app.py

import streamlit as st
import openai
import anthropic
import pandas as pd
import numpy as np
import time
from sklearn.metrics.pairwise import cosine_similarity
from sentence_transformers import SentenceTransformer
import os

# Initialize clients
openai.api_key = os.getenv("OPENAI_API_KEY")
anthropic_client = anthropic.Client(api_key=os.getenv("ANTHROPIC_API_KEY"))

# Load sentence transformer for similarity scoring
similarity_model = SentenceTransformer('all-MiniLM-L6-v2')

# ===== Helper Functions =====
def generate_response(prompt, model):
    """Generate responses from different models."""
    if model == "GPT-4":
        response = openai.chat.completions.create(
            model="gpt-4",
            messages=[{"role": "user", "content": prompt}]
        )
        return response.choices[0].message.content.strip()
    elif model == "Claude 3":
        response = anthropic_client.messages.create(
            model="claude-3-sonnet-20240229",
            max_tokens=1000,
            messages=[{"role": "user", "content": prompt}]
        )
        return response.content[0].text
    else:
        return "Model not supported."

def calculate_similarity(text1, text2):
    """Calculate cosine similarity between two texts."""
    embeddings = similarity_model.encode([text1, text2])
    return cosine_similarity([embeddings[0]], [embeddings[1]])[0][0]

def evaluate_response(response, golden_response):
    """Evaluate response quality."""
    similarity = calculate_similarity(response, golden_response)
    return {
        "Similarity Score": similarity,
        "Length": len(response),
        "Toxicity": "Low"  # Placeholder for actual toxicity detection
    }

def optimize_prompt(prompt, target_metrics):
    """Optimize prompt using a simple mutation strategy."""
    # Placeholder for genetic algorithm implementation
    optimized_prompt = prompt + " [Optimized for clarity and brevity]"
    return optimized_prompt

# ===== Streamlit UI =====
st.title("🚀 PromptCraft Studio")
st.write("The Ultimate Prompt Engineering Workbench")

# Sidebar for settings
with st.sidebar:
    st.header("⚙️ Settings")
    model_choice = st.selectbox("Choose Model", ["GPT-4", "Claude 3"])
    optimization_goal = st.selectbox("Optimization Goal", ["Clarity", "Brevity", "Engagement"])

# Main interface
st.markdown("### **Prompt Playground**")
user_prompt = st.text_area("Enter your prompt:", "Explain quantum computing in simple terms.")

if st.button("Generate Response"):
    with st.spinner("Generating response..."):
        response = generate_response(user_prompt, model_choice)
        st.markdown("### **Response**")
        st.write(response)

        # Evaluate response
        st.markdown("### **Evaluation Metrics**")
        golden_response = "Quantum computing uses qubits to perform complex calculations faster than classical computers."
        metrics = evaluate_response(response, golden_response)
        st.write(metrics)

# Prompt Optimization
st.markdown("### **Prompt Optimization**")
if st.button("Optimize Prompt"):
    optimized_prompt = optimize_prompt(user_prompt, optimization_goal)
    st.markdown("### **Optimized Prompt**")
    st.write(optimized_prompt)

# Multi-Model Comparison
st.markdown("### **Multi-Model Comparison**")
if st.button("Compare Models"):
    col1, col2 = st.columns(2)
    with col1:
        st.markdown("#### **GPT-4 Response**")
        gpt_response = generate_response(user_prompt, "GPT-4")
        st.write(gpt_response)
    with col2:
        st.markdown("#### **Claude 3 Response**")
        claude_response = generate_response(user_prompt, "Claude 3")
        st.write(claude_response)

    # Compare similarity
    similarity = calculate_similarity(gpt_response, claude_response)
    st.markdown(f"**Similarity Between Models:** {similarity:.2f}")

# Bulk Processing
st.markdown("### **Bulk Prompt Testing**")
uploaded_file = st.file_uploader("Upload a CSV with prompts", type=["csv"])
if uploaded_file:
    df = pd.read_csv(uploaded_file)
    if "Prompt" in df.columns:
        results = []
        for prompt in df["Prompt"]:
            response = generate_response(prompt, model_choice)
            results.append(response)
        df["Response"] = results
        st.write(df)
        st.download_button("Download Results", df.to_csv(index=False).encode('utf-8'), "results.csv", "text/csv")
    else:
        st.error("CSV must contain a 'Prompt' column.")

# Feedback
st.markdown("### **Feedback**")
if st.button("👍 Good Response"):
    st.toast("Thanks for your feedback!")
if st.button("👎 Needs Improvement"):
    st.toast("We'll do better next time!")