Update app.py

app.py

@@ -23,19 +23,19 @@ 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")
+    if "Llama 3B Service" in model:
+        return OpenAI(
+            base_url="http://host:60002/v1",
+            api_key="token-abc123"
+        )
     elif "Llama" in model:
         return InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
     elif "Mistral" in model:
         return InferenceClient("mistralai/Mistral-7B-Instruct-v0.2")
     elif "Phi" in model:
         return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
-    elif "Llama 3B" in model:
-        return OpenAI(
-            base_url="http://52.76.81.56:60002/v1",
-            api_key="token-abc123"
-        )
+    elif "Mixtral" in model:
+        return InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")
     else: 
         return InferenceClient("microsoft/Phi-3-mini-4k-instruct")
 
@@ -44,7 +44,7 @@ def randomize_seed_fn(seed: int) -> int:
     return seed
 
 system_instructions1 = """
-[SYSTEM] Answer as Real Optimus OPTIMUS, Made by 'Jaward.' 
+[SYSTEM] Answer as Real Jarvis JARVIS, Created by Jaward.
 Keep conversation friendly, short, clear, and concise. 
 Avoid unnecessary introductions and answer the user's questions directly. 
 Respond in a normal, conversational manner while being friendly and helpful.
@@ -57,7 +57,7 @@ def models(text, model="Mixtral 8x7B", seed=42):
     
     client = client_fn(model)
     
-    if "Llama 3B" in model:
+    if "Llama 3B Service" in model:
         messages = [
             {"role": "system", "content": system_instructions1},
             {"role": "user", "content": text}
@@ -72,7 +72,7 @@ def models(text, model="Mixtral 8x7B", seed=42):
             max_new_tokens=300,
             seed=seed
         )    
-        formatted_prompt = system_instructions1 + text + "[OPTIMUS]"
+        formatted_prompt = system_instructions1 + text + "[JARVIS]"
         stream = client.text_generation(
             formatted_prompt, **generate_kwargs, stream=True, details=True, return_full_text=False)
         output = ""
@@ -84,26 +84,24 @@ def models(text, model="Mixtral 8x7B", seed=42):
 async def respond(audio, model, seed):
     user = transcribe(audio)
     reply = models(user, model, seed)
-    communicate = edge_tts.Communicate(reply)
+    # Change the voice to a deep male voice
+    communicate = edge_tts.Communicate(reply, voice="en-US-GuyNeural")
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmp_file:
         tmp_path = tmp_file.name
         await communicate.save(tmp_path)
     yield tmp_path
 
-DESCRIPTION = """ # <center><b>OPTIMUS⚡</b></center>
-        ### <center>A personal Assistant of Jaward for YOU
-        ### <center>Voice Chat with your personal Assistant</center>
-        """
+DESCRIPTION = """ # <center><b>Hello, I am JARVIS your personal AI voice assistant</b></center>"""
 
 with gr.Blocks(css="style.css") as demo:    
     gr.Markdown(DESCRIPTION)
     with gr.Row():
         select = gr.Dropdown([
+            'Llama 3B Service',
             'Mixtral 8x7B',
             'Llama 3 8B',
             'Mistral 7B v0.3',
             'Phi 3 mini',
-            'Llama 3B'
         ],
         value="Mistral 7B v0.3",
         label="Model"