Unfortunately, Ubiquiti changed their (supposedly versioned) API in
3.2.0 which causes us to have to refer to cameras by id instead of
UUID. The firmware for 3.2.x also changed the on-camera login procedures
and snapshot functionality significantly.
This bumps the requirement for uvcclient to 0.9.0, which supports the
newer API and makes the tweaks necessary to interact properly.
This makes us track WeMo devices by serial number instead of URL. Per
@jaharkes' suggestion, tracking via URL could potentially be a problem
if a device changes to another IP and then appears to be discovered again.
Since we can't track based on mac (for static-configured devices), we
can use the serial number exposed to make sure we're always looking at
a list of unique devices.
The default configuration of the mFi controller generates self-signed
certificates which are valid for short periods of time, and which are
regnerated on start or as needed. This makes requests mad. Since most
people will be using the self-signed certificates anyway, add options
to let them choose non-TLS, or unverified connections if they want/need.
This bumps the mficlient requirement to 0.3.0 for proper handling of
verify=False.
I broke this with my fix, where I assumed that we'd always have a wemo
config section if we're running the wemo code. However, if we're fully
auto-detected, we might not.
I broke this with my fix, where I assumed that we'd always have a wemo
config section if we're running the wemo code. However, if we're fully
auto-detected, we might not.
The wemo component was tracking devices by mac to avoid adding duplicates,
but this means we'd never be able to load more than one static wemo, since
they all have mac=None.
This changes the code to track by url, which has to be unique per wemo
anyway, and which we also have for even static wemos.
The wemo component was tracking devices by mac to avoid adding duplicates,
but this means we'd never be able to load more than one static wemo, since
they all have mac=None.
This changes the code to track by url, which has to be unique per wemo
anyway, and which we also have for even static wemos.
The exception we were catching incorrectly referenced the client variable
in local scope instead of the module. Also, if we fail to connect we can
get a requests exception, so catch and handle that as well.
The older (unsupported AirCam) models behave differently and also apparently
suffer some under the last release of the NVR that supported them. Since they
are EOL and not supported by current software, filter them out so we don't
break while trying to extract an image from them.
This adds tests for the uvc camera module. It's a good thing too,
because I found a few bugs which are fixed here as well:
- Graceful handling of non-integer port
- Failure to take the first host that works when probing host,internalHost
- Failure to detect if neither of them actually work
This also converts the code to only call add_devices once with a listcomp.
The NVR tells us the admin username, but not the password for the
camera. Right now, we assume the default password, which obviously
doesn't work for people that have changed it. The uvcclient library
provides a way to set the cached admin password for a camera, which
is stored in a client-specific location. We can utilize that to
grab the password, falling back to the default if it's unset. With
this, people just need to run a command per camera to set the
admin password on their systems, if it has changed.
Ubiquiti's Unifi WAP infrastructure has a central controller (like mfi and uvc)
that can be queried for client status. This adds a device_tracker module that
can report the state of any client connected to the controller.
This adds a 'sensor_class' property and attribute, which should be either
None or one of several defined SENSOR_CLASSES to indicate contextual
information about what the sensor is measuring.
This adds verbose debugging which can be turned on to figure out what is
going on. It also adds a broad exception handler in the worker thread
to avoid dying. If you're running this such that stderr doesn't go to a
log, it can be easy to miss the thread's death.
I wrote all this to try to diagnose #1283, which seems to maybe have
healed itself. But since I have it, I figure we might as well keep it
in case we have trouble in the future.
This bumps the somecomfort requirement to 0.2.1 to pull in a change
that makes handling no-fan systems graceful. Adds a test that should
prove it gives us what we want.
If no fan, then fan is always idle and fanmode is None.
This allows you to get every door, window, smoke, etc zone from your security
panel into HA. This uses the live eventing feature of pynx584, which means you
get instantaneous signaling into HA when a door opens or something happens,
which is handy for automating lights on when doors open after dark, etc.
Requires update to pynx584 0.2
This makes the UVC camera_image() method not log into the camera on
every single image fetch, which reduces load on hass and the camera,
and lowers the latency to get an actual image fetched.
This lets you optionally only add thermostats by location or
specific device id, instead of all the thermostats in your
account. This would be helpful if you have two devices in different
houses (i.e vacation home), etc.
This adds tests for the influxdb component. It also fixes a bug,
where username and password are required, but not gracefully
handled if they're missing from config.
This also fixes issue #1214, and I think another bug. The splunk
code will just take the value of state.state and try to serialize
it to json if it can't make it into a number. It did this before
I generalized that code. Since json.dumps() will fail on most anything
complicated, I think the right thing to do is *not* try to do that.
This adds state_as_number(), a helper method that tries to interpret
state as a number, for cases we can predict. It's a generalization of
what is copy-and-paste-ed into multiple other places.
This is extremely useful if you want to support wemos that are on
another subnet or across a VPN. It also lets you sidestep the discovery
process, which is problematic for a lot of people and situations.
In order for this to work, we need to bump the pywemo requirement to
0.3.10, which includes my changes to make this possible.
WeMo devices can be manually configured by adding a static section to
the config, like this:
switch:
platform: wemo
static:
- 192.168.100.5
- 192.168.100.6
This extends the HoneywellUSThermostat functionality to find and add
all thermostats in your account. So, we add a new config element called
'region' that is the primary trigger for this, and remove the 'id'
trigger since it was never in a released version.
This does a few extra things:
- It names the thermostat what you have it named in your account,
which is not something we get to know *unless* we emumerate.
- It makes all thermostats on a given account use the same session,
and thus we have to avoid doing an explicit login every time we
refresh our data. That was causing some rate-limiting on their
side when I was debugging, so this is probably good. Now, we use
their existing keepalive pinger to determine if we're still logged
in and only re-login if we need to.
Like recorder, this component listens to all events and reports any
that it can to a graphite installation. This makes it easy to use
graphite for all your data collection and analysis. If you run
carbon-cache (the backend for graphite) on the local machine, no
configuration is required other than enabling the component.
For more info on graphite: http://graphite.wikidot.com/
Most of the mFi sensors are able to reasonably provide accurate
readings to a tenth of a unit or so. This patch rounds them for
better display in the UI. Normally, I would expect this to be a view
action instead of altering the actual data emitted, but since these
values are reasonable for sensor precision, we're not really losing
anything. I followed the model from the openweathermap component, which
rounds for readability in the backend.
When we update the mFi server for the state of a switch, the new
state is not always reported immediately if we update right after
the action (the server is not RESTful).
This patch adds some internal target-state handling to report the
desired state on the next poll, allowing any subsequent polls to
override that state.
Also, bump the version requirement for mficlient to 0.2.2 to absorb
a bug fix.
This adds support for NetworX-based alarm panels and should work for
any such panel equipped with a NX584 serial interface module. This
includes NX-4/6/8/8E, where the NX8E has this interface built-in.
It requires the pynx584 module, and requires running the server
component somewhere that has connectivity to the panel via serial,
which may include a serial-over-lan connection.
This adds support for Ubiquiti's UniFi Video cameras via their
NVR device (or service). By configuring just the address of the
NVR and a valid API key, all cameras are discovered and enabled,
including direct-to-camera image snapshot-based video support.
This adds support for the US variant of the Honeywell connected
thermostat. The interface is super simple, so this doesn't add
any external dependencies. It supports basic temperature, setpoint,
and control.
Issue #998 notes that the existing honeywell module doesn't work
for US models, which is because they are totally different. In order
to indicate to the honeywell platform module that the thermostat
is a US-type, we key off of whether or not the thermostat id is
provided. This is something that US people have (and require to
identify one of potentially multiple thermostats in their account)
and EU people will not.