pyhems 0.8.0

ECHONET Lite library for Home Energy Management System (HEMS)

pyhems

ECHONET Lite library for Home Energy Management System (HEMS).

🇯🇵 日本語ドキュメント

Features

• ECHONET Lite frame encoding/decoding
• UDP multicast device discovery
• Async runtime client with event subscription
• Device state management with DeviceManager
• Adaptive poll scheduler for non-notifying properties via PropertyPoller
• Entity definitions based on MRA data
• Full type hints (py.typed)

Requirements

• Python 3.14+
• bidict>=0.23.0

License

MIT License

Installation

pip install pyhems

Quick Start

import asyncio
from pyhems import EOJ, HemsClient, HemsFrameEvent, HemsInstanceListEvent

async def main():
    client = HemsClient(interface="0.0.0.0")
    await client.start()

    def on_event(event):
        if isinstance(event, HemsInstanceListEvent):
            print(f"Node: {event.node_id}, Instances: {event.instances}")
        elif isinstance(event, HemsFrameEvent):
            print(f"Frame from {event.node_id}: {event.frame}")

    unsubscribe = client.subscribe(on_event)

    # Read properties from a discovered device
    # node_id = "fe..."  # obtained from HemsInstanceListEvent
    # props = await client.get(node_id, EOJ(0x013001), [0x80, 0xB3])

    # Write a property (example: power ON)
    # await client.set_property(node_id, EOJ(0x013001), 0x80, b"\x30")

    await asyncio.sleep(60)
    unsubscribe()
    await client.stop()

asyncio.run(main())

Runtime API Overview

• HemsClient.start() / HemsClient.stop(): Start and stop UDP transport.
• HemsClient.subscribe(callback): Subscribe to runtime events.
• HemsClient.probe_nodes(): Trigger multicast node discovery.
• HemsClient.get(node_id, deoj, epcs): Read property values.
• HemsClient.set_property(node_id, deoj, epc, edt): Write a single property.
• HemsClient.set_properties(node_id, deoj, properties): Write multiple properties.

Adaptive Property Polling

PropertyPoller uses an adaptive scheduler designed for mixed ECHONET Lite devices:

• TID-correlated in-flight tracking avoids request pileups on slow devices.
• Poll cadence is adjusted from latency EWMA and failure backoff, starting from the base interval.
• Instantaneous-value EPCs can be scheduled on a separate fast lane (fast_poll_epcs).
• When devices truncate large responses, batch size is learned (observed_batch_capacity) and reused.
• Poll targets can be narrowed dynamically through DeviceManager.subscribe_epcs(...).

Definitions

from pyhems import REGISTRY

print(REGISTRY.version, REGISTRY.mra_version)

# Mapping: class_code -> tuple[EntityDefinition, ...]
ac_entities = REGISTRY.entities.get(0x0130, ())
for entity in ac_entities[:3]:
    print(entity.epc, entity.name_en)