Update app.py

app.py

@@ -21,18 +21,6 @@ def transcribe(audio):
 
 HF_TOKEN = os.environ.get("HF_TOKEN", None)
 
-def client_fn(model):
-    if "Mixtral" in model:
-        return InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
-    elif "Llama" in model:
-        return InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
-    elif "Mistral" in model:
-        return InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
-    elif "Phi" in model:
-        return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
-    else:
-        return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
-
 def randomize_seed_fn(seed: int) -> int:
     seed = random.randint(0, 999999)
     return seed
@@ -45,18 +33,17 @@ Respond in a normal, conversational manner while being friendly and helpful.
 [USER]
 """
 
-def models(text, model="Mixtral 8x7B", seed=42):
-
+def models(text, seed=42):
     seed = int(randomize_seed_fn(seed))
-    generator = torch.Generator().manual_seed(seed)
-
-    client = client_fn(model)
-
+    generator = torch.Generator().manual_seed(seed)  
+    
+    client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
+    
     generate_kwargs = dict(
         max_new_tokens=300,
         seed=seed
     )
-
+    
     formatted_prompt = system_instructions1 + text + "[JARVIS]"
     stream = client.text_generation(
         formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
@@ -76,32 +63,23 @@ async def respond(audio, model, seed):
         await communicate.save(tmp_path)
     return tmp_path
 
-st.title("JARVIS⚡")
-st.markdown("### A personal Assistant of Tony Stark for YOU")
-st.markdown("### Voice Chat with your personal Assistant")
+DESCRIPTION = """ # <center><b>JARVIS⚡</b></center>
+        ### <center>A personal Assistant of Tony Stark for YOU
+        ### <center>Voice Chat with your personal Assistant</center>
+        """
 
-with st.form("voice_form"):
-    model_choice = st.selectbox("Choose a model", ['Mixtral 8x7B', 'Llama 3 8B', 'Mistral 7B v0.3', 'Phi 3 mini'], key="voice_model")
-    audio_file = st.file_uploader("Upload Audio", type=["wav", "mp3"], key="voice_audio")
-    submit_button = st.form_submit_button("Submit")
+st.markdown(DESCRIPTION)
+st.title("JARVIS")
 
-    if submit_button:
-        if audio_file is not None:
-            with st.spinner("Transcribing and generating response..."):
-                audio_bytes = audio_file.read()
-                with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
-                    tmp_file.write(audio_bytes)
-                    tmp_path = tmp_file.name
-                    response = respond(tmp_path, model_choice, 42)
-                    st.audio(response, format='audio/wav')
+uploaded_file = st.file_uploader("Upload audio file", type=["wav"])
+seed = st.slider("Seed", min_value=0, max_value=999999, value=0)
+if uploaded_file is not None:
+    # Convert the uploaded file to a BytesIO object
+    audio_bytes = uploaded_file.read()
 
-with st.form("text_form"):
-    model_choice = st.selectbox("Choose a model", ['Mixtral 8x7B', 'Llama 3 8B', 'Mistral 7B v0.3', 'Phi 3 mini'], key="text_model")
-    user_text = st.text_area("Enter your message:", key="text_input")
-    submit_button = st.form_submit_button("Submit")
+    # Process the audio using the respond function
+    response_path = asyncio.run(respond(audio_bytes, models, seed))
 
-    if submit_button:
-        if user_text:
-            with st.spinner("Generating response..."):
-                response = models(user_text, model_choice, 42)
-                st.text_area("JARVIS Response", value=response, key="text_output", height=150)
+    # Display the audio response
+    st.audio(response_path, format="audio/wav")
+    os.remove(response_path)