WIP

.rubocop.yml

@@ -2,6 +2,9 @@ require:
   - rubocop-performance
   - rubocop-rspec
 
+AllCops:
+  NewCops: enable
+
 Style/FrozenStringLiteralComment:
   Enabled: false
 

components/app/Gemfile

@@ -2,6 +2,7 @@ source "https://rubygems.org"
 
 gem "adhearsion", github: "somleng/adhearsion"
 gem "aws-sdk-lambda"
+gem "aws-sdk-polly"
 gem "faraday"
 gem "http"
 gem "okcomputer"
@@ -11,6 +12,7 @@ gem "sentry-ruby"
 gem "sinatra"
 gem "sinatra-contrib", require: false
 gem "skylight"
+gem "tts_voices", github: "somleng/tts_voices"
 
 group :development, :test do
   gem "rubocop"

components/app/Gemfile.lock

@@ -36,6 +36,13 @@ GIT
       thor
       virtus (~> 1.0)
 
+GIT
+  remote: https://github.com/somleng/tts_voices.git
+  revision: e719a88769a77d08a85dc3aa99ef04ba4630ad18
+  specs:
+    tts_voices (0.1.0)
+      aws-sdk-polly
+
 GEM
   remote: https://rubygems.org/
   specs:
@@ -58,6 +65,9 @@ GEM
     aws-sdk-lambda (1.106.0)
       aws-sdk-core (~> 3, >= 3.184.0)
       aws-sigv4 (~> 1.1)
+    aws-sdk-polly (1.76.0)
+      aws-sdk-core (~> 3, >= 3.184.0)
+      aws-sigv4 (~> 1.1)
     aws-sigv4 (1.6.0)
       aws-eventstream (~> 1, >= 1.0.2)
     axiom-types (0.1.1)
@@ -274,6 +284,7 @@ PLATFORMS
 DEPENDENCIES
   adhearsion!
   aws-sdk-lambda
+  aws-sdk-polly
   faraday
   http
   okcomputer
@@ -290,6 +301,7 @@ DEPENDENCIES
   sinatra
   sinatra-contrib
   skylight
+  tts_voices!
   twilio-ruby
   vcr
   webmock

components/app/app/call_controllers/call_controller.rb

@@ -64,6 +64,7 @@ def build_call_properties
       api_version: response.api_version,
       to: response.to,
       from: response.from,
+      default_tts_voice_identifier: response.default_tts_voice_identifier,
       sip_headers: SIPHeaders.new(call_sid: response.call_sid, account_sid: response.account_sid)
     )
   end

components/app/app/models/call_properties.rb

@@ -10,7 +10,7 @@
   :to,
   :from,
   :sip_headers,
-  :default_tts_provider,
+  :default_tts_voice_identifier,
   keyword_init: true
 ) do
   def inbound?

components/app/app/models/execute_twiml.rb

@@ -10,10 +10,14 @@ class ExecuteTwiML
   NESTED_GATHER_VERBS = %w[Say Play].freeze
   MAX_LOOP = 100
   SLEEP_BETWEEN_REDIRECTS = 1
-  DEFAULT_VOICE_PROVIDER = "polly".freeze
   DEFAULT_TWILIO_VOICE = "man".freeze
   DEFAULT_TWILIO_LANGUAGE = "en".freeze
   FINISH_ON_KEY_PATTERN = /\A(?:\d|\*|\#)\z/.freeze
+  BASIC_TTS_MAPPING = {
+    "man" => "Basic.Kal",
+    "woman" => "Basic.Slt"
+  }.freeze
+
   DIAL_CALL_STATUSES = {
     no_answer: "no-answer",
     answer: "completed",
@@ -106,10 +110,10 @@ def execute_say(verb)
     answer unless answered?
 
     attributes = twiml_attributes(verb)
-    provider = resolve_voice_provider(attributes["voice"])
+    tts_voice = resolve_tts_voice(attributes)
 
     twiml_loop(attributes).each do
-      say(say_options(verb.content, attributes))
+      say(say_options(verb.content, tts_voice))
     end
 
     # call_platform_client.notify_tts_event(
@@ -141,7 +145,12 @@ def execute_gather(verb)
       end
 
       nested_verb_attributes = twiml_attributes(nested_verb)
-      content = nested_verb.name == "Say" ? say_options(nested_verb.content, nested_verb_attributes) : nested_verb.content
+      content = if nested_verb.name == "Say"
+        tts_voice = resolve_tts_voice(nested_verb_attributes)
+        say_options(nested_verb.content, tts_voice)
+      else
+        nested_verb.content
+      end
       result.concat(Array.new(twiml_loop(nested_verb_attributes).count, content))
     end
 
@@ -255,14 +264,9 @@ def execute_record(verb)
     )
   end
 
-  def say_options(content, attributes)
-    voice_params = {
-      name: attributes.fetch("voice", DEFAULT_TWILIO_VOICE),
-      language: attributes.fetch("language", DEFAULT_TWILIO_LANGUAGE)
-    }
-
+  def say_options(content, tts_voice)
     ssml = RubySpeech::SSML.draw do
-      voice(voice_params) do
+      voice(name: tts_voice.identifier, language: tts_voice.language) do
         # mod ssml doesn't support non-ascii characters
         # https://github.com/signalwire/freeswitch/issues/1348
         string(content + ".")
@@ -311,10 +315,24 @@ def normalize_recording_url(raw_recording_url)
     URL_PATTERN.match(raw_recording_url)[0]
   end
 
-  def resolve_voice_provider(voice)
-    return DEFAULT_VOICE_PROVIDER if voice.in?(["man", "woman"])
-    return DEFAULT_VOICE_PROVIDER if voice.start_with?("Polly.")
+  def resolve_tts_voice(attributes)
+    voice_attribute = attributes["voice"]
+    language_attribute = attributes["language"]
+
+    default_tts_voice = TTSVoices::Voice.find(call_properties.default_tts_voice_identifier)
+    voice_attribute = BASIC_TTS_MAPPING.fetch(voice_attribute) if BASIC_TTS_MAPPING.key?(voice_attribute)
 
-    raise Errors::TwiMLError, "Unsupported voice"
+    if voice_attribute.blank? && language_attribute.present?
+      tts_voice = resolve_tts_voice_by_language(default_tts_voice.provider, language_attribute)
+      voice_attribute = tts_voice&.identifier
+    end
+
+    TTSVoices::Voice.find(voice_attribute) || default_tts_voice
+  end
+
+  def resolve_tts_voice_by_language(provider, language_attribute)
+    TTSVoices::Voice.all.find do |voice|
+      voice.provider == provider && voice.language.casecmp(language_attribute).zero?
+    end
   end
 end

components/app/app/models/outbound_call.rb

@@ -29,7 +29,7 @@ def initiate
           api_version: call_params.fetch("api_version"),
           from: call_params.fetch("from"),
           to: call_params.fetch("to"),
-          default_tts_provider: call_params.fetch("default_tts_provider"),
+          default_tts_voice_identifier: call_params.fetch("default_tts_voice_identifier"),
           sip_headers:
         )
       },

components/app/config/initializers/aws_stubs.rb

@@ -28,4 +28,15 @@ def to_h
       }
     }
   }
+
+  Aws.config[:polly] ||= {
+    stub_responses: {
+      describe_voices: {
+        voices: [
+          { gender: "Female", id: "Lotte", language_code: "nl-NL" },
+          { gender: "Female", id: "Joanna", language_code: "en-US" }
+        ]
+      }
+    }
+  }
 end

components/app/lib/call_platform/client.rb

@@ -15,6 +15,7 @@ class InvalidRequestError < StandardError; end
       :api_version,
       :to,
       :from,
+      :default_tts_voice_identifier,
       keyword_init: true
     )
 
@@ -56,7 +57,8 @@ def create_call(params)
         direction: json_response.fetch("direction"),
         to: json_response.fetch("to"),
         from: json_response.fetch("from"),
-        api_version: json_response.fetch("api_version")
+        api_version: json_response.fetch("api_version"),
+        default_tts_voice_identifier: json_response.fetch("default_tts_voice_identifier")
       )
     end
 

components/app/spec/call_controllers/gather_spec.rb

@@ -442,7 +442,9 @@
       it "handles nested <Say>" do
         controller = build_controller(
           stub_voice_commands: { ask: build_input_result(nil) },
-          call_properties: { voice_url: "https://www.example.com/gather.xml" }
+          call_properties: {
+            voice_url: "https://www.example.com/gather.xml"
+          }
         )
 
         stub_twiml_request(controller, response: <<~TWIML)
@@ -464,15 +466,15 @@
           outputs.first(3).each do |ssml|
             node = ssml.voice.children.first
             expect(node.content).to eq("Hello World.")
-            expect(node.attributes.fetch("name").value).to eq("woman")
-            expect(node.attributes.fetch("lang").value).to eq("de")
+            expect(node.attributes.fetch("name").value).to eq("Basic.Slt")
+            expect(node.attributes.fetch("lang").value).to eq("en-US")
           end
 
           outputs.last(5).each do |ssml|
             node = ssml.voice.children.first
             expect(node.content).to eq("Foobar.")
-            expect(node.attributes.fetch("name").value).to eq("man")
-            expect(node.attributes.fetch("lang").value).to eq("en")
+            expect(node.attributes.fetch("name").value).to eq("Basic.Kal")
+            expect(node.attributes.fetch("lang").value).to eq("en-US")
           end
         end
       end

components/app/spec/call_controllers/say_spec.rb

@@ -19,7 +19,12 @@
       # |             | Limited to 4KB (4,000 ASCII characters)  |
 
       it "outputs SSML" do
-        controller = build_controller(stub_voice_commands: :say)
+        controller = build_controller(
+          stub_voice_commands: :say,
+          call_properties: {
+            default_tts_voice_identifier: "Basic.Slt"
+          }
+        )
         stub_twiml_request(controller, response: <<~TWIML)
             <?xml version="1.0" encoding="UTF-8" ?>
             <!-- Handles whitespace -->
@@ -35,6 +40,8 @@
           expect(ssml).to be_a(RubySpeech::SSML::Speak)
           expect(ssml.text).to eq("Hola, buen día.")
           expect(ssml.to_xml).to include("Hola, buen día.")
+          expect(fetch_ssml_attribute(ssml, :name)).to eq("Basic.Slt")
+          expect(fetch_ssml_attribute(ssml, :lang)).to eq("en-US")
         end
       end
     end
@@ -58,35 +65,36 @@
         # | Attribute Name | Allowed Values | Default Value |
         # | voice          | man, woman     | man           |
 
-        it "defaults to man" do
+        it "sets a custom voice" do
           controller = build_controller(stub_voice_commands: :say)
           stub_twiml_request(controller, response: <<~TWIML)
             <?xml version="1.0" encoding="UTF-8" ?>
             <Response>
-              <Say>Hello World</Say>
+              <Say voice="woman">Hello World</Say>
             </Response>
           TWIML
 
           controller.run
 
           expect(controller).to have_received(:say) do |ssml|
-            expect(fetch_ssml_attribute(ssml, :name)).to eq("man")
+            expect(fetch_ssml_attribute(ssml, :name)).to eq("Basic.Slt")
           end
         end
 
-        it "sets a custom voice" do
+        it "supports Polly" do
           controller = build_controller(stub_voice_commands: :say)
           stub_twiml_request(controller, response: <<~TWIML)
             <?xml version="1.0" encoding="UTF-8" ?>
             <Response>
-              <Say voice="woman">Hello World</Say>
+              <Say voice="Polly.Lotte">Hello World</Say>
             </Response>
           TWIML
 
           controller.run
 
           expect(controller).to have_received(:say) do |ssml|
-            expect(fetch_ssml_attribute(ssml, :name)).to eq("woman")
+            expect(fetch_ssml_attribute(ssml, :name)).to eq("Polly.Lotte")
+            expect(fetch_ssml_attribute(ssml, :lang)).to eq("nl-NL")
           end
         end
       end
@@ -106,7 +114,12 @@
         # so the option is ignored
 
         it "sets the language to en by default" do
-          controller = build_controller(stub_voice_commands: :say)
+          controller = build_controller(
+            stub_voice_commands: :say,
+            call_properties: {
+              default_tts_voice_identifier: "Basic.Kal"
+            }
+          )
           stub_twiml_request(controller, response: <<~TWIML)
             <?xml version="1.0" encoding="UTF-8" ?>
             <Response>
@@ -117,23 +130,29 @@
           controller.run
 
           expect(controller).to have_received(:say) do |ssml|
-            expect(fetch_ssml_attribute(ssml, :lang)).to eq("en")
+            expect(fetch_ssml_attribute(ssml, :lang)).to eq("en-US")
           end
         end
 
         it "sets the language when specifying the language attribute" do
-          controller = build_controller(stub_voice_commands: :say)
+          controller = build_controller(
+            stub_voice_commands: :say,
+            call_properties: {
+              default_tts_voice_identifier: "Polly.Joanna"
+            }
+          )
           stub_twiml_request(controller, response: <<~TWIML)
             <?xml version="1.0" encoding="UTF-8" ?>
             <Response>
-              <Say language="pt-BR">Hello World</Say>
+              <Say language="nl-NL">Hello World</Say>
             </Response>
           TWIML
 
           controller.run
 
           expect(controller).to have_received(:say) do |ssml|
-            expect(fetch_ssml_attribute(ssml, :lang)).to eq("pt-BR")
+            expect(fetch_ssml_attribute(ssml, :name)).to eq("Polly.Lotte")
+            expect(fetch_ssml_attribute(ssml, :lang)).to eq("nl-NL")
           end
         end
       end

components/app/spec/models/outbound_call_spec.rb

@@ -15,7 +15,7 @@
       "account_auth_token" => "sample-auth-token",
       "direction" => "outbound-api",
       "api_version" => "2010-04-01",
-      "default_tts_provider" => "basic",
+      "default_tts_voice_identifier" => "Basic.Kal",
       "routing_parameters" => {
         "destination" => "85516701721",
         "dial_string_prefix" => nil,
@@ -47,7 +47,7 @@
           call_sid: "sample-call-sid",
           direction: "outbound-api",
           api_version: "2010-04-01",
-          default_tts_provider: "basic",
+          default_tts_voice_identifier: "Basic.Kal",
           to: "+85516701721",
           from: "2442",
           sip_headers: SIPHeaders.new(
@@ -141,7 +141,7 @@ def build_call_params(params)
       "account_auth_token" => "sample-auth-token",
       "direction" => "outbound-api",
       "api_version" => "2010-04-01",
-      "default_tts_provider" => "basic",
+      "default_tts_voice_identifier" => "Basic.Kal",
       "routing_parameters" => {
         "destination" => "85516701721",
         "dial_string_prefix" => nil,