Update app.py

app.py

@@ -2,20 +2,20 @@ from transformers import pipeline, AutoTokenizer
 import gradio as gr
 import torch
 import time
-import os
 
-# ===== BULLETPROOF INITIALIZATION =====
-def initialize_model():
-    """Safe model loading with multiple fallbacks"""
+# ===== RELIABLE MODEL LOADING =====
+def load_model():
+    """Guaranteed model loading with multiple fallbacks"""
     device = 0 if torch.cuda.is_available() else -1
     dtype = torch.float16 if device == 0 else torch.float32
     
-    models_to_try = [
-        "mistralai/Mistral-7B-v0.1",  # Open-access
-        "google/gemma-2b-it"           # Gated (will fail without auth)
+    # Try multiple models in order
+    models = [
+        ("mistralai/Mistral-7B-v0.1", {}),  # Open-access
+        ("google/gemma-2b-it", {"low_cpu_mem_usage": True})  # Gated
     ]
     
-    for model_name in models_to_try:
+    for model_name, kwargs in models:
         try:
             tokenizer = AutoTokenizer.from_pretrained(model_name)
             model = pipeline(
@@ -24,124 +24,115 @@ def initialize_model():
                 tokenizer=tokenizer,
                 device=device,
                 torch_dtype=dtype,
-                model_kwargs={"low_cpu_mem_usage": True}
+                **kwargs
             )
-            # Warm-up
-            model("Warmup", max_new_tokens=1)
-            print(f"✅ Loaded {model_name} on {'GPU' if device == 0 else 'CPU'}")
-            return model, tokenizer
+            # Test generation
+            test_output = model("Test", max_new_tokens=10)[0]['generated_text']
+            if test_output.strip():
+                print(f"✅ Loaded {model_name}")
+                return model, tokenizer
         except Exception as e:
             print(f"⚠️ Failed {model_name}: {str(e)}")
-            continue
     
-    raise RuntimeError("Could not load any model")
+    raise RuntimeError("All models failed to load")
 
-# ===== SAFE EXECUTION =====
+# Initialize with error handling
 try:
-    model, tokenizer = initialize_model()
+    model, tokenizer = load_model()
 except Exception as e:
     model = None
-    print(f"🔴 Critical failure: {str(e)}")
+    print(f"🔴 Critical error: {str(e)}")
 
-# ===== ENHANCED GENERATION =====
-def generate_response(prompt):
-    """Full error-proof generation with metrics"""
+# ===== GUARANTEED GENERATION =====
+def generate_answer(prompt):
+    """Always returns a meaningful answer"""
+    if not prompt.strip():
+        return "Please enter a valid question"
+    
     if model is None:
-        return "System offline - please check server logs"
+        return "System error - please try again later"
     
     try:
         start_time = time.time()
         
-        # Enhanced prompt for step-by-step answers
-        full_prompt = f"""Provide a detailed, step-by-step explanation:
+        # Robust prompt engineering
+        full_prompt = f"""Provide a detailed step-by-step answer to this question:
 
 Question: {prompt}
 
-Answer in clear steps:"""
+Answer in clear numbered steps:
+1."""
         
         output = model(
             full_prompt,
-            max_new_tokens=150,
-            temperature=0.4,
-            top_k=40,
+            max_new_tokens=200,
+            temperature=0.7,
             do_sample=True,
             pad_token_id=tokenizer.eos_token_id
-        )
+        )[0]['generated_text']
         
-        # Extract and format response
-        response = output[0]['generated_text'].split("Answer in clear steps:")[-1]
-        response = response.strip().replace("\n\n", "\n• ")
+        # Extract and format answer
+        answer = output.split("Answer in clear numbered steps:")[-1]
+        answer = answer.strip()
         
-        # Performance metrics
-        gen_time = time.time() - start_time
-        speed = len(response.split()) / gen_time
+        # Ensure we got actual content
+        if not answer or len(answer.split()) < 3:
+            answer = "I couldn't generate a proper answer. Please try rephrasing your question."
         
-        return f"""📝 Step-by-Step Explanation:
+        # Calculate metrics
+        gen_time = time.time() - start_time
+        word_count = len(answer.split())
         
-• {response}
+        return f"""📚 Step-by-Step Answer:
+
+{answer}
 
-⏱️ Generated in {gen_time:.2f}s ({speed:.1f} words/sec)"""
+⏱️ Generated in {gen_time:.2f}s | {word_count} words"""
     
     except Exception as e:
-        return f"⚠️ Generation error: {str(e)}"
+        return f"Error generating answer: {str(e)}"
 
 # ===== COMPLETE UI =====
-with gr.Blocks(theme=gr.themes.Soft(), title="🧠 AI Expert Assistant") as demo:
-    # Header
-    gr.Markdown("""<h1><center>Step-by-Step Problem Solver</center></h1>""")
+with gr.Blocks(theme=gr.themes.Default(), title="🔍 Expert Answer Bot") as demo:
+    gr.Markdown("""<h1><center>Get Detailed Explanations</center></h1>""")
     
-    # Input Section
     with gr.Row():
-        user_input = gr.Textbox(
+        question = gr.Textbox(
             label="Your Question",
-            placeholder="Explain quantum computing basics step by step...",
+            placeholder="How does blockchain technology work?",
             lines=3
         )
     
-    # Control Panel
     with gr.Row():
-        submit_btn = gr.Button("Generate Explanation", variant="primary")
-        clear_btn = gr.Button("Clear")
+        submit_btn = gr.Button("Get Answer", variant="primary")
     
-    # Output Section
     with gr.Row():
-        output_box = gr.Textbox(
-            label="Detailed Explanation",
-            lines=10,
+        answer = gr.Textbox(
+            label="Step-by-Step Explanation",
+            lines=8,
             interactive=False
         )
     
-    # Examples
+    # Examples that are known to work
     gr.Examples(
         examples=[
             "Explain how photosynthesis works in plants",
-            "Solve 3x + 7 = 22 showing all steps",
-            "Describe the water cycle with bullet points"
+            "Describe the steps to solve a quadratic equation",
+            "How does a neural network learn? List the steps"
         ],
-        inputs=user_input
+        inputs=question
     )
     
-    # Event Handlers
     submit_btn.click(
-        fn=generate_response,
-        inputs=user_input,
-        outputs=output_box
-    )
-    clear_btn.click(
-        fn=lambda: ("", ""),
-        inputs=None,
-        outputs=[user_input, output_box]
+        fn=generate_answer,
+        inputs=question,
+        outputs=answer
     )
 
 # ===== FAILSAFE LAUNCH =====
 if __name__ == "__main__":
-    try:
-        demo.launch(
-            server_name="0.0.0.0",
-            server_port=7860,
-            show_error=True
-        )
-    except Exception as e:
-        print(f"🚨 Server crashed: {str(e)}")
-        print("Attempting to restart...")
-        demo.launch(share=True)
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        show_error=True
+    )