Automation foundation model for tiny devices: 2-bit, 8-29 MB, tool calls, ASR, structured extraction and embeddings on phones, wearables, smart homes, robots, cars and microcontrollers.
[project]
name = "cactus-needle"
version = "3.2.1"
description = "A 14MB foundation tool-calling model for tiny devices: inference, LoRA finetuning, and build."
requires-python = ">=3.9"
readme = "README.md"
license = { text = "Apache-2.0" }
dependencies = [
"huggingface_hub",
]
[project.optional-dependencies]
train = [
"numpy",
"jax",
"jaxlib",
"flax>=0.10.2",
"optax",
"safetensors",
"sentencepiece",
"whisper-normalizer",
"soxr"
]
gpu = ["jax[cuda12]"]
metal = [
"jax==0.4.38",
"jaxlib==0.4.38",
"jax-metal"
]
mic = ["numpy", "soxr", "sounddevice"]
compare = ["openai-whisper", "moonshine-voice"]
test = ["pytest", "pydantic"]
# mlx-cpu (the CPU backend Linux needs) is gated to Linux by mlx itself; macOS gets mlx + Metal.
mlx = ["mlx[cpu]"]
[project.scripts]
needle = "needle.cli:main"
[tool.setuptools.packages.find]
include = ["needle*"]
[tool.setuptools.package-data]
"needle.model" = ["*.model", "*.vocab"]
"needle.playground" = ["index.html", "app.js", "style.css"]
[build-system]
requires = ["setuptools>=68.0"]
build-backend = "setuptools.build_meta"
[tool.pytest.ini_options]
testpaths = ["tests"]
markers = ["slow: end-to-end JAX build/finetune tests (init a tiny model)"]