app.py

import gradio as gr
import cohere
import pickle
import os
import torch

# Step 1: Load the saved model, tokenizer, embeddings, FAISS index, and prompts from the pickle file
with open('chatbot_state_v1.pkl', 'rb') as f:
    data = pickle.load(f)

# Restore the saved components
tokenizer = data['tokenizer']
model = data['model']
embeddings = data['embeddings']
index = data['faiss_index']
prompts = data['prompts']

# Fetch the API key from the environment variable
api_key = os.getenv('COHERE_API_KEY')

# Ensure the API key is set
if api_key is None:
    raise ValueError("API Key not found. Please set COHERE_API_KEY in the environment variables.")

# Set up the Cohere client using the API key
cohere_client = cohere.Client(api_key)

# Initialize an empty list to store the conversation history
conversation_history = []

# Function to retrieve the closest dialog based on a user's query
def retrieve_closest_dialog(query, tokenizer, model, index, prompts):
    inputs = tokenizer(query, return_tensors='pt', padding=True, truncation=True)
    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
    inputs = {k: v.to(device) for k, v in inputs.items()}

    with torch.no_grad():
        outputs = model(**inputs)
        cls_embedding = outputs.last_hidden_state[:, 0, :].cpu().numpy()

    # Perform FAISS search to find the closest match
    D, I = index.search(cls_embedding, k=1)

    # Retrieve the closest matching dialog from the dataset
    closest_dialog = prompts[I[0][0]]

    return closest_dialog

# Function to handle the chatbot conversation
def chatbot(user_input):
    global conversation_history

    # Step 1: Retrieve the closest dialog
    closest_dialog = retrieve_closest_dialog(user_input, tokenizer, model, index, prompts)

    # Step 2: Append user input and retrieved dialog to conversation history
    conversation_history.append(f"User: {user_input}")
    conversation_history.append(f"Previous conversation: {closest_dialog}")

    # Step 3: Build the prompt from the conversation history
    prompt = "\n".join(conversation_history[-6:])  # Limit to last 6 exchanges
    prompt += f"\nYour response:"

    # Step 4: Generate response using Cohere
    try:
        response = cohere_client.generate(
            model='command-r-plus-04-2024',
            prompt=prompt,
            max_tokens=100
        )
        generated_response = response.generations[0].text
    except Exception as e:
        generated_response = f"Error generating response: {str(e)}"
    
    # Step 5: Append the generated response to the conversation history
    conversation_history.append(f"Bot: {generated_response}")

    # Return the generated response
    return generated_response

# Gradio Interface
def gradio_interface(user_input):
    return chatbot(user_input)

# Set up Gradio interface
interface = gr.Interface(fn=gradio_interface, inputs="text", outputs="text", title="Conversational Chatbot")

# Launch the Gradio app
interface.launch()