lorawan-connection

Documentation · Building a device library

Backend-neutral LoRaWAN events and device collections for Python 3.12+. The base package has no runtime dependencies. Install the optional ChirpStack backend to connect to a server.

Feed inventory and live events into a vendor DeviceCollection. It creates supported models and reports them through device-added callbacks. Consumers then observe model state. The caller owns the network connection.

from lorawan_connection import Device, DeviceCollection, DeviceDescriptor, DeviceEvent


class Sensor(Device):
    vendor_id = 123
    catalog_model_id = "known-model"

    def __init__(self, descriptor: DeviceDescriptor) -> None:
        super().__init__(descriptor)
        self.temperature: float | None = None
        self.humidity: float | None = None

    def handle_event(self, event: DeviceEvent) -> None:
        # Decode an event, update model attributes, then call self.notify().
        pass


class Sensors(DeviceCollection[Sensor]):
    DEVICES = (Sensor,)


sensors = Sensors(network_id="my-network")
stop = sensors.subscribe_device_added(lambda device: print(device.descriptor.name))
# Feed inventory and live events into sensors.handle_event(event).

The collection builds its catalog lookup from DEVICES. You can also pass classes at construction: DeviceCollection([Sensor], network_id="my-network"). The Device base supplies identity, add_update_listener(), notify(), and cleanup. Models define their own attributes, including multiple measurements or channels. Listeners take no arguments and read model attributes after a complete update.

Install

pip install lorawan-connection

The quickstart replays a SenseCAP S2101 capture and prints 21.4 °C and 31.4% humidity.

Included

• Read-only event and payload Protocols. Generated payloads can pass by reference.
• Immutable descriptors and fixture dataclasses for every supported event payload.
• A Device base with synchronous update listeners and explicit notifications.
• A generic DeviceCollection with inventory replay, model replacement, and retirement.
• Synchronous callback helpers with independent, idempotent unsubscribe functions.
• Typed exports (py.typed), a tested SenseCAP example, and Astro/Starlight documentation.

The optional lorawan_connection.chirpstack backend supplies inventory and live events. The application owns its connection lifecycle; provisioning, QR parsing, and vendor decoders belong in separate libraries. The SenseCAP implementation under examples/ illustrates a separate device library.

Events are live notifications. The package does not persist history, reconnect a transport, or request replay after a gap. Providers report inventory before activity. Collections select models from catalog identity, never from names or payload guesses.

Sending commands

Pass an async send_downlink callback to the collection. Device models encode their commands and call async_send_downlink(data=..., f_port=...). The ChirpStack backend queues the bytes and returns a queue-item ID. Device reports update model attributes; queue acceptance does not confirm that an action happened.

The Dragino relay example controls both LT-22222-L relays independently. It uses the existing ChirpStack catalog identity and preserves the other relay when sending a command.

Device-library CLI

Pass the library's supported model classes to the shared helper:

from lorawan_connection.cli_helper import run
from my_sensors import Sensors

run(Sensors.DEVICES)

The helper discovers supported devices and prints their state. Models supply catalog identity and update listeners through the shared Device base. The CLI reads public model attributes and properties. Use --list for inventory or --json for machine-readable output.

pip install "lorawan-connection[chirpstack]"
python -m my_sensors --server https://chirpstack.example.com:443 --api-key-file /path/to/key

See the CLI guide and ChirpStack backend.

Migrating from 0.2

Rename model product_id to catalog_model_id. Replace state callbacks with add_update_listener(callback); callbacks take no arguments and read model attributes. Models inherit Device, initialize it with super().__init__(descriptor), and call self.notify() after updating their data. Declare DEVICES on the collection to replace manual lookup dictionaries.

Models now inherit Device. Pass the network ID by keyword when constructing a collection: Sensors(network_id="my-network"). Existing _create_device() overrides remain available for custom matching.

Migrating from 0.3

Inherit Device without a state type parameter. Call super().__init__(descriptor) without state=. Define your model's data as attributes or properties. An existing state object can remain a vendor-defined attribute, but the base does not require it. The CLI's JSON state field now contains the model's public attributes and properties.

Home Assistant

The Home Assistant guide covers provider subscriptions, discovery, config-entry lifecycle, and entities that observe library models. The provider API described there exists on the linked Core POC branch; it is not yet an upstream HA API.

Development

uv sync --group compatibility
uv run ruff check .
uv run ruff format --check .
uv run mypy
uv run pytest --cov --cov-report=term-missing
uv run python examples/replay.py
cd docs
npm ci
npm run build

CI tests Python 3.12, 3.13, and 3.14. Compatibility tests check the payload contracts against chirpstack-api==4.19.0. The optional backend uses the same bindings. The documentation build validates internal links and produces llms.txt.

The repository follows the packaging, documentation, and release approach of modbus-connection. The source version stays 0.0.0; the publish workflow sets the version from the release tag. See releasing for the first PyPI release setup.