app.py

import os
import torch
import gradio as gr
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.document_loaders import TextLoader
from langchain_community.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains import RetrievalQA
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from langchain_huggingface import HuggingFacePipeline

# Detect device (GPU or CPU)
device = "cuda" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if device == "cuda" else torch.float32

# ✅ Step 1: Check if dataset exists
DATASET_FILE = "dataset.txt"
if not os.path.exists(DATASET_FILE):
    raise FileNotFoundError(f"❌ Error: '{DATASET_FILE}' not found. Please create and add some text.")

# ✅ Step 2: Load and split dataset
print("📄 Loading dataset...")
doc_loader = TextLoader(DATASET_FILE)
docs = doc_loader.load()

print("🔀 Splitting documents...")
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
split_docs = text_splitter.split_documents(docs)

# ✅ Step 3: Initialize FAISS Vector Store
print("🧠 Creating embeddings...")
embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")

print("📌 Creating FAISS Vector DB...")
vectordb = FAISS.from_documents(split_docs, embeddings)
retriever = vectordb.as_retriever(search_kwargs={"k": 5})

# ✅ Step 4: Load Hugging Face Model
MODEL_NAME = "mistralai/Mistral-7B-Instruct"  # More memory-efficient than Yi-Coder-9B
print(f"🚀 Loading Model: {MODEL_NAME} on {device}...")

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
model = AutoModelForCausalLM.from_pretrained(
    MODEL_NAME,
    device_map="auto",
    torch_dtype=torch_dtype,
)

# ✅ Step 5: Create QA Pipeline
print("🔧 Setting up pipeline...")
qa_pipeline = pipeline(
    "text-generation",
    model=model,
    tokenizer=tokenizer,
    max_new_tokens=750,
    pad_token_id=tokenizer.eos_token_id,
)

llm = HuggingFacePipeline(pipeline=qa_pipeline)

qa_chain = RetrievalQA.from_chain_type(
    retriever=retriever,
    chain_type="stuff",
    llm=llm,
    return_source_documents=False
)

# ✅ Step 6: Define Helper Functions
def preprocess_query(query):
    """Prepares query text based on type of request"""
    if "script" in query.lower() or "code" in query.lower():
        return f"Write a CPSL script: {query}"
    return query

def clean_response(response):
    """Cleans and extracts the response from model output"""
    result = response.get("result", "")
    if "Answer:" in result:
        return result.split("Answer:")[1].strip()
    return result.strip()

def chatbot_response(user_input):
    """Processes user input and returns AI response"""
    processed_query = preprocess_query(user_input)
    raw_response = qa_chain.invoke({"query": processed_query})
    return clean_response(raw_response)

# ✅ Step 7: Build Gradio UI
with gr.Blocks() as chat_interface:
    gr.Markdown("# 🤖 CPSL Chatbot")
    chat_history = gr.Chatbot(label="Chat History", height=300)
    user_input = gr.Textbox(label="Your Message:")
    send_button = gr.Button("Send")
    clear_button = gr.Button("Clear")

    def interact(user_message, history):
        """Handles user interaction and updates chat history"""
        bot_reply = chatbot_response(user_message)
        history.append(("🧑‍💻 You:", user_message))
        history.append(("🤖 Bot:", bot_reply))
        return history, ""

    send_button.click(interact, inputs=[user_input, chat_history], outputs=[chat_history, user_input])
    clear_button.click(lambda: [], outputs=[chat_history])

# ✅ Step 8: Run Gradio App
if __name__ == "__main__":
    print("🚀 Launching CPSL Chatbot...")
    chat_interface.launch(share=True)  # share=True allows public link