There are a couple of important concepts in
pyatv that is good to understand. This page will cover
the most important ones, giving them relations to how they are used in code.
Devices and configuration
A physical device, e.g. an Apple TV 4K, is represented by a configuration. From a code perspective, you find all the important pieces in conf.AppleTV. It stores all the necessary information about the device it represents, e.g. name, IP-address, supported protocols, etc. You can read more about this in the following sections.
The simplest way to create a configuration is to to scan for devices, as pyatv.scan returns a list of conf.AppleTV objects with almost all information already filled in. You can then just pick the configuration you want and either connect to or pair with the device. Scanning and pairing is explained below.
It is also possible to manually create a configuration, although this is not the recommended way. One reason for this is that some information is not static. The
Media Remote Protocol uses a randomized port that might change at any time. If the currently used port is stored in a static configuration, then the configuration will not work in case the port changes. When using pyatv.scan, the correct port will be automatically identified and filled in for you.
Services and protocols
Each configuration keeps track of all the different protocols a device supports. The term service is also used in this context and it is exactly the same thing (this will likely be consolidated into a single term in the future, but for now they are used interchangeably).
pyatv supports three protocols:
|Digital Media Access Protocol (DMAP)||Control device and get metadata||Apple TV <= 3 and tvOS <= 12|
|Media Remote Protocol (MRP)||Control device and get metadata||tvOS (any version), e.g. Apple TV 4 and later|
|AirPlay||Stream video to a device||All devices|
At least one of
MRP must be present in order to connect to a device, otherwise pyatv.connect will raise an exception. Prior to version 0.4.1,
pyatv would return configurations for pure
AirPlay devices, e.g. wireless speakers. But since
pyatv can’t connect to these devices, they are no longer returned by pyatv.scan.
There are methods in conf.AppleTV to add and retrieve services. When using pyatv.scan, all discovered protocols will be added automatically and all relevant information (e.g. which port is used) is stored as well. When manually creating a configuration, you have to provide this information yourself (e.g. via conf.DmapService for
DMAP, and so on).
In some cases it’s interesting to show some more user friendly metadata, like which operating system and version a device runs. There’s no “good” way of getting this information, but
pyatv will pull bits and pieces from the metadata received during scanning to make a good guess. Information from both the main protocol (
AirPlay are used, so if you have disabled
AirPlay, you will lose some metadata (e.g. tvOS version).
An important concept for
pyatv is to be able to uniquely identify a device. You should be able to say: “I want to find and connect to this specific device” and
pyatv should be able to do that for you. This of course means that you cannot use common identifiers, like IP-address or device name, as they can change at any time. And how many devices named “Living Room” aren’t there in the world?
pyatv extracts unique identifiers from the different services it finds when scanning. You can then specify that you want to scan for a device with one of these identifiers and be sure that you find the device you expect. What is a unique identifier then? It can be anything, as long as it is unique of course. In case of
MRP, it actually exposes a property called
UniqueIdentifer. Looking at
AirPlay, you can get the device MAC-address via a property called
deviceid. In practice this means that a device has multiple identifiers and you can use any of them when scanning. There is a convencience property, conf.AppleTV.identifier, used to get an identifier. It just picks one from all of the available.
If you perform a scan with
atvremote, you can see all the available identifiers:
$ atvremote scan ======================================== Name: Living Room Model/SW: 4K tvOS 13.3.1 build 17K795 Address: 10.0.0.10 MAC: AA:BB:CC:DD:EE:FF Identifiers: - 01234567-89AB-CDEF-0123-4567890ABCDE - 00:11:22:33:44:55 Services: - Protocol: MRP, Port: 49152, Credentials: None - Protocol: AirPlay, Port: 7000, Credentials: None
Why is this concept so important then? Well, it is important that you connect to the device you expect. Especially for
MRP, where the port might change at any time and you need to be able to find your way back (reconnect) when that happens. It also makes
pyatv more reliable in environments using DHCP, which is the case in most peoples homes.
All protocols has some sort of “authentication” process and require credentials when connecting. You obtain said credentials via pairing, which is explained below. The credentials are then stored together with the protocol in a configuration.
The usual flow is:
- Perform scan
- Pick device of interest
- For each protocol, get the service and insert credentials (stored externally)
This means that
pyatv does not store credentials for you. In the future, persistent storage will be added to
pyatv (see issue #243), but for now you have to deal with this yourself.
One exception from this is
DMAP, when Home Sharing is enabled. In this case, required credentials are automatically filled in and the service is ready to use. Fact is, in most cases you can talk to a device via
MRP without using credentials at all. It is not known what the “rules” are for this though (when and what works).
To find devices, you perform a scan with the pyatv.scan function. It uses Zeroconf to discover devices, just like the Remote app/Control Center widget in iOS. You get a list of all available devices with information pre-filled and you can pretty much connect immediately (you might have to add credentials first though).
Scanning is not 100% reliable, that comes with the territory. Sometimes a device is not found, or even a service might not be found. It will happen and that is just life. Scan again and hope for the best. Usually it works just fine, so it shouldn’t be that much of an issue (but bear it in mind when writing your software).
As a workaround for scanning issues,
pyatv comes with support for unicast scanning. It allows
you to specify a list of devices to scan for. When doing so,
pyatv will not rely on multicast
but rather send a request to the devices directly. This is much more reliable but of course comes
with the downside of not having devices auto discovered. You can simply try this out with
please see atvremote.
Please note that unicast scanning does not work across different network as the Apple TV will ignore
packets from a different subnet. This is according to the multicast DNS specification (see chapter 5.5
in RFC 6762 here in case you are interested) and is
not something that can be fixed in
Deep Sleep Detection
When a device is in deep sleep mode, another node on the network called a
sleep proxy will announce its prescence. It will also wake up the device
using Wake-On-LAN in case a particular service is requested. When scanning,
pyatv can detect if the response originates from a sleep proxy and
deduce that a device is sleeping. This is indicated via the flag
Please do note that this is an experimental feature.
Pairing is the process of obtaining credentials. You provide a configuration to the pyatv.pair method together with the protocol you want to pair. Usually you have to input a PIN code and then the credentials are returned via the
credentials property in the service. This means that you can scan for a device, pass the configuration you got from pyatv.scan to pyatv.pair and finish off by passing the same configuration to pyatv.connect. As simple as that.
The final step is to connect to a device. It is done via pyatv.connect, which sets up the connection and validates the credentials (e.g. setup encryption). You will get an error if it doesn’t work. You can provide which protocol you want to use when connecting, but you don’t have to. In the latter case,
MRP will be preferred over
DMAP if both are available. If neither are present, an exception will be raised.
Metadata and push updates
The object you get when connecting follows the interface specified in the
interface module. You can get metadata, e.g. what is currently playing via the
metadata property. You don’t have to poll the device for that information, you can use the interface.PushUpdater interface to receive updates instantly as they happen via a callback interface.
Depending on hardware model and software, various features might be supported or not. Siri is for
instance only supported by devices running tvOS. Certain actions might not be possible to perform
due to the device state, e.g. pressing “pause” is not possible unless something is playing. In
it’s possible to obtain information about the availability of features using interface.Features.