feat: add local speech provider support

README.md

@@ -39,16 +39,17 @@ On the new component panel, copy and paste the following attribute template into
 
 ```json
 {
-  "speech_provider": "google|elevenlabs",
+  "speech_provider": "google|elevenlabs|local",
   "speech_provider_key": "<SECRET-KEY>",
+  "speech_service_name": "<service name>",
   "speech_voice": "<VOICE-OPTION>",
   "completion_provider": "openai",
   "completion_model": "gpt-4|gpt-3.5-turbo",
   "completion_provider_org": "<org-abc123>",
   "completion_provider_key": "<sk-mykey>",
   "completion_persona": "<PERSONA>",
   "listen": true,
-  "listen_provider": "google",
+  "listen_provider": "google|<service name>",
   "listen_trigger_say": "<TRIGGER-PHRASE>",
   "listen_trigger_completion": "<COMPLETION-PHRASE>",
   "listen_trigger_command": "<COMMAND-TO-RETRIEVE-STORED-TEXT>",
@@ -69,8 +70,9 @@ The following attributes are available for the `viam-labs:speech:speechio` speec
 
 | Name    | Type   | Inclusion    | Description |
 | ------- | ------ | ------------ | ----------- |
-| `speech_provider` | string | Optional | The speech provider for the voice service: `"google"` or `"elevenlabs"`. Default: `"google"`.  |
+| `speech_provider` | string | Optional | The speech provider for the voice service: `"google"`, `"elevenlabs"`, or `"local"`. Default: `"google"`.  |
 | `speech_provider_key` | string | **Required** | The secret key for the provider - only required for elevenlabs. Default: `""`. |
+| `speech_provider_name` | string | **Required** | The name of the service to provide text-to-speech (TTS) capabilities - only required when `"local"` is set for `speech_provider` field. Default: `""`. |
 | `speech_voice`  | string | Optional | If the speech_provider (example: elevenlabs) provides voice options, you can select the voice here. Default: `"Josh"`. |
 | `completion_provider`  | string | Optional | `"openai"`. Other providers may be supported in the future. [completion_provider_org](#completion_provider_org) and [completion_provider_key](#completion_provider_key) must also be provided. Default: `"openai"`. |
 | `completion_model`  | string | Optional | `gpt-4` or `gpt-3.5-turbo`. Other models may be supported in the future.  [completion_provider_org](#completion_provider_org) and [completion_provider_key](#completion_provider_key) must also be provided. Default: `"gpt-4"`. |
@@ -108,6 +110,26 @@ The following attributes are available for the `viam-labs:speech:speechio` speec
 ```
 In the above case, the `listen_provider` and `depends_on` value are set to the name of the configured `viam-labs:speech:stt-vosk` service for the robot config.
 
+\*If the `speech_provider` is another speech service (set to `"local"`), it should be set as a dependency for the "speechio" service. This must be done using the "Raw JSON" editor within the robot configuration by setting the `"depends_on"` field for the service:
+
+```json
+{
+  "name": "speechio",
+  "type": "speech",
+  "namespace": "viam-labs",
+  "model": "viam-labs:speech:speechio",
+  "attributes": {
+    "speech_provider": "local",
+    "speech_service_name": "piper",
+    /* other configuration for the service */
+  },
+  "depends_on": [
+    "piper"
+  ]
+}
+```
+In the above case, the `speech_service_name` and `depends_on` value are set to the name of the configured [`viam-labs:speech:tts-piper` service](https://github.com/viam-labs/tts-piper) for the robot config.
+
 ### Example configuration
 
 The following configuration sets up listening mode with local speech-to-text, uses an ElevenLabs voice "Antoni", makes AI completions available, and uses a 'Gollum' persona for AI completions:

pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "speech"
-version = "0.5.2"
+version = "0.6.0"
 authors = [
   { name="Matt Vella", email="[email protected]" },
 ]

src/__main__.py

@@ -2,7 +2,9 @@
 
 from viam.module.module import Module
 
-from .speech import SpeechIOService, SpeechService
+from speech_service_api import SpeechService
+from .speech import SpeechIOService
+
 
 async def main():
     """This function creates and starts a new module, after adding all desired resources.
@@ -15,4 +17,4 @@ async def main():
 
 
 if __name__ == "__main__":
-    asyncio.run(main())
+    asyncio.run(main())

src/speech/speechio.py

@@ -30,6 +30,7 @@
 class SpeechProvider(str, Enum):
     google = "google"
     elevenlabs = "elevenlabs"
+    local = "local"
 
 
 class CompletionProvider(str, Enum):
@@ -51,6 +52,7 @@ class RecState:
 
 rec_state = RecState()
 
+
 class SpeechIOService(SpeechService, Reconfigurable):
     """This is the specific implementation of a ``SpeechService`` (defined in api.py)
 
@@ -95,6 +97,9 @@ async def say(self, text: str, blocking: bool, cache_only: bool = False) -> str:
         if str == "":
             raise ValueError("No text provided")
 
+        if self.tts is not None:
+            return await self.tts.say(text=text, blocking=blocking)
+
         LOGGER.info("Generating audio...")
         if not os.path.isdir(CACHEDIR):
             os.mkdir(CACHEDIR)
@@ -109,7 +114,7 @@ async def say(self, text: str, blocking: bool, cache_only: bool = False) -> str:
         )
         try:
             if not os.path.isfile(file):  # read from cache if it exists
-                if self.speech_provider == "elevenlabs":
+                if self.speech_provider == SpeechProvider.elevenlabs:
                     audio = eleven.generate(text=text, voice=self.speech_voice)
                     eleven.save(audio=audio, filename=file)
                 else:
@@ -153,6 +158,8 @@ async def listen_trigger(self, type: str) -> str:
         return "OK"
 
     async def is_speaking(self) -> bool:
+        if self.tts is not None:
+            return await self.tts.is_speaking()
         return mixer.music.get_busy()
 
     async def completion(
@@ -245,9 +252,13 @@ async def to_text(self, speech: bytes, format: str = "mp3"):
         return ""
 
     async def to_speech(self, text):
-        if self.speech_provider == "elevenlabs":
+        if self.speech_provider == SpeechProvider.elevenlabs:
             audio = eleven.generate(text=text, voice=self.speech_voice)
             return audio
+
+        if self.speech_provider == SpeechProvider.local and self.tts is not None:
+            audio = await self.tts.to_speech(text=text)
+            return audio
         else:
             mp3_fp = BytesIO()
             sp = gTTS(text=text, lang="en", slow=False)
@@ -322,6 +333,7 @@ def reconfigure(
             str(attrs.get("speech_provider", "google"))
         ]
         self.speech_provider_key = str(attrs.get("speech_provider_key", ""))
+        self.speech_service_name = str(attrs.get("speech_service_name", ""))
         self.speech_voice = str(attrs.get("speech_voice", "Josh"))
         self.completion_provider = CompletionProvider[
             str(attrs.get("completion_provider", "openaigpt35turbo"))
@@ -350,13 +362,22 @@ def reconfigure(
         self.command_list = []
         self.trigger_active = False
         self.active_trigger_type = ""
-        self.stt = None
+        self.stt: None | SpeechService = None
+        self.tts: None | SpeechService = None
 
         if (
             self.speech_provider == SpeechProvider.elevenlabs
             and self.speech_provider_key != ""
         ):
             eleven.set_api_key(self.speech_provider_key)
+        elif (
+            self.speech_provider == SpeechProvider.local
+            and self.speech_service_name != ""
+        ):
+            tts = dependencies[
+                SpeechService.get_resource_name(self.speech_service_name)
+            ]
+            self.tts = cast(SpeechService, tts)
         else:
             self.speech_provider = SpeechProvider.google