app.py

"""
Rakesh Kumar Portfolio Chatbot
Gradio app with Groq LLM integration and session management
"""
import os
import uuid
import time
import gradio as gr
from groq import Groq

# Session management with 1-hour TTL
sessions = {}
SESSION_TTL = 3600  # 1 hour in seconds

def cleanup_expired_sessions():
    """Remove sessions older than TTL"""
    current_time = time.time()
    expired = [sid for sid, data in sessions.items() if current_time - data['created_at'] > SESSION_TTL]
    for sid in expired:
        del sessions[sid]

def get_or_create_session(session_id: str) -> list:
    """Get existing session or create new one"""
    cleanup_expired_sessions()
    
    if session_id not in sessions:
        sessions[session_id] = {
            'history': [],
            'created_at': time.time()
        }
    return sessions[session_id]['history']

# System prompt with Rakesh's background
SYSTEM_PROMPT = """You are a helpful AI assistant representing Rakesh Kumar's portfolio. Answer questions about Rakesh's experience, skills, projects, and background.

## About Rakesh Kumar
- **Current Role**: AI Product Manager & AI Lead at Dr. Reddy's (Svaas Wellness) since Feb 2023
- **Experience**: 7+ years in AI, healthcare, and e-commerce
- **Education**: MBA from IIM Ahmedabad (2018-2020), B.Tech from IIT Ropar (2011-2015)
- **Location**: India
- **Contact**: p18rakeshk@iima.ac.in | +91-9004260456

## Key Projects

### AI Playground
Low-code platform for business owners to create prototype AI agents. Features RAG pipeline creation, evaluation frameworks, orchestration layer, database integration hooks, and configuration layer for validating model responses.

### Medical Deep Research Agent
Secondary research tool for pharma teams covering epidemiology, clinical trials, pharmacokinetics, safety, unmet needs. Built with RAG + Knowledge Graph, multi-agent architecture using SLMs and TinyLLM for knowledge extraction.

### SemaAI Agent
Fully on-premise enterprise chatbot using TinyLLM, LoRA fine-tuning, RAG-grounded responses, and custom MLP layer for intent classification.

### Recommendation Engine (eShakti)
Hybrid recommendation system using User-User, Item-Item, User-Item collaborative filtering with SVD, dimensionality reduction, and neural networks. Achieved +$50k/month revenue, +7% conversion.

## Key Achievements
- 70%+ manual tasks reduced using agentic AI workflows
- 80% faster insight generation with BI dashboards
- 10× clinical research turnaround acceleration
- 85% ML + LLM diet engine accuracy
- 20k+ users onboarded in 90 days

## Skills
Product Strategy, Roadmapping, AI/ML, LLM & RAG, GenAI Voice, Experimentation, SQL, Python, Data Platforms, Health Tech, E-commerce

## Certifications
IBM AI Product Manager, IBM AI Developer, Google Digital Marketing, Six Sigma Green Belt

Be concise, professional, and helpful. If asked something not about Rakesh, politely redirect to portfolio-related topics."""

# Initialize Groq client
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))

def chat(message: str, history: list, request: gr.Request) -> str:
    """Process chat message with Groq LLM"""
    
    # Get session ID from cookies or create new one
    session_id = str(uuid.uuid4()) if not hasattr(request, 'session_hash') else request.session_hash
    
    # Get session history
    session_history = get_or_create_session(session_id)
    
    # Build messages for Groq
    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
    
    # Add session history
    for h in session_history:
        messages.append({"role": "user", "content": h[0]})
        messages.append({"role": "assistant", "content": h[1]})
    
    # Add current message
    messages.append({"role": "user", "content": message})
    
    try:
        # Call Groq API
        completion = client.chat.completions.create(
            model="llama-3.3-70b-versatile",
            messages=messages,
            temperature=0.7,
            max_tokens=1024,
        )
        
        response = completion.choices[0].message.content
        
        # Update session history
        session_history.append((message, response))
        
        return response
        
    except Exception as e:
        return f"I apologize, but I encountered an error. Please try again. Error: {str(e)}"

# Create Gradio interface
with gr.Blocks(
    title="Chat with Rakesh's Portfolio",
) as demo:
    gr.Markdown("""
    # 💬 Ask About Rakesh Kumar
    *AI Product Manager & AI Lead*
    
    Ask me anything about Rakesh's experience, projects, skills, or background!
    """)
    
    chatbot = gr.ChatInterface(
        fn=chat,
        examples=[
            "What is Rakesh's current role?",
            "Tell me about the AI Playground project",
            "What are Rakesh's key achievements?",
            "What is his educational background?",
        ],
    )

if __name__ == "__main__":
    demo.launch()