app.py

import os
import gradio as gr
from langchain.chains import RetrievalQA
from langchain_community.vectorstores import Chroma  # ✅ Fixed Import
from langchain.llms import OpenAI
from langchain_huggingface import HuggingFaceEndpoint  # ✅ Corrected Import
from langchain.embeddings import OpenAIEmbeddings
from langchain_community.embeddings import HuggingFaceEmbeddings  # ✅ Corrected Import
from langchain_community.document_loaders import PyPDFLoader  # ✅ Corrected Import
import time

# Define paths for cybersecurity training PDFs
PDF_FILES = [
    "ISOIEC 27001_2ef522.pdf",
    "ISO-IEC-27005-2022.pdf",
    "MITRE ATLAS Overview Combined_v1.pdf",
    "NIST_CSWP_04162018.pdf"
]

# Fetch Hugging Face API token securely from environment variables
HUGGINGFACE_API_KEY = os.getenv("HUGGINGFACEHUB_API_TOKEN")
if HUGGINGFACE_API_KEY is None:
    raise ValueError("❌ Hugging Face API token is missing! Set it in Hugging Face Spaces Secrets.")

# Load PDFs into ChromaDB
def load_data():
    """Loads multiple PDFs and stores embeddings in ChromaDB"""
    all_docs = []
    for pdf in PDF_FILES:
        if os.path.exists(pdf):  # Ensure the PDF exists in the Hugging Face Space
            loader = PyPDFLoader(pdf)
            all_docs.extend(loader.load())

    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")  # ✅ Use updated embeddings

    return Chroma.from_documents(all_docs, embeddings)

# Load the knowledge base
vector_db = load_data()

# Load LLM from Hugging Face securely
llm = HuggingFaceEndpoint(
    repo_id="google/flan-t5-large",
    temperature=0.5,  # ✅ Explicitly set temperature
    max_length=512,  # ✅ Explicitly set max_length
    huggingfacehub_api_token=HUGGINGFACE_API_KEY
)

# Create Retrieval QA chain
qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=vector_db.as_retriever())

# Function to simulate futuristic typing effect
def chatbot_response(question):
    """Handles chatbot queries with a typing effect"""
    response = qa_chain.run(question)
    displayed_response = ""
    for char in response:
        displayed_response += char
        time.sleep(0.02)  # Simulate typing delay
        yield displayed_response

# Custom futuristic CSS styling
custom_css = """
body {background-color: #0f172a; color: #0ff; font-family: 'Orbitron', sans-serif;}
.gradio-container {background: linear-gradient(to bottom, #020c1b, #001f3f);}
textarea {background: #011627; color: #0ff; font-size: 18px;}
button {background: #0088ff; color: white; font-size: 20px; border-radius: 5px; border: none; padding: 10px;}
button:hover {background: #00ffff; color: #000;}
"""

# Create Gradio Chatbot Interface
iface = gr.Interface(
    fn=chatbot_response,
    inputs="text",
    outputs="text",
    title="🤖 Cybersecurity AI Assistant",
    description="Ask me about NIST, ISO/IEC 27001, MITRE ATLAS, and ISO/IEC 27005. Powered by AI.",
    theme="default",
    css=custom_css,
    live=True,  # Enables real-time updates for typing effect
)

# Launch chatbot
iface.launch()