safera-sense-ble

Async Python library for talking to Safera Sense cooking sensors over Bluetooth Low Energy — as found in Røros Hetta cooker hoods (Safera Sense Integral) and Safera stove guards.

Credits

Most of the protocol reverse engineering comes from two projects that did the heavy lifting:

• magicus/safera-ble — the byte-level protocol documentation, derived from packet captures and analysis of the decompiled Android app;
• havardgulldahl/rorossense-ble — a working Python client and further protocol exploration against a RørosHetta hood, including the consolidated protocol docs.

This library is an independent, from-scratch implementation of that protocol, extended with findings made during its development (verified on an IFU10CR-PRO, firmware 13/75):

• the proprietary characteristics require BLE bonding — the device answers ATT "Insufficient authentication" until paired; this library pairs on demand and retries;
• grease filter saturation lives at byte 59 of the extended sensor report and resets via the SET_HOOD_FILTER_CHANGED command (confirmed with a before/after diff around the vendor app's filter reset);
• the "particle index" field tracks the vendor app's PM2.5 reading (µg/m³, 0.2 µg/m³ raw resolution).

Unofficial project — not affiliated with Safera Oy or Røros Metall AS.

Features

• Live sensor stream: subscribe to notifications (~1 Hz) parsed into a typed SensorReport — ambient/surface/pan temperature, humidity, ambient light, eCO2, tVOC, air quality index, PM2.5, stove power draw, cooking activity, alarm level, grease filter saturation, device state and error bitfields.
• Control: hood fan speeds 1–4 (level 4 = boost), auto mode; light levels 1–3; identify; grease-filter reset.
• Smart Cooking event log: read or subscribe to the timeline of cooking events (cooking/frying/boiling/heating started, stove alarms, button presses); clear it too.
• Device info: model, serial, hardware/firmware revisions, Wi-Fi status (SSID, RSSI, device name).
• Sensible precision: values are quantized to each sensor's meaningful resolution (0.1 °C temperatures, whole-percent humidity, whole lux, 0.1 µg/m³ PM2.5) so consumers don't see measurement jitter as state changes.
• Built on bleak and bleak-retry-connector; plays well with Home Assistant's Bluetooth stack but has no Home Assistant dependency.

Installation

pip install safera-sense-ble

Usage

Find the device

Safera devices advertise the proprietary service UUID and, depending on branding, a name like Røroshetta, iSense… or Sense_…:

import asyncio
from bleak import BleakScanner
from safera_sense_ble import SAFERA_SERVICE_UUID

async def find():
    devices = await BleakScanner.discover(return_adv=True)
    for device, adv in devices.values():
        if SAFERA_SERVICE_UUID in adv.service_uuids:
            print(device.address, adv.local_name)

asyncio.run(find())

Stream sensor data

import asyncio
from bleak import BleakScanner
from safera_sense_ble import SaferaSenseClient, SensorReport

async def monitor(address: str):
    ble_device = await BleakScanner.find_device_by_address(address)
    client = SaferaSenseClient(ble_device)
    await client.connect()

    info = await client.fetch_device_info()
    print(f"Connected to {info.model} (fw {info.firmware_rev})")

    def on_report(report: SensorReport) -> None:
        print(
            f"{report.ambient_temperature:.1f} °C  "
            f"{report.humidity} %RH  "
            f"eCO2 {report.co2_ppm} ppm  "
            f"PM2.5 {report.particle_index} µg/m³  "
            f"filter {report.grease_filter} %"
        )

    # The device requires bonding for this; the client pairs on demand.
    await client.subscribe_sensor_reports(on_report)
    await asyncio.sleep(30)
    await client.disconnect()

asyncio.run(monitor("D4:6A:C8:XX:XX:XX"))

Control the hood

from safera_sense_ble import FanSpeed, LightLevel

await client.set_fan_speed(FanSpeed.LEVEL_2)   # speeds 1-4
await client.set_fan_speed(FanSpeed.BOOST)     # top step (level 4)
await client.set_fan_auto()                    # air-quality controlled
await client.set_fan_speed(FanSpeed.OFF)

await client.set_light_level(LightLevel.LEVEL_3)  # light levels 1-3
await client.set_light_level(LightLevel.OFF)
await client.toggle_light_auto()               # toggle presence-based auto

await client.identify()                        # make the device identify itself
await client.reset_grease_filter()             # after cleaning the filter

One-shot reads

report = await client.fetch_sensor_report()    # single parsed snapshot
wifi = await client.fetch_wifi_status()        # SSID, RSSI, device name, IP

Smart Cooking event log

from safera_sense_ble import CookingEvent

events = await client.fetch_event_log()        # current timeline (newest first)
for e in events:
    print(e.name, e.timestamp)                 # e.g. "frying_start", device clock

def on_events(events: list[CookingEvent]) -> None:
    if events:
        print("latest:", events[0].name)

await client.subscribe_event_log(on_events)    # notified on new events
await client.clear_event_log()                 # the app's "Clear Timeline"

Event timestamps are in device-clock seconds; correlate them with the live SensorReport.device_clock if you need wall-clock times.

See examples/monitor.py for a runnable script.

API notes

• All methods are coroutines; the client is designed for a single long-lived connection (the device accepts one central at a time — close the vendor app while connected).
• SensorReport.from_bytes accepts both the documented 54-byte record and the extended (~69-byte) record sent by hood-integrated devices; hood-only fields (fan_speed_raw, light_raw, grease_filter, …) are None when absent.
• SaferaSenseClient.dump_characteristics() hex-dumps every readable characteristic — useful for decoding the remaining unknown fields (byte 58 of the sensor report, the event log, READ_SETTINGS, …); contributions welcome here and upstream.

License

MIT — see LICENSE.