Create app.py

app.py

@@ -0,0 +1,235 @@
+import gradio as gr
+from model_utils import CodeThinkingAssistant
+import time
+
+# Initialize the assistant
+# IMPORTANT: Replace with your fine-tuned model ID after training
+MODEL_ID = "meta-llama/Llama-3.2-1B-Instruct"  # CHANGE THIS to your model after fine-tuning
+
+print("🚀 Initializing Llama 3.2 Codex Assistant...")
+assistant = CodeThinkingAssistant(MODEL_ID)
+print("✅ Ready to help with coding!")
+
+# Custom CSS for better UI
+custom_css = """
+<style>
+    .thinking-mode {
+        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
+        padding: 2px;
+        border-radius: 10px;
+    }
+    .fast-mode {
+        background: linear-gradient(135deg, #f093fb 0%, #f5576c 100%);
+        padding: 2px;
+        border-radius: 10px;
+    }
+    .thinking-box {
+        background-color: #f0f4ff;
+        border-left: 4px solid #667eea;
+        padding: 10px;
+        margin: 10px 0;
+        border-radius: 5px;
+        font-family: monospace;
+    }
+    .code-box {
+        background-color: #1e1e1e;
+        color: #d4d4d4;
+        padding: 15px;
+        border-radius: 5px;
+        font-family: 'Courier New', monospace;
+        overflow-x: auto;
+    }
+    footer {
+        visibility: visible;
+        text-align: center;
+        margin-top: 20px;
+        font-size: 12px;
+    }
+</style>
+"""
+
+def format_code_output(code: str) -> str:
+    """Format code for display"""
+    return f'<div class="code-box"><pre>{code}</pre></div>'
+
+def format_thinking_output(thinking: str) -> str:
+    """Format thinking process for display"""
+    return f'<div class="thinking-box">💭 <strong>Thinking process:</strong><br>{thinking}</div>'
+
+def respond(
+    message: str,
+    history: list,
+    thinking_mode: bool,
+    reasoning_style: str,
+    temperature: float,
+    max_tokens: int
+):
+    """Main response function for Gradio chat"""
+    
+    if not message.strip():
+        yield "Please enter a coding question or task."
+        return
+    
+    # Show thinking indicator
+    yield "🤔 Thinking" + "." * 3
+    
+    try:
+        if thinking_mode:
+            if reasoning_style == "Step-by-step thinking":
+                result = assistant.generate_with_thinking(
+                    message,
+                    max_thought_tokens=300,
+                    max_code_tokens=max_tokens
+                )
+            else:  # Chain-of-thought
+                result = assistant.generate_with_chain_of_thought(message)
+            
+            # Format output with both thinking and code
+            formatted_output = ""
+            if result.get("thinking"):
+                formatted_output += format_thinking_output(result["thinking"])
+            if result.get("code"):
+                formatted_output += "\n\n" + format_code_output(result["code"])
+            
+            yield formatted_output
+        else:
+            # Fast mode
+            code = assistant.generate_fast(message, max_tokens=max_tokens)
+            yield format_code_output(code)
+            
+    except Exception as e:
+        yield f"❌ Error: {str(e)}\n\nPlease make sure the model is loaded correctly."
+
+# Build the Gradio interface
+with gr.Blocks(css=custom_css, title="Llama 3.2 Codex - AI Coding Assistant") as demo:
+    
+    gr.Markdown("""
+    # 🤖 Llama 3.2 1B Codex
+    
+    ### Your AI Pair Programmer with Thinking Mode
+    
+    Built with Llama 3.2 - Specialized for code generation with explicit reasoning.
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=2):
+            # Chat interface
+            chatbot = gr.Chatbot(
+                label="Code Assistant",
+                height=500,
+                bubble_full_width=False
+            )
+            
+            with gr.Row():
+                msg = gr.Textbox(
+                    label="Ask for code help",
+                    placeholder="Example: 'Write a function to sort a list of dictionaries by a key' or 'Explain this algorithm...'",
+                    scale=4
+                )
+                send_btn = gr.Button("Send", variant="primary", scale=1)
+            
+            with gr.Row():
+                clear_btn = gr.Button("Clear Chat")
+                example_btn = gr.Button("Load Example")
+        
+        with gr.Column(scale=1):
+            gr.Markdown("### ⚙️ Settings")
+            
+            thinking_mode = gr.Checkbox(
+                label="🧠 Enable Thinking Mode",
+                value=True,
+                info="Shows reasoning process before generating code"
+            )
+            
+            reasoning_style = gr.Radio(
+                choices=["Step-by-step thinking", "Chain-of-thought"],
+                value="Step-by-step thinking",
+                label="Reasoning style",
+                visible=True
+            )
+            
+            temperature = gr.Slider(
+                minimum=0.1,
+                maximum=1.5,
+                value=0.7,
+                step=0.1,
+                label="Temperature (creativity)",
+                info="Lower = more focused, Higher = more creative"
+            )
+            
+            max_tokens = gr.Slider(
+                minimum=100,
+                maximum=1500,
+                value=600,
+                step=50,
+                label="Max response length",
+                info="Maximum tokens in response"
+            )
+            
+            gr.Markdown("""
+            ---
+            ### 📌 Tips
+            - **Thinking Mode ON**: Best for complex problems
+            - **Thinking Mode OFF**: Faster responses for simple code
+            - Be specific in your requests
+            - Ask for explanations or optimizations
+            """)
+    
+    # Example prompts
+    examples = gr.Examples(
+        examples=[
+            "Write a Python function to find the longest common prefix in a list of strings",
+            "Implement a binary search tree with insert and search methods",
+            "Explain the difference between deep and shallow copy in Python with examples",
+            "Write a recursive function to generate all permutations of a string",
+            "Create a decorator that measures function execution time",
+            "Implement a simple URL shortener using dictionary",
+            "Write a function to check if two strings are anagrams",
+            "Create a class for a bank account with deposit, withdraw, and interest calculation"
+        ],
+        inputs=msg,
+        label="Example Prompts (click to try)"
+    )
+    
+    # Event handlers
+    def respond_wrapper(message, history, thinking_mode, reasoning_style, temperature, max_tokens):
+        response_generator = respond(message, history, thinking_mode, reasoning_style, temperature, max_tokens)
+        for response in response_generator:
+            history.append((message, response))
+            yield history, ""
+            # Reset after yielding
+            history = []
+            yield history, ""
+    
+    # Wire up the events
+    send_btn.click(
+        respond,
+        [msg, chatbot, thinking_mode, reasoning_style, temperature, max_tokens],
+        [chatbot, msg]
+    )
+    
+    msg.submit(
+        respond,
+        [msg, chatbot, thinking_mode, reasoning_style, temperature, max_tokens],
+        [chatbot, msg]
+    )
+    
+    clear_btn.click(lambda: None, None, chatbot, queue=False)
+    
+    example_btn.click(
+        lambda: "Write a function to check if a string is a palindrome (ignoring spaces, punctuation, and case)",
+        None,
+        msg
+    )
+    
+    # Footer with required attribution
+    gr.Markdown("""
+    ---
+    <footer>
+    <b>Built with Llama</b> • Llama 3.2 1B Codex • <a href="https://llama.meta.com/" target="_blank">Meta Llama 3.2</a><br>
+    Licensed under <a href="/LICENSE.txt">Llama 3.2 Community License</a>
+    </footer>
+    """)
+
+if __name__ == "__main__":
+    demo.launch(share=True)