app.py

import gradio as gr
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import time
from threading import Lock
import os

# Global variables for model caching
model = None
tokenizer = None
model_lock = Lock()

def load_model():
    """Load the trained model using standard transformers (CPU compatible)"""
    global model, tokenizer

    with model_lock:
        if model is None:
            try:
                print("🔄 Loading Career Guidance AI model...")
                model_path = "./gemma_career_final"

                # Load tokenizer
                tokenizer = AutoTokenizer.from_pretrained(model_path)

                # Add pad token if missing
                if tokenizer.pad_token is None:
                    tokenizer.pad_token = tokenizer.eos_token

                # Load model for CPU inference
                model = AutoModelForCausalLM.from_pretrained(
                    model_path,
                    torch_dtype=torch.float32,  # Use float32 for CPU
                    device_map="cpu",           # Force CPU
                    low_cpu_mem_usage=True,     # Optimize CPU memory
                    trust_remote_code=True      # Trust model code
                )

                # Set to evaluation mode
                model.eval()

                print("✅ Model loaded successfully on CPU!")
                return True

            except Exception as e:
                print(f"❌ Error loading model: {str(e)}")
                print("📝 Trying fallback loading method...")

                try:
                    # Fallback: Load base model if fine-tuned model fails
                    base_model = "google/gemma-2-2b-it"
                    print(f"🔄 Loading base model: {base_model}")

                    tokenizer = AutoTokenizer.from_pretrained(base_model)
                    if tokenizer.pad_token is None:
                        tokenizer.pad_token = tokenizer.eos_token

                    model = AutoModelForCausalLM.from_pretrained(
                        base_model,
                        torch_dtype=torch.float32,
                        device_map="cpu",
                        low_cpu_mem_usage=True
                    )
                    model.eval()

                    print("✅ Base model loaded successfully!")
                    return True

                except Exception as fallback_error:
                    print(f"❌ Fallback loading failed: {str(fallback_error)}")
                    return False

    return True

def generate_career_response(message, history):
    """Generate career guidance response using transformers"""

    # Load model if not loaded
    if not load_model():
        return "❌ I'm having trouble loading. Please refresh and try again."

    # Handle empty messages
    if not message.strip():
        return "Please ask me a career-related question! I'm here to help with career planning, job search, interviews, skills, and professional development."

    try:
        # Format the conversation prompt for Gemma
        prompt = f"""<start_of_turn>user
{message}<end_of_turn>
<start_of_turn>model
"""

        # Tokenize input
        inputs = tokenizer(
            prompt,
            return_tensors="pt",
            max_length=1024,           # Limit input length for CPU
            truncation=True,
            padding=True
        )

        # Generate response with CPU-optimized settings
        with torch.no_grad():
            outputs = model.generate(
                inputs.input_ids,
                attention_mask=inputs.attention_mask,
                max_new_tokens=200,        # Shorter for faster CPU inference
                temperature=0.7,
                do_sample=True,
                top_p=0.9,
                top_k=50,
                repetition_penalty=1.1,
                pad_token_id=tokenizer.pad_token_id,
                eos_token_id=tokenizer.eos_token_id,
                no_repeat_ngram_size=3     # Reduce repetition
            )

        # Decode response
        response = tokenizer.decode(outputs[0], skip_special_tokens=False)

        # Extract model response
        if "<start_of_turn>model" in response:
            response = response.split("<start_of_turn>model")[-1]

        if "<end_of_turn>" in response:
            response = response.split("<end_of_turn>")[0]

        response = response.strip()

        # Fallback responses for common issues
        if not response or len(response.split()) < 5:
            if "career" in message.lower() or "job" in message.lower():
                response = "I'd be happy to help with your career question! Could you provide more specific details about what aspect of your career you'd like guidance on?"
            else:
                response = "I specialize in career guidance and professional development. I can help with career planning, job search strategies, interview preparation, skill development, and professional growth. How can I assist with your career goals?"

        return response

    except Exception as e:
        print(f"💥 Generation error: {str(e)}")

        # Provide helpful fallback response based on query type
        career_keywords = ["career", "job", "interview", "resume", "skill", "work", "salary", "promotion"]
        if any(keyword in message.lower() for keyword in career_keywords):
            return """I understand you're looking for career guidance. While I'm experiencing some technical difficulties with my AI processing, here are some general tips:

**For Career Planning:**
- Identify your strengths and interests
- Research industry trends and requirements
- Network with professionals in your field
- Consider additional training or certifications

**For Job Search:**
- Tailor your resume to each position
- Practice common interview questions
- Build a strong LinkedIn profile
- Apply consistently and follow up professionally

Would you like to try rephrasing your question? I'll do my best to provide helpful career advice!"""
        else:
            return "I'm a career guidance assistant. I can help with career planning, job interviews, skill development, and professional growth. What career-related question can I help you with?"

# Enhanced CSS for professional appearance
css = """
#chatbot {
    height: 650px !important;
}

.gradio-container {
    max-width: 900px !important;
    margin: auto !important;
}

.message.user {
    background-color: #f0f8ff !important;
    border-left: 4px solid #007bff !important;
    padding-left: 15px !important;
    margin: 10px 0 !important;
}

.message.bot {
    background-color: #f8f9fa !important;
    border-left: 4px solid #28a745 !important;
    padding-left: 15px !important;
    margin: 10px 0 !important;
}

.gradio-container .wrap {
    max-width: 100% !important;
}

#component-0 {
    max-height: none !important;
}
"""

# Career guidance examples
examples = [
    ["What skills do I need to become a data scientist?"],
    ["How should I prepare for a software engineering interview?"],
    ["What's the best career path for someone interested in AI?"],
    ["How do I transition from marketing to product management?"],
    ["What certifications are valuable for cybersecurity careers?"],
    ["How do I negotiate salary in my first job?"],
    ["What should I include in my LinkedIn profile?"],
    ["How do I network effectively in my industry?"]
]

# Create the Gradio interface
with gr.Blocks(css=css, title="Career Guidance AI Assistant", theme=gr.themes.Soft()) as demo:

    gr.HTML("""
    <div style="text-align: center; padding: 20px;">
        <h1 style="color: #007bff; margin-bottom: 10px;">🚀 Career Guidance AI Assistant</h1>
        <p style="font-size: 18px; color: #666; max-width: 700px; margin: 0 auto;">
            Your personal AI career advisor, ready to help with career planning, job search strategies, 
            interview preparation, and professional development guidance.
        </p>
    </div>
    """)

    with gr.Row():
        with gr.Column():
            gr.Markdown("""
            ### 💼 I can help you with:
            - **Career Planning** & goal setting
            - **Job Search** strategies & tips  
            - **Interview Preparation** & practice
            - **Skill Development** recommendations
            - **Resume & LinkedIn** optimization
            - **Salary Negotiation** guidance
            - **Career Transitions** & pivots
            - **Professional Networking** strategies
            """)

    # Main chat interface
    chatbot = gr.ChatInterface(
        generate_career_response,
        chatbot=gr.Chatbot(
            elem_id="chatbot",
            height=600,
            show_label=True,
            show_copy_button=True,
            bubble_full_width=False,
            avatar_images=("👨‍💼", "🤖"),
            show_share_button=False
        ),
        textbox=gr.Textbox(
            placeholder="💬 Ask me anything about careers, jobs, interviews, skills, or professional development...",
            container=False,
            scale=7,
            max_lines=3
        ),
        title=None,  # Already added above
        retry_btn="🔄 Try Again",
        undo_btn="↩️ Undo Last",
        clear_btn="🗑️ Clear Chat",
        submit_btn="Send 📤"
    )

    # Example questions section
    with gr.Row():
        with gr.Column():
            gr.Markdown("### 💡 Try these example questions:")
            with gr.Row():
                for i in range(0, len(examples), 2):
                    with gr.Column():
                        if i < len(examples):
                            gr.Examples(
                                examples=[examples[i]],
                                inputs=chatbot.textbox,
                                label=None
                            )
                        if i + 1 < len(examples):
                            gr.Examples(
                                examples=[examples[i + 1]],
                                inputs=chatbot.textbox,
                                label=None
                            )

    # Footer section
    gr.HTML("""
    <div style="margin-top: 40px; padding: 20px; background-color: #f8f9fa; border-radius: 10px;">
        <div style="text-align: center;">
            <h3 style="color: #007bff; margin-bottom: 15px;">📋 Important Notes</h3>
            <div style="display: flex; justify-content: space-around; flex-wrap: wrap; gap: 20px;">
                <div>
                    <strong>🔒 Privacy</strong><br>
                    <small>Conversations are not stored</small>
                </div>
                <div>
                    <strong>⚡ Response Time</strong><br>
                    <small>~10-30 seconds per query</small>
                </div>
                <div>
                    <strong>🎯 Specialization</strong><br>
                    <small>Career guidance & professional development</small>
                </div>
                <div>
                    <strong>📝 Disclaimer</strong><br>
                    <small>General guidance - consult professionals for specific advice</small>
                </div>
            </div>
        </div>
    </div>
    """)

# Launch configuration
if __name__ == "__main__":
    demo.launch(
        server_name="0.0.0.0",
        server_port=7860,
        show_error=True,
        share=False
    )