app.py

import os
import streamlit as st
from PyPDF2 import PdfReader
import numpy as np
from groq import Groq
import faiss

# Set up Groq API client
groq_client = Groq(api_key="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")

# Function to extract text from PDF
def extract_pdf_content(pdf_file):
    reader = PdfReader(pdf_file)
    content = ""
    for page in reader.pages:
        content += page.extract_text()
    return content

# Function to split content into chunks
def chunk_text(text, chunk_size=500):
    words = text.split()
    return [" ".join(words[i:i + chunk_size]) for i in range(0, len(words), chunk_size)]

# Function to compute embeddings using Groq's Llama3-70B-8192 model
def compute_embeddings(text_chunks):
    embeddings = []
    for chunk in text_chunks:
        response = groq_client.chat.completions.create(
            messages=[{"role": "user", "content": chunk}],
            model="llama3-70b-8192"
        )
        embeddings.append(np.array(response['choices'][0]['message']['content']))
    return np.array(embeddings)

# Function to build FAISS index
def build_faiss_index(embeddings):
    dimension = embeddings.shape[1]
    index = faiss.IndexFlatL2(dimension)  # L2 distance for similarity
    index.add(embeddings)
    return index

# Function to search in FAISS index
def search_faiss_index(index, query_embedding, text_chunks, top_k=3):
    distances, indices = index.search(query_embedding, top_k)
    return [(text_chunks[idx], distances[0][i]) for i, idx in enumerate(indices[0])]

# Function to generate professional content using Groq's Llama3-70B-8192 model
def generate_professional_content_groq(topic):
    response = groq_client.chat.completions.create(
        messages=[{"role": "user", "content": f"Explain '{topic}' in bullet points, highlighting key concepts, examples, and applications for electrical engineering students."}],
        model="llama3-70b-8192"
    )
    return response['choices'][0]['message']['content'].strip()

# Function to compute query embedding using Groq's Llama3-70B-8192 model
def compute_query_embedding(query):
    response = groq_client.chat.completions.create(
        messages=[{"role": "user", "content": query}],
        model="llama3-70b-8192"
    )
    return np.array(response['choices'][0]['message']['content']).reshape(1, -1)

# Streamlit app
st.title("Generative AI for Electrical Engineering Education with FAISS and Groq")
st.sidebar.header("AI-Based Tutor with Vector Search")

# File upload section
uploaded_file = st.sidebar.file_uploader("Upload Study Material (PDF)", type=["pdf"])
topic = st.sidebar.text_input("Enter a topic (e.g., Newton's Third Law)")

if uploaded_file:
    # Extract and process file content
    content = extract_pdf_content(uploaded_file)
    st.sidebar.success(f"{uploaded_file.name} uploaded successfully!")

    # Chunk and compute embeddings
    chunks = chunk_text(content)
    embeddings = compute_embeddings(chunks)

    # Build FAISS index
    index = build_faiss_index(embeddings)

    st.write("**File Processed and Indexed for Search**")
    st.write(f"Total chunks created: {len(chunks)}")

# Generate study material
if st.button("Generate Study Material"):
    if topic:
        st.header(f"Study Material: {topic}")
        
        # Compute query embedding
        query_embedding = compute_query_embedding(topic)

        # Search FAISS index
        if uploaded_file:
            results = search_faiss_index(index, query_embedding, chunks, top_k=3)
            st.write("**Relevant Content from Uploaded File:**")
            for result, distance in results:
                st.write(f"- {result} (Similarity: {distance:.2f})")
        else:
            st.warning("No file uploaded. Generating AI-based content instead.")
        
        # Generate content using Groq's Llama3-70B-8192 model
        ai_content = generate_professional_content_groq(topic)
        st.write("**AI-Generated Content (Groq - Llama3-70B-8192):**")
        st.write(ai_content)
    else:
        st.warning("Please enter a topic!")