axintai 0.1.0

Python SDK for Axint — define Apple App Intents in Python, ship to Siri.

axintai — Python SDK for Axint

Define Apple App Intents in Python. Ship them to Siri, Shortcuts, and Spotlight through the same open-source compiler pipeline that powers the TypeScript SDK.

The Python SDK includes a native Swift generator — no Node.js dependency required. Parse, validate, and compile intents entirely from Python.

Install

Note: The PyPI package is not published yet. Install from source for now:

git clone https://github.com/agenticempire/axint.git
cd axint/python
pip install -e '.[dev]'

Define an intent

from axintai import define_intent, param

create_event = define_intent(
    name="CreateCalendarEventIntent",
    title="Create Calendar Event",
    description="Creates a new event on the user's calendar",
    domain="productivity",
    params={
        "event_title": param.string("Title of the event"),
        "start_date": param.date("When the event starts"),
        "duration_minutes": param.int("Length of the event in minutes"),
        "is_all_day": param.boolean("Whether the event is all-day", optional=True, default=False),
    },
    entitlements=["com.apple.developer.calendars"],
    info_plist_keys=["NSCalendarsUsageDescription"],
)

Compile it

# Parse and inspect the IR
axintai parse intents/create_event.py
axintai parse intents/create_event.py --json

# Compile Python → Swift (native, no Node.js needed)
axintai compile intents/create_event.py --stdout
axintai compile intents/create_event.py --out ios/Intents/

# With companion fragments
axintai compile intents/create_event.py --out ios/Intents/ --emit-info-plist --emit-entitlements

# Validate without generating Swift
axintai validate intents/create_event.py

# Machine-readable output
axintai compile intents/create_event.py --json

Use it as a library

from axintai import define_intent, param, generate_swift, validate_intent

intent = define_intent(
    name="SendMessage",
    title="Send Message",
    description="Sends a message",
    domain="messaging",
    params={"body": param.string("Message text")},
)

ir = intent.to_ir()
diagnostics = validate_intent(ir)
swift_code = generate_swift(ir)

Cross-language bridge

The Python SDK produces compatible IR JSON that the TypeScript compiler can consume. You can pipe it in for additional validation and Swift generation:

axintai parse intent.py --json | axint compile - --from-ir --stdout

Why Python?

Every language-agnostic analysis layer in Axint — the IR, the validator, the generator — works with a stable JSON schema. The Python SDK implements the full pipeline natively, unlocking a massive population of developers who shouldn't have to learn TypeScript to build Siri integrations.

The Python parser never runs your code. It walks the Python AST the same way the TypeScript compiler walks the TS AST, so axintai compile is deterministic, sandboxable, and reproducible.

Parity with the TypeScript SDK

Feature	TypeScript	Python
define_intent / defineIntent	✅	✅
param.string/int/double/...	✅	✅
entitlements, infoPlistKeys	✅	✅
isDiscoverable	✅	✅
Multi-intent files	✅	✅
Swift codegen (native)	✅	✅
IR validation	✅	✅
Info.plist fragment	✅	✅
Entitlements fragment	✅	✅
CLI (parse/compile/validate)	✅	✅
Return-type inference	✅	🟡 v0.3
MCP server	✅	🟡 v0.3

Development

pip install -e '.[dev]'
pytest -v
ruff check .
mypy axintai

License

Apache 2.0 — see LICENSE.

Part of the Axint project by Agentic Empire.