin"

[patch.crates-io]
sentencepiece-sys = { path = "vendor/sentencepiece-sys" }