speechmatics.client

Wrapper library to interface with Real-time ASR v2 API. Based on http://asyncio.readthedocs.io/en/latest/producer_consumer.html

class speechmatics.client.WebsocketClient(connection_settings_or_auth_token: Optional[Union[str, speechmatics.models.ConnectionSettings]] = None)

Manage a transcription session with the server.

The best way to interact with this library is to instantiate this client and then add a set of handlers to it. Handlers respond to particular types of messages received from the server.

Parameters

connection_settings (speechmatics.models.ConnectionSettings) – Settings for the WebSocket connection, including the URL of the server.

add_event_handler(event_name, event_handler)

Add an event handler (callback function) to handle an incoming message from the server. Event handlers are passed a copy of the incoming message from the server. If event_name is set to ‘all’ then the handler will be added for every event.

For example, a simple handler that just prints out the speechmatics.models.ServerMessageType.AddTranscript messages received:

>>> client = WebsocketClient(
        ConnectionSettings(url="wss://localhost:9000"))
>>> handler = lambda msg: print(msg)
>>> client.add_event_handler(ServerMessageType.AddTranscript, handler)

Parameters

• event_name (str) – The name of the message for which a handler is being added. Refer to speechmatics.models.ServerMessageType for a list of the possible message types.

• event_handler (Callable[[dict], None]) – A function to be called when a message of the given type is received.

Raises

ValueError – If the given event name is not valid.

add_middleware(event_name, middleware)

Add a middleware to handle outgoing messages sent to the server. Middlewares are passed a reference to the outgoing message, which they may alter. If event_name is set to ‘all’ then the handler will be added for every event.

Parameters

• event_name (str) – The name of the message for which a middleware is being added. Refer to the V2 API docs for a list of the possible message types.

• middleware (Callable[[dict, bool], None]) – A function to be called to process an outgoing message of the given type. The function receives the message as the first argument and a second, boolean argument indicating whether or not the message is binary data (which implies it is an AddAudio message).

Raises

ValueError – If the given event name is not valid.

get_language_pack_info() → dict

Get the language_pack_info which is a subset of information from the manifest in the language pack which we expose to end users.

Can be None if this field has not yet been set - i.e. if the RecognitionStarted message has not been received yet.

async run(stream, transcription_config: speechmatics.models.TranscriptionConfig, audio_settings=AudioSettings(encoding=None, sample_rate=44100, chunk_size=4096), from_cli=False)

Begin a new recognition session. This will run asynchronously. Most callers may prefer to use run_synchronously() which will block until the session is finished.

Parameters

• stream (io.IOBase) – File-like object which an audio stream can be read from.

• transcription_config (speechmatics.models.TranscriptionConfig) – Configuration for the transcription.

• audio_settings (speechmatics.models.AudioSettings) – Configuration for the audio stream.

• from_cli (bool) – Indicates whether the caller is the command-line interface or not.

Raises

Exception – Can raise any exception returned by the consumer/producer tasks.

run_synchronously(*args, timeout=None, **kwargs)

Run the transcription synchronously. :raises asyncio.TimeoutError: If the given timeout is exceeded.

stop()

Indicates that the recognition session should be forcefully stopped. Only used in conjunction with run. You probably don’t need to call this if you’re running the client via run_synchronously().

update_transcription_config(new_transcription_config)

Updates the transcription config used for the session. This results in a SetRecognitionConfig message sent to the server.

Parameters

new_transcription_config (speechmatics.models.TranscriptionConfig) – The new config object.