Update speaker context handling to use safe references (partial fix…

… for incompatible plugins) (#139)

Update `speak` message handling to use `Message.data['lang']` instead of hard-coded default
Updated logging and dependencies

neon_audio/service.py

@@ -135,15 +135,16 @@ def handle_get_tts(self, message):
         ident = message.context.get("ident") or "neon.get_tts.response"
         LOG.info(f"Handling TTS request: {ident}")
         if not message.data.get("speaker"):
-            LOG.warning(f"No speaker data with request, "
-                        f"core defaults will be used.")
+            LOG.info(f"No speaker data with request, "
+                     f"core defaults will be used.")
         if text:
             if not isinstance(text, str):
                 self.bus.emit(message.reply(
                     ident, data={"error": f"text is not a str: {text}"}))
                 return
             try:
                 responses = self.tts.get_multiple_tts(message)
+                LOG.debug(f"Emitting response: {responses}")
                 self.bus.emit(message.reply(ident, data=responses))
             except Exception as e:
                 LOG.exception(e)

neon_audio/tts/neon.py

@@ -63,12 +63,13 @@ def get_requested_tts_languages(msg) -> list:
         # If speaker data is present, use it
         if msg.data.get("speaker"):
             speaker = msg.data.get("speaker")
-            tts_reqs.append({"speaker": speaker["name"],
-                             "language": speaker["language"],
-                             "gender": speaker["gender"],
+            tts_reqs.append({"speaker": speaker.get("name", "Neon"),
+                             "language": speaker.get("language",
+                                                     msg.data.get("lang")),
+                             "gender": speaker.get("gender", default_gender),
                              "voice": speaker.get("voice")
                              })
-            LOG.debug(f">>> speaker={speaker}")
+            LOG.info(f">>> speaker={speaker}")
 
         # If multiple profiles attached to message, get TTS for all
         elif profiles:
@@ -245,42 +246,44 @@ def get_multiple_tts(self, message, **kwargs) -> dict:
         LOG.debug(f"tts_requested={tts_requested}")
         sentence = message.data["text"]
         sentence = self.validate_ssml(sentence)
+        skill_lang = message.data.get('lang') or self.lang
+        LOG.debug(f"utterance_lang={skill_lang}")
         responses = {}
         for request in tts_requested:
-            lang = kwargs["lang"] = request["language"]
+            tts_lang = kwargs["lang"] = request["language"]
             # Check if requested tts lang matches internal (text) lang
             # TODO: `self.lang` should come from the incoming message
-            if lang.split("-")[0] != self.lang.split("-")[0]:
-                self.cached_translations.setdefault(lang, {})
+            if tts_lang.split("-")[0] != skill_lang.split("-")[0]:
+                self.cached_translations.setdefault(tts_lang, {})
 
-                tx_sentence = self.cached_translations[lang].get(sentence)
+                tx_sentence = self.cached_translations[tts_lang].get(sentence)
                 if not tx_sentence:
-                    tx_sentence = self.translator.translate(sentence, lang,
-                                                            self.lang)
-                    self.cached_translations[lang][sentence] = tx_sentence
+                    tx_sentence = self.translator.translate(sentence, tts_lang,
+                                                            skill_lang)
+                    self.cached_translations[tts_lang][sentence] = tx_sentence
                     self.cached_translations.store()
                 LOG.info(f"Got translated sentence: {tx_sentence}")
             else:
                 tx_sentence = sentence
             wav_file, phonemes = self._get_tts(tx_sentence, request, **kwargs)
 
             # If this is the first response, populate translation and phonemes
-            if not responses.get(lang):
-                responses[lang] = {"sentence": tx_sentence,
-                                   "translated": tx_sentence != sentence,
-                                   "phonemes": phonemes,
-                                   "genders": list()}
+            responses.setdefault(tts_lang, {"sentence": tx_sentence,
+                                            "translated": tx_sentence != sentence,
+                                            "phonemes": phonemes,
+                                            "genders": list()})
 
             # Append the generated audio from this request
             if os.path.isfile(wav_file):
-                responses[lang][request["gender"]] = wav_file
-                responses[lang]["genders"].append(request["gender"])
+                responses[tts_lang][request["gender"]] = wav_file
+                responses[tts_lang]["genders"].append(request["gender"])
                 # If this is a remote request, encode audio in the response
                 if message.context.get("klat_data") or \
                         message.msg_type == "neon.get_tts":
-                    responses[lang].setdefault("audio", {})
-                    responses[lang]["audio"][request["gender"]] = \
+                    responses[tts_lang].setdefault("audio", {})
+                    responses[tts_lang]["audio"][request["gender"]] = \
                         encode_file_to_base64_string(wav_file)
+                    LOG.debug(f"Got {tts_lang} {request['gender']} response")
             else:
                 raise RuntimeError(f"No audio generated for request: {request}")
         return responses

requirements/requirements.txt

@@ -1,7 +1,7 @@
 ovos-audio~=0.0.1,>=0.0.2a12,<0.0.2a14
 # TODO: ovos-audio 0.0.2a14 introduces a breaking change around `ident` handling
-ovos-utils~=0.0.34
-ovos-config~=0.0.7
+ovos-utils==0.0.35
+ovos-config~=0.0.10
 phoneme-guesser~=0.1
 ovos-plugin-manager~=0.0.22,>=0.0.24a5
 neon-utils[network]~=1.6