Comwatt Python Client

Overview

The Comwatt Python Client is a Python library that provides a convenient way to interact with the Comwatt API. It allows you to authenticate users, retrieve authenticated user information, and access site and device data.

Please note that the Comwatt client is exclusively for gen4 devices: it use energy.comwatt.com/api. Versions below gen4 will not be compatible. If you're looking for the go.comwatt.com/api go to python-comwatt-client-legacy.

Features

The client currently supports the following methods:

• authenticate(self, username, password): Authenticates a user with the provided username and password. The client keeps the session alive automatically: it re-authenticates proactively before a request once the session expiry (session_expires_at) is within 30 s, and falls back to re-authenticating on a 401 response, retrying the request once; pass ComwattClient(auto_reauth=False) to disable both.
• is_authenticated(self): Returns whether the current session is still valid (probes the API; True/False, re-raises unexpected errors).
• session_expires_at: Property — the instant the current session cookie dies, as a UTC-aware datetime (captured from the login response's x-cwt-token header; login time + 2 h). Refreshed on every (re-)authentication, reset to None on logout(); None before the first login or if the server omits the header.
• logout(self): Issues POST /v1/logout, which expires the cwt_session cookie client-side (Set-Cookie 1970) and drops it from the local jar — but does not invalidate the session server-side (live-verified: the old cookie keeps working until its natural expiry, ~login + 2 h). Also clears the stored credentials so auto_reauth cannot silently restore the session. Idempotent (a 401 when already logged out is a no-op). Note this is distinct from close(), which only releases the local HTTP session without logging out.
• get_authenticated_user(self): Retrieves information about the authenticated user.
• get_sites(self): Retrieves a list of sites associated with the authenticated user.
• get_site(self, site_id): Retrieves a single site. Takes the numeric site id (from site["id"]), not the short siteUid string.
• get_site_by_uid(self, site_uid): Retrieves a single site. Takes the short siteUid string (from site["siteUid"]), not the numeric site id.
• get_site_networks_ts_time_ago(self, site_id, measure_kind = "FLOW", aggregation_level = "NONE", aggregation_type = None, time_ago_unit = "HOUR", time_ago_value = 1, start = None, end = None): Retrieves the time series data for the networks of a specific site, based on the provided parameters. Deprecated — use get_site_time_series instead (emits a DeprecationWarning; the endpoint still works but the app has moved to site-time-series).
• get_site_consumption_breakdown_time_ago(self, site_id, aggregation_level = "HOUR", time_ago_unit = "DAY", time_ago_value = 1, start = None, end = None) Retrieves the consumption breakdown data for a specific site, based on the provided parameters. Deprecated — use get_top_consumption instead (emits a DeprecationWarning; the endpoint still works but the app has moved to top-consumption).
• get_devices(self, site_id): Retrieves a list of devices for the specified site.
• get_connected_objects(self, site_id=None, gateway_uid=None): Retrieves the connected objects for a site or a gateway. Exactly one of site_id / gateway_uid is required (raises ValueError otherwise).
• get_connected_object(self, connected_object_id): Retrieves information about a specific connected object.
• get_connected_object_devices(self, connected_object_id): Retrieves the devices exposed by a specific connected object (GET /connectedobjects/{id}/devices) — narrower than get_devices(site_id). Full device documents (capacities, deviceKind, site, uuid, ...; live-verified 2026-10-27).
• get_gateway(self, gateway_id): Retrieves a gateway (the Comwatt box) by its numeric id (gateway["id"]) — the gatewayUid string is rejected with 400. Returns the full gateway document (JSON-with-refs, returned as-is); connected / updating / status are the "box offline / being updated" flags.
• get_gateway_by_uid(self, gateway_uid): Retrieves a gateway by its opaque gatewayUid string. Returns the same document as get_gateway.
• get_gateway_network(self, gateway_uid): Retrieves the network details of a gateway (signal strength, channel, ...). ⚠ Takes the gatewayUid, not the numeric id — the numeric id answers 412 gateway.not.found.for.fetching.network.details (bare JSON array error body).
• get_gateway_ssids(self, gateway_id): Retrieves the Wi-Fi networks visible to a gateway, by numeric id. ⚠ Privacy-sensitive: the response includes neighbouring Wi-Fi networks, not just the gateway's own.
• get_measure_keys(self, site_id): Retrieves the measure keys (flat measurement inventory) for a site — each a (device, measureKind) pair with a stable id and measureKey UUID.
• get_tiles(self, site_id): Retrieves the dashboard tile configuration for a site (tile type + which device each points at; configuration only, no live values).
• get_electricity_price(self, site_id): Retrieves the EDF Tempo calendar / tariff structure for a site (tempoSyntheses, daily, ...).
• get_device_kinds(self, site_uid): Retrieves the device-kind catalogue for a site (the "add a device" picker). Takes the short siteUid string (from site["siteUid"]), not the numeric site id.
• get_products(self): Retrieves the Comwatt hardware catalogue (MONITOR_GEN_4, POWER_GEN_4, ...; {id, name, brand, cardinality, features}).
• get_product(self, product_id): Retrieves a single catalogue product by its numeric id — same shape as the get_products() entries.
• get_timezones(self): Retrieves the static timezone list ({timeZone, posix} pairs: IANA name + POSIX TZ string).
• get_modbus_configurations(self): Retrieves the Modbus configuration catalogue (one entry per supported Modbus device family: {id, name, label, port, slave, device_kinds} — note device_kinds is snake_case). Modbus pairing itself is an installer flow and is not exposed.
• get_plannings(self, site_id): Retrieves the plannings (schedules) for a site. Auto-paginates through the content/totalPages envelope and returns a flat list. The server currently ignores siteId (returns the authenticated user's plannings); the trailing-slash variant 404s, and the ?deviceId= variant is deliberately not offered.
• update_planning(self, planning_id, payload): Updates a planning via PUT /plannings/{planning_id} with the payload sent as-is. Second-hand pitfalls (not live-tested): the inline device object must carry "@class": "Device" plus a numeric id (else 400 "Failed to read request"), each schedule's typicalDay must be a full inline object (id-only → 500), and the PUT replaces typicalDaySchedules wholesale — include every surviving schedule and re-read after writing; optimalPlanning: true schedules are server-managed and not writable.
• get_typical_days(self, site_id): Retrieves the "typical day" templates for a site (ON / OFF / COMWATT time ranges driving the auto-consumption optimizer). Auto-paginates and returns a flat list.
• create_typical_day(self, site_id, payload): Creates a "typical day" template for a site (POST /api/typicaldays?siteId=). ⚠ siteId goes in the query string, not the body — second-hand from the solareco reverse-engineering, not live-tested. The payload is passed through unvalidated: build it from get_typical_days output; the response shape is not live-verified.
• update_typical_day(self, typical_day_id, payload): Updates an existing "typical day" template (PUT /api/typicaldays/{id}). The payload is passed through unvalidated — build it from get_typical_days output; the response shape is not live-verified.
• get_alert_configs(self, device_id): Retrieves the alert config (consumption threshold) of a device (GET /alertconfigs/v2?deviceId=). Returns a single {period, id, min, max, activated, deviceId} object; devices without a configured alert answer 200 with a skeleton object (id/min/max/deviceId null, activated false) instead of 404. min and max are independently nullable; observed period: SLIDING_24_HOURS. device_id is mandatory server-side (bare GET → 400).
• set_alert_config(self, device_id, min, max, period = "SLIDING_24_HOURS", activated = True): Sets the alert config (consumption threshold) of a device. Two-step flow: resolves the numeric config id via get_alert_configs, then sends PUT /alertconfigs/v2/{id} with {deviceId, min, max, period, activated}. Write body shape bundle-derived, not live-verified. Raises ComwattAPIError when the device has no configured alert (skeleton id null) — the creation path (POST) is not exposed. The min/max parameter names shadow the builtins to mirror the API body fields.
• get_thermal_control(self, device_id): Retrieves the thermal-control configuration (set-point / mode) of a device (GET /thermalcontrol?deviceId=). Returns None when the API answers 200 with an empty body (non-thermal devices). The payload shape is unverified — no thermal/SG-Ready equipment on the probe site, so the keys are presumed rather than observed.
• set_thermal_control(self, thermal_control_id, device_id, set_point = None, mode = None): Updates the thermal-control configuration (set-point / mode) of a device (PUT /thermalcontrol/{id}). thermal_control_id is the thermal-control config id (a device id 404s), obtained from get_thermal_control(device_id); validation runs before the path lookup. The body's inline device object is the only required field and must carry the Jackson discriminator "@class": "Device" + the numeric id (else 400 "Failed to read request"); set_point / mode are optional and only sent when not None. Success path unverified (no thermal-control config on the probe account); mode values unconfirmed at this endpoint (capacity-level thermal-mode enum: OFF/HEAT/COOL).
• get_device_ts_time_ago(self, device_id, measure_kind = "FLOW", aggregation_level = "HOUR", aggregation_type = "MAX", time_ago_unit = "DAY", time_ago_value = 1, start = None, end = None): Retrieves the time series data for a specific device, based on the provided parameters.
• get_site_time_series(self, site_id, measure_kind = "FLOW", aggregation_level = "HOUR", aggregation_type = None, time_ago_unit = "DAY", time_ago_value = 1, start = None, end = None): Retrieves the whole-site rollup time series data for a specific site, based on the provided parameters.
• get_site_network_time_series(self, site_id, site_network_type, measure_kind = "FLOW", aggregation_level = "HOUR", aggregation_type = None, time_ago_unit = "DAY", time_ago_value = 1, start = None, end = None): Retrieves the time series of one of a site's virtual networks — PRODUCTION / CONSUMPTION / INJECTION / WITHDRAWAL / CHARGE / DISCHARGE (battery flows via the directional CHARGE / DISCHARGE; BATTERY and GRID are rejected with 400). Site-addressed siteNetworkType overload of the device time-series endpoint.
• get_devices_ts_time_ago(self, device_ids, *, max_workers = None, **kwargs): Fans get_device_ts_time_ago out over a concurrent.futures.ThreadPoolExecutor (one call per device, default max_workers = min(8, len(device_ids))). Results are returned in device_ids order; **kwargs are forwarded to every call. If one or more devices fail, all calls are still awaited and the first failure in device_ids order is raised. Returns [] for an empty device_ids without starting any threads.
• get_top_consumption(self, site_id, aggregation_level = "DAY", time_ago_unit = "DAY", time_ago_value = 1, start = None, end = None): Retrieves the per-device consumption breakdown (top 5 devices + "others") for a specific site.
• get_ecowatt(self): Retrieves the RTE EcoWatt grid-stress forecast (array of daily entries with a GREEN/ORANGE/RED status and 24 hourly values). Takes no parameters.
• client.weather.forecast(*, city=None, zip=None, country='FR', units='metric', lang='fr'): Retrieves the OpenWeatherMap 5-day/3-hour forecast through the unauthenticated same-origin weather proxy (no authenticate() needed). Exactly one of city / zip (raises ValueError otherwise). Returns the raw OWM payload (cnt=40 entries + city block). Note it lives outside /api.
• switch_capacity(self, capacity_id, enable): Turns a switch/relay capacity on or off (PUT /capacities/{id}/switch). Takes the numeric capacity id.
• set_pilot_wire(self, capacity_id, state): Sets the pilot-wire order of a heating capacity (PUT /capacities/{id}/pilot-wire). Requires a "select" capacity (else 412 capacity.is.not.a.select). state is not server-side enum-bound — the 2026-10-27 binding sweep passed any string — so valid orders remain unverified (no pilot-wire hardware on the probe account); check the capacity's selectValues.
• set_thermal_mode(self, capacity_id, state): Sets the thermal mode of a thermostat capacity to OFF, HEAT or COOL (PUT /capacities/{id}/thermal-mode). state is server-side enum-bound (live-pinned 2026-10-27 by a sweep of 42 candidates): any other value is rejected with 400 "Failed to convert 'state'". Requires a "select" capacity (else 412 capacity.is.not.a.select); check the capacity's selectValues for the values valid on your hardware.
• set_thermostat_set_point(self, capacity_id, value): Sets the target set-point of a thermostat capacity (PUT /capacities/{id}/thermostat-set-point). value (temperature) is passed through as-is.

start/end accept a datetime or ISO-8601 string and select an absolute window instead of the relative time_ago_* params; a naive datetime is treated as UTC:

from datetime import datetime

client.get_device_ts_time_ago(
    "device-1",
    start=datetime(2026, 7, 4, 0, 0, 0),
    end=datetime(2026, 7, 5, 0, 0, 0),
)

• get_device(self, device_id): Retrieves information about a specific device.
• get_device_capacities(self, device_id): Retrieves the capacities of a specific device as full objects (no @ref indirection, unlike the site-wide listing). Each capacity has two ids: the numeric id (expected by switch_capacity / set_pilot_wire / ...) and the opaque capacityId string (echoed in WebSocket frames).
• put_device(self, device_id, payload): Updates a specific device with the provided payload.
• set_device_mode(self, device_id, mode): Sets the control mode of a specific device (PUT /devices/{id}/mode) — the "Off / On / Auto" control: ON / OFF / COMWATT (COMWATT = solar-driven auto, same family as the planning modes). The mode is sent as a JSON body {"mode": mode} (bundle-derived and not live-verified; if the API answers 400, it may instead expect a query parameter). Mode values were observed read-only in typical-day / planning modes; no write-based enum validation was performed.
• stream_measurements(self, site, *, reconnect=False, reconnect_backoff=1.0, reconnect_backoff_max=60.0, reconnect_max_attempts=None): Streams live measurements (FLOW / STATE) for a single site over STOMP-over-WebSocket. Takes one site dict (as returned by get_sites); yields Measurement / CapacityChanged events. Requires the optional [stream] extra (websocket-client). Pass reconnect=True to enable opt-in exponential-backoff reconnection on socket drops; reconnect_backoff sets the initial delay (seconds), reconnect_backoff_max caps the delay, and reconnect_max_attempts limits consecutive failed connect attempts before raising (None = unlimited). Reconnect is off by default; an auth rejection is always terminal. The server declines STOMP heart-beats, so the client sends the raw __ping__ text as a keep-alive every 60 s (mirroring the SPA); silent-socket timeouts no longer end the stream. Teardown (generator exhaustion, close(), or socket death) best-effort SENDs /app/streaming/stop before the STOMP DISCONNECT, mirroring the SPA. Before each WebSocket handshake (initial and reconnect), the client proactively refreshes the session when it is within 30 s of expiry (auto_reauth + stored credentials); a handshake 401 still falls back to reactive re-auth.

Installation

You can install the Comwatt Python Client using pip. Run the following command:

pip install comwatt-client

Usage

Here's a simple example of how to use the Comwatt Python Client:

from comwatt_client import ComwattClient

# Create a Comwatt client instance
client = ComwattClient()

# Authenticate the user
client.authenticate('username', 'password')

# Get information about the authenticated user
user_info = client.get_authenticated_user()
print(user_info)

# Get a list of sites associated with the authenticated user
sites = client.get_sites()
print(sites)

# Get the whole-site rollup time series (productions, consumptions, injections, ...)
site_time_series_data = client.get_site_time_series(sites[0]['id'])
print(site_time_series_data)

# Get the per-device consumption breakdown for a specific site (top 5 + "others")
top_consumption_data = client.get_top_consumption(sites[0]['id'])
print(top_consumption_data)

# Get a list of devices for a specific site
devices = client.get_devices(sites[0]['id'])
print(devices)

# Get time series data for a specific device
time_series_data = client.get_device_ts_time_ago(devices[0]['id'])
print(time_series_data)

# Set the control mode of a specific device to MANUAL
device = client.get_device(devices[0]['id'])
device['configuration']['controlMode'] = 'MANUAL'
client.put_device(device['id'], device)

# Switch the POWER_SWITCH capacity
for feature in device['features']:
    for capacity in feature['capacities']:
        if capacity.get('capacity', {}).get('nature') == "POWER_SWITCH":
            capacity_id = capacity['capacity']['id']
client.switch_capacity(capacity_id, False)
client.switch_capacity(capacity_id, True)

Make sure to replace 'username', 'password' with the actual values for your Comwatt account.

Error handling

All errors raised by the client subclass ComwattError:

• ComwattAuthError — the credentials were rejected (HTTP 401/403 on login) or the session expired (HTTP 401). Re-authenticate.

• ComwattAPIError — any other unexpected HTTP status, including non-credential failures of authenticate() (e.g. a 5xx during an outage, or a 200 response missing the session cookie). Exposes .status_code, .url, .detail.

  detail is extracted from the API's two error body shapes: the RFC-7807 detail field, and — for the bare JSON array of Spring validation errors some endpoints return instead (e.g. 412, also seen on 404/403) — each entry's code (falling back to codes[0], then defaultMessage), comma-joined.

from comwatt_client import ComwattClient, ComwattAuthError, ComwattAPIError

client = ComwattClient()
try:
    client.authenticate("username", "password")
    sites = client.get_sites()
except ComwattAuthError:
    ...  # credentials wrong or session expired — re-authenticate
except ComwattAPIError as e:
    print(e.status_code, e.url, e.detail)

Realtime streaming (optional)

For live measurements, install the optional streaming extra (adds websocket-client):

pip install comwatt-client[stream]

Then iterate a single site's measurements as they arrive over STOMP-over-WebSocket:

from comwatt_client import ComwattClient

client = ComwattClient()
client.authenticate("username", "password")

sites = client.get_sites()
for ev in client.stream_measurements(sites[0], reconnect=True):
    print(ev)

The generator yields Measurement (live FLOW / STATE values) and CapacityChanged (switch/state changes) events. By default (reconnect=False) it stops when the WebSocket drops. Pass reconnect=True to reconnect automatically with full-jitter exponential backoff:

for ev in client.stream_measurements(
    sites[0],
    reconnect=True,
    reconnect_backoff=1.0,
    reconnect_backoff_max=60.0,
    reconnect_max_attempts=10,
):
    print(ev)

The delay between failed connection attempts starts at reconnect_backoff seconds and doubles up to reconnect_backoff_max. reconnect_max_attempts counts consecutive failed attempts (resets to zero after any successful connection). An auth rejection (HTTP 401/403) is always terminal and will not be retried. QUANTITY measures are emitted only on bucket rollover, so keep a slow REST fallback (e.g. get_site_time_series) for absolute totals.

Mapping streamed measurements to devices

The stream yields measurements keyed by an opaque capacityId string. Sensor and switch capacities use AZUREIOT-co.{N}.instances.{i}.{sensor|switch}.{j}.data; grid-exchange and battery capacities keep the sensor.{j} segment and append the quantity instead — ...sensor.{j}.withdrawal.data, ...sensor.{j}.injection.data, ...sensor.{j}.battery_charge.data, ...sensor.{j}.battery_discharge.data. To route them to a device and interpret them, join them against get_connected_objects(site_id):

• capacityId ↔ capacity object — Measurement.capacity_id is the capacityId string found on each capacity object. That same object also carries deviceId, nature (e.g. CLAMP, POWER_SENSOR, POWER_SWITCH), measureKind, and production (bool) — enough to route a measurement to its device and determine its sign with no separate map.

• Multi-instance / polyphase devices — a device can expose several SENSOR capacities (a 3-phase solar inverter pushes instances.0, instances.1, instances.2). The device's instantaneous power is the sum of its FLOW capacities, not any single instance. Polyphase capacities carry a phase field such as POSITIVE_1 / POSITIVE_2 / POSITIVE_3 (not every capacity has one).

• Typed value fields — Measurement parses the raw string into value_float (when it is numeric — watts for FLOW) and value_bool (when it is true/false — on/off for STATE). In practice FLOW populates value_float and STATE populates value_bool; Measurement.value is the raw string the server sent — prefer the typed field.

• Cadence — the server pushes measurements in bursts roughly every 15 seconds, grouped per connected object (the co.{N} segment of the capacityId). It is not a fixed per-device poll: expect a burst of many measures close together, then a gap.

• Use get_connected_objects(), not get_devices(), to build the route — get_connected_objects(site_id) returns each connected object (the co.{N} segment, via its coId) with all its capacities, each carrying capacityId / deviceId / nature / measureKind. get_devices() returns the user-facing device list, but the grid-meter and battery parents come back with empty features and capacities, so their routable capacities are not exposed there.

  This is the only way to map grid exchange and battery: ...sensor.{j}.withdrawal.data routes to a WITHDRAWAL part-child device, ...sensor.{j}.injection.data to an INJECTION one, ...sensor.{j}.battery_charge.data to a BATTERY_CHARGE device, and ...sensor.{j}.battery_discharge.data to a BATTERY_DISCHARGE device. These are real devices — part-children of the GRID_METER / BATTERY parent (identified by deviceKind.code, not a nature field) — that also show up in get_measure_keys(site_id). Some physical clamps on a multi-channel box are present as capacities but have deviceId: null until they are assigned to a device — drop those from the route.

• Switch state — a device with a POWER_SWITCH / RELAY capacity receives its on/off state on the stream as STATE measurements on a ...switch.N.data capacityId, parallel to the ...sensor.N.data FLOW capacityId on the same instance.

• CapacityChanged vs the route — CapacityChanged is also yielded for switch state changes, but its capacity_id is the numeric capacity id (the one switch_capacity takes), not the capacityId string — don't match it against the string route. Switch changes were not observed during this check (no switch was toggled manually).

Building the capacityId → (deviceId, nature, production) route from get_connected_objects() (covers every mappable capacity, including grid exchange and battery):

from comwatt_client import ComwattClient, Measurement

client = ComwattClient()
client.authenticate("username", "password")
sites = client.get_sites()

route = {}
for obj in client.get_connected_objects(site_id=sites[0]["id"]):
    for cap in obj["capacities"]:
        device_id = cap["deviceId"]
        if device_id is None:
            continue
        route[cap["capacityId"]] = (device_id, cap["nature"], cap.get("production", False))

for ev in client.stream_measurements(sites[0], reconnect=True):
    if isinstance(ev, Measurement):
        device_id, nature, production = route.get(ev.capacity_id, (None, None, None))
        if device_id is not None:
            print(f"device={device_id} nature={nature} production={production} "
                  f"value={ev.value_float if ev.value_float is not None else ev.value_bool}")

Architecture

ComwattClient composes one mixin per API domain (auth, sites, devices, gateways, ...), each contributing its get_* / set_* methods flat on the client instance. The single exception is client.weather, a namespaced proxy.

The rule deciding the split:

A domain gets a namespaced proxy if and only if its transport differs from the authenticated REST path — a different base URL/host, protocol, or auth model. Method count is never a trigger.

• Weather is a proxy because it hits an unauthenticated OpenWeatherMap forecast proxy outside /api, while reusing the client's HTTP session.
• Streaming is transport-divergent too (WebSocket, different host) but stays a flat method (stream_measurements): it is authenticated and session-coupled. It becomes a proxy only if a second streaming method ever lands.

Related projects

Actively maintained projects worth knowing about. This is deliberately not an inventory of everything Comwatt-related on GitHub — most of it is abandoned or pre-gen4, and pointing you at a dead scraper would be worse than saying nothing.

• homeassistant-comwatt — Home Assistant integration built on this library (gen 4). If domotics is your goal, install this via HACS rather than wiring the client up yourself.
• comwatt_indepbox — Home Assistant integration for pre-gen4 boxes (go.comwatt.com). Useful if you landed here with older hardware and want a working integration rather than a Python library.
• solareco — Kotlin Multiplatform app (Android / desktop) with its own gen-4 client. The only other actively maintained implementation, and a handy second opinion when an undocumented endpoint changes behaviour.

API field notes

The API this client talks to is private and undocumented, so docs/energy.comwatt.com/ collects reverse-engineered notes on it: authentication, the aggregation/time-series endpoints, capacities, the STOMP-over-WebSocket stream, and the gotchas that bite implementors (two ids per site, @ref graphs, mandatory aggregationLevel, ...). They are best-effort and dated — verify anything load-bearing against the live API.

Contributing

Contributions to the Comwatt Python Client are welcome! If you find any issues or have suggestions for improvement, please open an issue or submit a pull request on the GitHub repository.