Authoring intents in Python

The Python authoring API mirrors the TypeScript surface exactly — same parameter types, same field names (with snake_case where appropriate), same semantics.

define_intent

from axint 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"),
    },
    entitlements=["com.apple.developer.calendars"],
    info_plist_keys=["NSCalendarsUsageDescription"],
    is_discoverable=True,
)

All arguments except name, title, description, and domain are optional.

param helpers

Python	TypeScript	Swift type
`param.string(desc)`	`param.string(desc)`	`String`
`param.int(desc)`	`param.int(desc)`	`Int`
`param.double(desc)`	`param.double(desc)`	`Double`
`param.float(desc)`	`param.float(desc)`	`Float`
`param.boolean(desc)`	`param.boolean(desc)`	`Bool`
`param.date(desc)`	`param.date(desc)`	`Date`
`param.duration(desc)`	`param.duration(desc)`	`Measurement<UnitDuration>`
`param.url(desc)`	`param.url(desc)`	`URL`
`param.number(desc)`	`param.number(desc)`	`Int` (legacy alias)

Every helper accepts optional=True and default=<value>:

param.boolean("Whether to include attachments", optional=True, default=False)

Multiple intents per file

Python intent files can define multiple intents — axint parse will pick up every top-level define_intent(...) assignment:

from axint import define_intent, param

send = define_intent(name="SendMessageIntent", title="Send", description="...", domain="messaging")
reply = define_intent(name="ReplyIntent", title="Reply", description="...", domain="messaging")

Return the IR

from axint import define_intent

intent = define_intent(name="X", title="X", description="...", domain="x")
ir = intent.to_ir()        # -> IntentIR
print(ir.to_dict())        # -> camelCase JSON matching the TS schema

The to_dict() output uses camelCase keys (isDiscoverable, infoPlistKeys) because that’s the cross-language contract shared with the TypeScript compiler.