Migrating to google-cloud-speech 1.0

The 1.0 release of the google-cloud-speech client is a significant upgrade based on a next-gen code generator, and includes substantial interface changes. Existing code written for earlier versions of this library will likely require updates to use this version. This document describes the changes that have been made, and what you need to do to update your usage.

To summarize:

  • The library has been broken out into multiple libraries. The new gems google-cloud-speech-v1 and google-cloud-speech-v1p1beta1 contain the actual client classes for versions V1 and V1p1beta1 of the Speech service, and the gem google-cloud-speech now simply provides a convenience wrapper. See Library Structure for more info.
  • The library uses a new configuration mechanism giving you closer control over endpoint address, network timeouts, and retry. See Client Configuration for more info. Furthermore, when creating a client object, you can customize its configuration in a block rather than passing arguments to the constructor. See Creating Clients for more info.
  • Previously, positional arguments were used to indicate required arguments. Now, all method arguments are keyword arguments, with documentation that specifies whether they are required or optional. Additionally, you can pass a proto request object instead of separate arguments. See Passing Arguments for more info.
  • Previously, the client included a method supporting bidirectional streaming recognition requests, both incremental audio and incremental results. The current client retains this method, but simplifies the interface to match streaming methods in other Ruby clients. See Streaming Interface for more info.
  • Previously, clients reported RPC errors by raising instances of Google::Gax::GaxError and its subclasses. Now, RPC exceptions are of type Google::Cloud::Error and its subclasses. See Handling Errors for more info.
  • Some classes have moved into different namespaces. See Class Namespaces for more info.

Library Structure

Older 0.x releases of the google-cloud-speech gem were all-in-one gems that included potentially multiple clients for multiple versions of the Speech service. The Google::Cloud::Speech.new factory method would return you an instance of a Google::Cloud::Speech::V1::SpeechClient object for the V1 version of the service, or a Google::Cloud::Speech::V1p1beta1::SpeechClient object for the V1p1beta1 version of the service. All these classes were defined in the same gem.

With the 1.0 release, the google-cloud-speech gem still provides factory methods for obtaining clients. (The method signatures will have changed. See Creating Clients for details.) However, the actual client classes have been moved into separate gems, one per service version. The Google::Cloud::Speech::V1::Speech::Client class, along with its helpers and data types, is now part of the google-cloud-speech-v1 gem. Similarly, the Google::Cloud::Speech::V1p1beta1::Speech::Client class is part of the google-cloud-speech-v1p1beta1 gem.

For normal usage, you can continue to install the google-cloud-speech gem (which will bring in the versioned client gems as dependencies) and continue to use factory methods to create clients. However, you may alternatively choose to install only one of the versioned gems. For example, if you know you will only use V1 of the service, you can install google-cloud-speech-v1 by itself, and construct instances of the Google::Cloud::Speech::V1::Speech::Client client class directly.

Client Configuration

In older releases, if you wanted to customize performance parameters or low-level behavior of the client (such as credentials, timeouts, or instrumentation), you would pass a variety of keyword arguments to the client constructor. It was also extremely difficult to customize the default settings.

With the 1.0 release, a configuration interface provides control over these parameters, including defaults for all instances of a client, and settings for each specific client instance. For example, to set default credentials and timeout for all Speech V1 clients:

Google::Cloud::Speech::V1::Speech::Client.configure do |config|
  config.credentials = "/path/to/credentials.json"
  config.timeout = 10.0
end

Individual RPCs can also be configured independently. For example, to set the timeout for the recognize call:

Google::Cloud::Speech::V1::Speech::Client.configure do |config|
  config.rpcs.recognize.timeout = 20.0
end

Defaults for certain configurations can be set for all Speech versions and services globally:

Google::Cloud::Speech.configure do |config|
  config.credentials = "/path/to/credentials.json"
  config.timeout = 10.0
end

Finally, you can override the configuration for each client instance. See the next section on Creating Clients for details.

Creating Clients

In older releases, to create a client object, you would use the Google::Cloud::Speech.new class method. Keyword arguments were available to select a service version and to configure parameters such as credentials and timeouts.

With the 1.0 release, use the Google::Cloud::Speech.speech class method to create a client object. You may select a service version using the :version keyword argument. However, other configuration parameters should be set in a configuration block when you create the client.

Old:

client = Google::Cloud::Speech.new credentials: "/path/to/credentials.json"

New:

client = Google::Cloud::Speech.speech do |config|
  config.credentials = "/path/to/credentials.json"
end

The configuration block is optional. If you do not provide it, or you do not set some configuration parameters, then the default configuration is used. See Client Configuration.

Passing Arguments

In older releases, required arguments would be passed as positional method arguments, while most optional arguments would be passed as keyword arguments.

With the 1.0 release, all RPC arguments are passed as keyword arguments, regardless of whether they are required or optional. For example:

Old:

client = Google::Cloud::Speech.new

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

# Config and audio are positional arguments
response = client.recognize config, audio

New:

client = Google::Cloud::Speech.speech

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

# Config and audio are keyword arguments
response = client.recognize config: config, audio: audio

In the 1.0 release, it is also possible to pass a request object, either as a hash or as a protocol buffer.

New:

client = Google::Cloud::Speech.speech

request_object = Google::Cloud::Speech::V1::RecognizeRequest.new(
  config: {
    language_code: "en-US",
    sample_rate_hertz: 44_100,
    encoding: :FLAC
  },
  audio: {
    uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
  }
)

# Pass a request object as a positional argument:
response = client.recognize request_object

Finally, in older releases, to provide call options, you would pass a Google::Gax::CallOptions object with the :options keyword argument. In the 1.0 release, pass call options using a second set of keyword arguments.

Old:

client = Google::Cloud::Speech.new

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

options = Google::Gax::CallOptions.new timeout: 10.0

response = client.recognize config, audio, options: options

New:

client = Google::Cloud::Speech.speech

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

# Use a hash to wrap the normal call arguments (or pass a request object), and
# then add further keyword arguments for the call options.
response = client.batch_annotate_images(
  { config: config, audio: audio },
  timeout: 10.0)

Streaming Interface

The client library includes one special streaming method streaming_recognize. In the older client, this method employed a streaming interface in which you would register callbacks for different incremental result events. In version 1.0, we have standardized the streaming interfaces across the various Ruby client libraries. The streaming_recognize call returns a lazy enumerable that you can query for incremental results.

Old:

client = Google::Cloud::Speech.new

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
stream = client.streaming_recognize config: config

stream.on_interim { |result| puts "received interim" }
stream.on_complete { |result| puts "received complete" }

stream.send File.read("my_input.flac", mode: "rb")
stream.stop

stream.wait_until_complete!
results = stream.results

New:

client = Google::Cloud::Speech.speech

input_stream = Gapic::StreamInput.new
output_stream = client.streaming_recognize input_stream

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
streaming_config = { config: config }

input_stream.push({ streaming_config: streaming_config })
input_stream.push({ audio_content: File.read("my_input.flac", mode: "rb") })
input_stream.close

output_stream.each do |response|
  puts "received: #{response}"
end

Handling Errors

The client reports standard gRPC error codes by raising exceptions. In older releases, these exceptions were located in the Google::Gax namespace and were subclasses of the Google::Gax::GaxError base exception class, defined in the google-gax gem. However, these classes were different from the standard exceptions (subclasses of Google::Cloud::Error) thrown by other client libraries such as google-cloud-storage.

The 1.0 client library now uses the Google::Cloud::Error exception hierarchy, for consistency across all the Google Cloud client libraries. In general, these exceptions have the same name as their counterparts from older releases, but are located in the Google::Cloud namespace rather than the Google::Gax namespace.

Old:

client = Google::Cloud::Speech.new

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

begin
  response = client.recognize config, audio
rescue Google::Gax::Error => e
  # Handle exceptions that subclass Google::Gax::Error
end

New:

client = Google::Cloud::Speech.speech

config = {
  language_code: "en-US",
  sample_rate_hertz: 44_100,
  encoding: :FLAC
}
audio = {
  uri: "gs://cloud-samples-data/speech/brooklyn_bridge.flac"
}

begin
  response = client.recognize config: config, audio: audio
rescue Google::Cloud::Error => e
  # Handle exceptions that subclass Google::Cloud::Error
end

Class Namespaces

In older releases, the client object was of classes with names like: Google::Cloud::Speech::V1::SpeechClient. In the 1.0 release, the client object is of a different class: Google::Cloud::Speech::V1::Speech::Client. Note that most users will use the factory methods such as Google::Cloud::Speech.speech to create instances of the client object, so you may not need to reference the actual class directly. See Creating Clients.

In older releases, the credentials object was of class Google::Cloud::Speech::V1::Credentials. In the 1.0 release, each service has its own credentials class, e.g. Google::Cloud::Speech::V1::Speech::Credentials. Again, most users will not need to reference this class directly. See Client Configuration.