diff --git a/src/control_backend/agents/llm/llm_instructions.py b/src/control_backend/agents/llm/llm_instructions.py index e3aed7e..6922fca 100644 --- a/src/control_backend/agents/llm/llm_instructions.py +++ b/src/control_backend/agents/llm/llm_instructions.py @@ -30,7 +30,7 @@ class LLMInstructions: "You are a Pepper robot engaging in natural human conversation.", "Keep responses between 1–3 sentences, unless told otherwise.\n", "You're given goals to reach. Reach them in order, but make the conversation feel " - "natural. Some turns you should not try to achieve your goals.\n" + "natural. Some turns you should not try to achieve your goals.\n", ] if self.norms: diff --git a/src/control_backend/agents/transcription/speech_recognizer.py b/src/control_backend/agents/transcription/speech_recognizer.py index 9e61fd7..527d371 100644 --- a/src/control_backend/agents/transcription/speech_recognizer.py +++ b/src/control_backend/agents/transcription/speech_recognizer.py @@ -87,7 +87,6 @@ class MLXWhisperSpeechRecognizer(SpeechRecognizer): return mlx_whisper.transcribe( audio, path_or_hf_repo=self.model_name, - initial_prompt="You're a robot called Pepper, talking with a person called Twirre.", **self._get_decode_options(audio), )["text"].strip() @@ -105,8 +104,4 @@ class OpenAIWhisperSpeechRecognizer(SpeechRecognizer): def recognize_speech(self, audio: np.ndarray) -> str: self.load_model() - return whisper.transcribe( - self.model, - audio, - **self._get_decode_options(audio) - )["text"] + return whisper.transcribe(self.model, audio, **self._get_decode_options(audio))["text"]