Update app.py

app.py

@@ -146,6 +146,37 @@ def auto_reset_state():
     time.sleep(5)
     return None, ""
 
+
+def transcribe(
+    audio: tuple[int, np.ndarray],
+    transformers_chat: list[dict],
+    conversation: list[dict],
+):
+    original_sr = audio[0]
+    target_sr = 16000
+
+    audio_sr = librosa.resample(
+        audio[1].astype(np.float32) / 32768.0, orig_sr=original_sr, target_sr=target_sr
+    )
+
+    tf_input = [d for d in transformers_chat]
+
+    output = pipe(
+        {"audio": audio_sr, "turns": tf_input, "sampling_rate": target_sr},
+        max_new_tokens=512,
+    )
+    transcription = whisper({"array": audio_sr.squeeze(), "sampling_rate": target_sr})
+    yield AdditionalOutputs
+
+    # conversation.append({"role": "user", "content": transcription["text"]})
+    # conversation.append({"role": "assistant", "content": output})
+    # transformers_chat.append({"role": "user", "content": transcription["text"]})
+    # transformers_chat.append({"role": "assistant", "content": output})
+
+    # yield AdditionalOutputs(transformers_chat, conversation)
+
+
+
 def transcribe_function(stream, new_chunk):
     try:
         sr, y = new_chunk[0], new_chunk[1]
@@ -548,8 +579,8 @@ def response(audio: tuple[int, np.ndarray], conversation_llm_format: list[dict],
    
     #prompt = "Hey Annie how are you"
     #prompt = transcribe_function (client_state.grState, audio)
-    #prompt = client_state.hf.automatic_speech_recognition(audio_to_bytes(audio),model="https://kttcvz41e0htmjpp.us-east-1.aws.endpoints.huggingface.cloud").text
-    prompt = client_state.hf.automatic_speech_recognition(audio_to_bytes(audio),model="https://u3fcydn2o5vvwyd0.us-east-1.aws.endpoints.huggingface.cloud").text
+    prompt = client_state.hf.automatic_speech_recognition(audio_to_bytes(audio),model="https://kttcvz41e0htmjpp.us-east-1.aws.endpoints.huggingface.cloud").text
+    #prompt = client_state.hf.automatic_speech_recognition(audio_to_bytes(audio),model="https://u3fcydn2o5vvwyd0.us-east-1.aws.endpoints.huggingface.cloud").text
 
     # ADDED BY AL TO USE THE SPEECH TO TEXT GRADIO CLIENT - not working 
     # # client = Client("on1onmangoes/radarheyanniebackend112724")