code

app.py

@@ -7,11 +7,14 @@ import os
 import torch
 import librosa
 
-
-pipe = transformers.pipeline(model='fixie-ai/ultravox-v0_4_1-llama-3_1-8b', trust_remote_code=True,
-                             device=torch.device('cuda'))
-whisper = transformers.pipeline(model="openai/whisper-large-v3-turbo",
-                                device=torch.device('cuda'))
+pipe = transformers.pipeline(
+    model="fixie-ai/ultravox-v0_4_1-llama-3_1-8b",
+    trust_remote_code=True,
+    device=torch.device("cuda"),
+)
+whisper = transformers.pipeline(
+    model="openai/whisper-large-v3-turbo", device=torch.device("cuda")
+)
 
 account_sid = os.environ.get("TWILIO_ACCOUNT_SID")
 auth_token = os.environ.get("TWILIO_AUTH_TOKEN")
@@ -29,29 +32,29 @@ else:
     rtc_configuration = None
 
 
-
-def transcribe(audio: tuple[int, np.ndarray], conversation: list[dict], gradio_convo: list[dict]):
+def transcribe(audio: tuple[int, np.ndarray], 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)
-   
-    output = pipe({"audio": audio_sr, "turns": conversation, "sampling_rate": target_sr},
-                  max_new_tokens=512)
+    audio_sr = librosa.resample(
+        audio[1].astype(np.float32) / 32768.0, orig_sr=original_sr, target_sr=target_sr
+    )
+
+    output = pipe(
+        {"audio": audio_sr, "turns": conversation, "sampling_rate": target_sr},
+        max_new_tokens=512,
+    )
     transcription = whisper({"array": audio_sr.squeeze(), "sampling_rate": target_sr})
 
-    conversation.append({"role": "user", "content": transcription})
+    conversation.append({"role": "user", "content": transcription["text"]})
     conversation.append({"role": "assistant", "content": output})
-    gradio_convo.append({"role": "user", "content": transcription})
-    gradio_convo.append({"role": "assistant", "content": output})
 
-    yield AdditionalOutputs(conversation, gradio_convo)
+    yield AdditionalOutputs(conversation)
 
 
 with gr.Blocks() as demo:
     gr.HTML(
-    """
+        """
     <h1 style='text-align: center'>
     Talk to Ultravox Llama 3.1 8b (Powered by WebRTC ⚡️)
     </h1>
@@ -64,24 +67,34 @@ with gr.Blocks() as demo:
     </p>
     """
     )
-    transformers_convo = gr.State(value=[{
-        "role": "system",
-        "content": "You are a friendly and helpful character. You love to answer questions for people."
-        }])
     with gr.Row():
-        with gr.Column():
+        with gr.Group():
+            transcript = gr.Chatbot(
+                label="transcript",
+                type="messages",
+                value=[
+                    {
+                        "role": "system",
+                        "content": "You are a friendly and helpful character. You love to answer questions for people.",
+                    }
+                ],
+            )
             audio = WebRTC(
                 rtc_configuration=rtc_configuration,
                 label="Stream",
                 mode="send",
                 modality="audio",
             )
-        with gr.Column():
-            transcript = gr.Chatbot(label="transcript", type="messages")
 
-    audio.stream(ReplyOnPause(transcribe), inputs=[audio, transformers_convo, transcript], outputs=[audio], time_limit=90)
-    audio.on_additional_outputs(lambda s,a: (s,a), outputs=[transformers_convo, transcript],
-                                queue=False, show_progress="hidden")
+    audio.stream(
+        ReplyOnPause(transcribe),
+        inputs=[audio, transcript],
+        outputs=[audio],
+        time_limit=90,
+    )
+    audio.on_additional_outputs(
+        lambda s: s, outputs=[transcript], queue=False, show_progress="hidden"
+    )
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch()