Update app.py

app.py

@@ -187,16 +187,6 @@ def respond(
         yield custom_response
         return
 
-    if is_image_request(message):
-        try:
-            image = generate_image(message)
-            if image:
-                return f"Here's your generated image based on: {message}"
-            else:
-                return "Sorry, I couldn't generate the image. Please try again."
-        except Exception as e:
-            return f"An error occurred while generating the image: {str(e)}"
-
     # Prepare conversation history
     messages = [{"role": "system", "content": system_message}]
     for val in history:
@@ -207,18 +197,61 @@ def respond(
     
     messages.append({"role": "user", "content": message})
 
-    # Get response from model
-    response = ""
-    for message in text_client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-        response += token
-        yield response
+    # Check if the user wants a fast or critical response
+    if "critical" in message.lower():
+        # Critical thinker logic (like the 'nemo' function)
+        budget = 3  # Starting budget for steps
+        step_response = ""
+        message_critical = f"""[INST] [SYSTEM] You are a helpful assistant in normal conversation.
+When given a problem to solve, you are an expert problem-solving assistant.
+Your task is to provide a detailed, step-by-step solution to a given question.
+Follow these instructions carefully:
+1. Read the given question carefully and reset counter between <count> and </count> to {budget} (maximum 3 steps).
+2. Think critically like a human researcher or scientist. Break down the problem using first principles to conceptually understand and answer the question.
+3. Generate a detailed, logical step-by-step solution.
+4. Enclose each step of your solution within <step> and </step> tags.
+5. You are allowed to use at most {budget} steps (starting budget), keep track of it by counting down within tags <count> </count>, STOP GENERATING MORE STEPS when hitting 0, you don't have to use all of them.
+6. Do a self-reflection when you are unsure about how to proceed, based on the self-reflection and reward, decide whether you need to return to the previous steps.
+7. After completing the solution steps, reorganize and synthesize the steps into the final answer within <answer> and </answer> tags.
+8. Provide a critical, honest, and subjective self-evaluation of your reasoning process within <reflection> and </reflection> tags.
+9. Assign a quality score to your solution as a float between 0.0 (lowest quality) and 1.0 (highest quality), enclosed in <reward> and </reward> tags.
+Example format:
+<count> [starting budget] </count>
+<step> [Content of step 1] </step>
+<count> [remaining budget] </count>
+<step> [Content of step 2] </step>
+<reflection> [Evaluation of the steps so far] </reflection>
+<reward> [Float between 0.0 and 1.0] </reward>
+<count> [remaining budget] </count>
+<step> [Content of step 3 or Content of some previous step] </step>
+<count> [remaining budget] </count>
+...
+<step> [Content of final step] </step>
+<count> [remaining budget] </count>
+<answer> [Final Answer] </answer> (must give final answer in this format)
+<reflection> [Evaluation of the solution] </reflection>
+<reward> [Float between 0.0 and 1.0] </reward> [/INST] [INST] [QUERY] {message} [/INST] [ASSISTANT] """
+        
+        # Send critical thinking message to client and stream response
+        stream = text_client.text_generation(message_critical, max_new_tokens=4096, stream=True, details=True, return_full_text=False)
+        critical_response = ""
+        for response in stream:
+            critical_response += response.token.text
+        yield critical_response
+        
+    else:
+        # Standard fast response logic (like 'models')
+        response = ""
+        for message in text_client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            token = message.choices[0].delta.content
+            response += token
+            yield response
 
     yield response
 
@@ -260,7 +293,6 @@ Focus on accuracy, respect for sensitive topics, transparency, and professionali
 Version: Xylaria-1.2.9
 """
 
-
 # Gradio chat interface
 demo = gr.ChatInterface(
     respond,
@@ -293,4 +325,4 @@ demo = gr.ChatInterface(
     ],
     css=custom_css
 )
-demo.launch()
+demo.launch()