Update app.py

app.py

@@ -1,27 +1,34 @@
+import os
+import time
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-client = InferenceClient("google/gemma-1.1-2b-it")
-client = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407")
+# Retrieve API token from Hugging Face Secrets
+HF_TOKEN = os.getenv("HF_TOKEN")
 
-def models(Query): 
-    
-    messages = []
-    
-    messages.append({"role": "user", "content": f"[SYSTEM] You are ASSISTANT who answer question asked by user in short and concise manner. [USER] {Query}"})
-
-    Response = ""
+# Ensure the token is available
+if not HF_TOKEN:
+    raise ValueError("Missing Hugging Face API Token! Set 'HF_TOKEN' in Hugging Face Secrets.")
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=2048,
-        stream=True
-    ):
-        token = message.choices[0].delta.content
+# Initialize clients with authentication
+client_gemma = InferenceClient("google/gemma-1.1-2b-it", token=HF_TOKEN)
+client_mistral = InferenceClient("mistralai/Mistral-Nemo-Instruct-2407", token=HF_TOKEN)
 
-        Response += token
-        yield Response
+# Function for normal fast responses
+def models(query): 
+    messages = [{"role": "user", "content": f"[SYSTEM] You are an AI assistant. Answer concisely. [USER] {query}"}]
+    
+    response = ""
+    try:
+        for message in client_gemma.chat_completion(messages, max_tokens=1024, stream=True):
+            token = message.choices[0].delta.content
+            response += token
+            yield response
+            time.sleep(0.3)  # Prevent rate limiting
+    except Exception as e:
+        yield f"Error: {str(e)}"
 
+# Function for detailed critical thinking responses
 def nemo(query):
     budget = 3
     message = f"""[INST] [SYSTEM] You are a helpful assistant in normal conversation.
@@ -54,22 +61,30 @@ Example format:
 <reflection> [Evaluation of the solution] </reflection>
 <reward> [Float between 0.0 and 1.0] </reward> [/INST] [INST] [QUERY] {query} [/INST] [ASSISTANT] """
 
-    stream = client.text_generation(message, max_new_tokens=4096, stream=True, details=True, return_full_text=False)
     output = ""
+    try:
+        stream = client_mistral.text_generation(message, max_new_tokens=2048, stream=True, details=True, return_full_text=False)
+        for response in stream:
+            output += response.token.text
+            yield output
+            time.sleep(0.3)  # Prevent rate limiting
+    except Exception as e:
+        yield f"Error: {str(e)}"
 
-    for response in stream:
-        output += response.token.text
-    return output
-
-description="# Chat GO\n### Enter your query and Press enter and get lightning fast response"
+# Description for Gradio UI
+description = "# Chat GO\n### Enter your query and get lightning-fast responses"
 
+# Gradio Interfaces
 with gr.Blocks() as demo1:
-    gr.Interface(description=description,fn=models, inputs=["text"], outputs="text")
+    gr.Interface(fn=models, inputs=["text"], outputs="text", description=description)
+
 with gr.Blocks() as demo2:
-    gr.Interface(description="Very low but critical thinker",fn=nemo, inputs=["text"], outputs="text", api_name="critical_thinker", concurrency_limit=10)
+    gr.Interface(fn=nemo, inputs=["text"], outputs="text", description="Critical Thinking Mode", api_name="critical_thinker", concurrency_limit=5)
 
+# Create a tabbed interface
 with gr.Blocks() as demo:
-    gr.TabbedInterface([demo1, demo2] , ["Fast", "Critical"])
+    gr.TabbedInterface([demo1, demo2], ["Fast", "Critical"])
 
-demo.queue(max_size=300000)
+# Launch the application
+demo.queue(max_size=100000)  # Adjust max queue size
 demo.launch()