app.py

import gradio as gr
import chromadb
from sentence_transformers import SentenceTransformer
from transformers import pipeline
import re

# --- 1. Load Models ---
print("Loading sentence-transformer model for retrieval...")
retriever_model = SentenceTransformer('all-MiniLM-L6-v2')
print("Retriever model loaded.")

# --- THIS IS THE UPDATED LINE ---
print("Loading generative model for answering (google/flan-t5-base)...")
# Using the balanced 'base' model for better performance and reliability.
generator_pipe = pipeline("text2text-generation", model="google/flan-t5-base", device=-1)
print("Generative model loaded.")
# --- END OF UPDATE ---


# --- 2. Setup ChromaDB ---
client = chromadb.Client()

try:
    collection = client.create_collection("whatsapp_chat_v2")
    print("ChromaDB collection created.")
    
    # --- Data Loading and Cleaning ---
    try:
        print("Loading data from my_data.txt...")
        with open('my_data.txt', 'r', encoding='utf-8') as f:
            lines = [line.strip() for line in f if line.strip()]
        
        message_pattern = re.compile(r'^\[?.*?\]?\s*.*?:\s*(.*)')
        
        cleaned_documents = []
        for line in lines:
            match = message_pattern.match(line)
            if match and match.group(1):
                cleaned_documents.append(match.group(1).strip())

        if not cleaned_documents:
            print("ERROR: Could not extract any valid messages from my_data.txt.")
            cleaned_documents = ["Error: The data file 'my_data.txt' could not be processed."]
        else:
            print(f"Successfully loaded and cleaned {len(cleaned_documents)} messages.")

        documents = cleaned_documents
        
    except FileNotFoundError:
        print("Error: my_data.txt not found.")
        documents = ["Error: my_data.txt not found. Please make sure the file is uploaded."]
    
    # --- Batch Processing ---
    batch_size = 5000 
    print("Starting to process and add documents in batches...")
    for i in range(0, len(documents), batch_size):
        end_i = min(i + batch_size, len(documents))
        batch_docs = documents[i:end_i]
        print(f"Processing batch of {len(batch_docs)} documents...")
        batch_embeddings = retriever_model.encode(batch_docs)
        batch_ids = [f"id_{j}" for j in range(i, end_i)]
        collection.add(
            embeddings=batch_embeddings.tolist(),
            documents=batch_docs,
            ids=batch_ids
        )
    print("All documents have been successfully added to ChromaDB.")

except ValueError:
    collection = client.get_collection("whatsapp_chat_v2")
    print("ChromaDB collection loaded.")


# --- 3. Define Chatbot Logic ---
def chatbot_response(message, history):
    query_embedding = retriever_model.encode([message]).tolist()
    results = collection.query(
        query_embeddings=query_embedding,
        n_results=5 # Using 5 results is a good balance for the base model
    )
    retrieved_documents = results['documents'][0]
    
    if not retrieved_documents or "Error:" in retrieved_documents[0]:
        return "I'm sorry, I couldn't find any relevant information in the chat history. 🤔"
    
    context = "\n- ".join(retrieved_documents)
    prompt = f"""
    Based on the following excerpts from a WhatsApp chat, provide a helpful and accurate answer to the user's question.

    Chat Context:
    - {context}

    Question:
    {message}

    Answer:
    """
    
    generated_text = generator_pipe(prompt, max_length=150, num_beams=5, early_stopping=True)
    response = generated_text[0]['generated_text']
    
    return response

# --- 4. Create the Gradio Interface ---
iface = gr.ChatInterface(
    fn=chatbot_response,
    title="WhatsApp Chat Bot ⚡️",
    description="Ask me anything about this WhatsApp chat history. (Powered by flan-t5-base)",
    theme="soft", 
    examples=["What was the final decision on the project deadline?", "Summarize the conversation about the event."],
    cache_examples=False
)

# Launch the app
iface.launch()