pyproject.toml

# `pip install AutoCua` — one wheel per OS/arch holding the AutoCua package
# and both Rust extensions prebuilt, so no Rust is needed on the user's
# machine: the web agent's (AutoCua/web), which maturin builds, and the
# Android agent's (AutoCua/android), which the release machine builds into
# place first and `include` below carries. Built with maturin:
#
#   pip install maturin
#   python -c "import AutoCua.android.agent"   # builds AutoCua/android/agent_native.*
#   maturin build --release -o dist            # -> dist/autocua-*.whl
#
# note/pypi-release.md has the per-machine commands. A checkout keeps working
# exactly as before (python main.py, the setup scripts, cargo building both
# agents on first import).

[build-system]
requires = ["maturin>=1.15,<2"]
build-backend = "maturin"

[project]
name = "AutoCua"
version = "0.1.7"
description = "AutoCua: an AI agent that operates your computer, shell, browser and iPhone"
readme = "README.md"
requires-python = ">=3.10"
license = "MIT"
license-files = ["LICENSE", "THIRD_PARTY_NOTICES.md", "THIRD_PARTY_LICENSES.txt"]
authors = [{ name = "Ashish Yadav" }]
keywords = ["agent", "automation", "computer-use", "browser-agent", "llm", "ios"]
classifiers = [
    "Development Status :: 4 - Beta",
    "Intended Audience :: Developers",
    "Operating System :: MacOS",
    "Operating System :: Microsoft :: Windows",
    "Operating System :: POSIX :: Linux",
    "Programming Language :: Python :: 3",
    "Programming Language :: Rust",
]

# Everything main.py needs on either platform, terminal and UI alike. The
# markers pick the OS; the floors are the versions the Windows list pinned.
dependencies = [
    "requests>=2.34.2",
    "python-dotenv>=1.2.4",
    "Pillow>=12.3.0",
    "openai>=2.17.0",
    "anthropic>=0.83.0",
    "groq>=1.0.0",
    "google-genai>=1.64.0",
    "flask>=3.1.3",
    "werkzeug>=3.1.9",
    "pywebview>=6.1",
    "python-telegram-bot>=21.6",
    "pyautogui>=0.9.54; sys_platform == 'darwin' or sys_platform == 'win32'",
    # macOS computer use
    "pyobjc-core>=11; sys_platform == 'darwin'",
    "pyobjc-framework-Cocoa>=11; sys_platform == 'darwin'",
    "pyobjc-framework-Quartz>=11; sys_platform == 'darwin'",
    "pyobjc-framework-ApplicationServices>=11; sys_platform == 'darwin'",
    "pyobjc-framework-Vision>=11; sys_platform == 'darwin'",
    "pyobjc-framework-WebKit>=11; sys_platform == 'darwin'",
    "pynput>=1.7; sys_platform == 'darwin'",
    # Each run's recording is encoded with OpenCV on macOS and with
    # imageio-ffmpeg on Linux and Windows: OpenCV's Linux wheels cannot write
    # H.264, and its Windows wheels only at a fixed ~900 MB a minute
    # (utils/run_recorder.py). imageio-ffmpeg's ffmpeg then compresses it on
    # every OS (utils/video_compress.py).
    "opencv-python>=4.9; sys_platform == 'darwin'",
    "imageio-ffmpeg>=0.5",
    # Linux computer use — OCR fusion (PP-OCRv6 via onnxruntime, see
    # AutoCua/linux/tree/ocr.py); AT-SPI itself comes from the distro's
    # python3-gi, which pip cannot supply (requirements_linux.txt explains).
    "onnxruntime>=1.20; sys_platform == 'linux'",
    # Windows computer use
    "pywinauto>=0.6.9; sys_platform == 'win32'",
    "pywin32>=311; sys_platform == 'win32'",
    "comtypes>=1.4; sys_platform == 'win32'",
    "interception-python>=1.13; sys_platform == 'win32'",
    "keyboard>=0.13.5; sys_platform == 'win32'",
    "psutil>=5.9; sys_platform == 'win32'",
    "pyperclip>=1.8; sys_platform == 'win32'",
    "numpy>=1.26; sys_platform == 'win32' or sys_platform == 'linux'",
    "PySide6>=6.6; sys_platform == 'win32'",
    "winrt-runtime>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Foundation>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Foundation.Collections>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Media.Ocr>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Globalization>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Graphics.Imaging>=3.2; sys_platform == 'win32'",
    "winrt-Windows.Storage.Streams>=3.2; sys_platform == 'win32'",
]

[project.optional-dependencies]
# Paired iPhone/iPad (the simulator path needs only Xcode).
ios = ["pymobiledevice3>=10; sys_platform == 'darwin'"]

[project.urls]
Homepage = "https://github.com/AutoCua/AutoCua"
Source = "https://github.com/AutoCua/AutoCua"

[tool.maturin]
# The crate lives inside the package (AutoCua/web) and is imported as
# AutoCua.web.agent_native; python-source = "." makes ./AutoCua the package.
manifest-path = "AutoCua/web/Cargo.toml"
module-name = "AutoCua.web.agent_native"
python-source = "."
# The Android agent's extension, AutoCua/android/agent_native, the second
# crate: maturin builds one crate per wheel (manifest-path above), so each
# release machine builds this one into place before `maturin build`
# (`python -c "import AutoCua.android.agent"` — AutoCua/utils/rust.py; on Linux
# `cargo zigbuild` against the wheel's glibc floor; note/pypi-release.md), and
# it is named here because maturin's package walk honours .gitignore, where
# *.py[cod] and *.so cover it: unlisted, it never reaches the wheel. The
# crate's sources ship too (unlike the web crate's, below): a wheel that pip
# builds from the sdist on a platform without one of ours has no module in
# it, and rust.py builds it from them on the first Android run. One line per
# platform's suffix; the one that matches nothing is skipped.
include = [
    { path = "AutoCua/android/agent_native.pyd", format = "wheel" },   # Windows
    { path = "AutoCua/android/agent_native.so", format = "wheel" },    # macOS, Linux
]
exclude = [
    # wheel-only: the crate sources stay in the sdist so it can build, but a
    # wheel ships the compiled extension — no Cargo.toml beside it means the
    # loader never tries cargo.
    { path = "AutoCua/web/**/*.rs", format = "wheel" },
    { path = "AutoCua/web/*.rs", format = "wheel" },
    { path = "AutoCua/web/Cargo.toml", format = "wheel" },
    { path = "AutoCua/web/Cargo.lock", format = "wheel" },
    { path = "AutoCua/web/build.rs", format = "wheel" },
    # never shipped
    "AutoCua/web/target/**",
    "AutoCua/web/scratchpad/**",
    "AutoCua/android/target/**",
    "AutoCua/android/scratchpad/**",
    # the licence-file generator (AutoCua/utils/third_party_licenses.py): a
    # maintainer's tool that reads the cargo registry, not part of the app
    { path = "AutoCua/utils/third_party_licenses.py", format = "wheel" },
    # rust.py's staging copy of either extension (agent_native.pyd.new),
    # which only a crash leaves behind — and which .gitignore does not cover
    "AutoCua/*/agent_native.*.new",
    "AutoCua/mac/scratchpad/**",
    "AutoCua/windows/scratchpad/**",
    "AutoCua/ios/scratchpad/**",
    "AutoCua/linux/scratchpad/**",
    # the PP-OCRv6 models ocr_models.py downloads on the first Linux scan
    "AutoCua/linux/tree/models/**",
    # legacy in-package secret locations (keys now live in AutoCua_data/)
    "AutoCua/api_key/**",
    "AutoCua/agent_conversation/*.json",
    "AutoCua/agent_conversation/chat_*/**",
    "AutoCua/web/HANDOFF.md",
    "AutoCua/Structure.md",
    "AutoCua/**/test.py",
    "AutoCua/**/old_element.py",
    "AutoCua/**/__pycache__/**",
    "AutoCua/ios_connector/WebDriverAgent/**",
    "AutoCua/ios_connector/paired_devices.json",
    # the Chromium desktop shell's npm install: Electron itself, ~300 MB and
    # built per OS (a checkout runs `npm install` in AutoCua/desktop; a pip
    # install fetches it into a per-user cache, AutoCua/utils/electron.py)
    "AutoCua/desktop/node_modules/**",
    # the UI's npm install (React, Vite, TypeScript: `npm ci` in AutoCua/ui builds
    # ui/index.js, which is committed and ships; pip users need none of this)
    "AutoCua/ui/node_modules/**",
]