app.py

import gradio as gr
from sentence_transformers import SentenceTransformer, util
import openai
import os

os.environ["TOKENIZERS_PARALLELISM"] = "false"

# Initialize paths and model identifiers for easy configuration and maintenance
filename = "output_topic_details.txt"  # Path to the file storing chess-specific details
retrieval_model_name = 'output/sentence-transformer-finetuned/'

openai.api_key = os.environ["OPENAI_API_KEY"]

system_message = "You are a sorting chatbot specialized in providing information about how to sort trash."
# Initial system message to set the behavior of the assistant
messages = [{"role": "system", "content": system_message}]

# Attempt to load the necessary models and provide feedback on success or failure
try:
    retrieval_model = SentenceTransformer(retrieval_model_name)
    print("Models loaded successfully.")
except Exception as e:
    print(f"Failed to load models: {e}")

def load_and_preprocess_text(filename):
    """
    Load and preprocess text from a file, removing empty lines and stripping whitespace.
    """
    try:
        with open(filename, 'r', encoding='utf-8') as file:
            segments = [line.strip() for line in file if line.strip()]
        print("Text loaded and preprocessed successfully.")
        return segments
    except Exception as e:
        print(f"Failed to load or preprocess text: {e}")
        return []

segments = load_and_preprocess_text(filename)

def find_relevant_segment(user_query, segments):
    """
    Find the most relevant text segment for a user's query using cosine similarity among sentence embeddings.
    This version finds the best match based on the content of the query.
    """
    try:
        # Lowercase the query for better matching
        lower_query = user_query.lower()
        
        # Encode the query and the segments
        query_embedding = retrieval_model.encode(lower_query)
        segment_embeddings = retrieval_model.encode(segments)
        
        # Compute cosine similarities between the query and the segments
        similarities = util.pytorch_cos_sim(query_embedding, segment_embeddings)[0]
        
        # Find the index of the most similar segment
        best_idx = similarities.argmax()
        
        # Return the most relevant segment
        return segments[best_idx]
    except Exception as e:
        print(f"Error in finding relevant segment: {e}")
        return ""

def generate_response(user_query, relevant_segment):
    """
    Generate a response emphasizing the bot's capability in providing information about how to sort your trash.
    """
    try:
        user_message = f"Here's the information on how to sort your trash: {relevant_segment}"

        # Append user's message to messages list
        messages.append({"role": "user", "content": user_message})
        
        response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=messages,
            max_tokens=300,
            temperature=0.2,
            top_p=1,
            frequency_penalty=0,
            presence_penalty=0
        )
        
        # Extract the response text
        output_text = response['choices'][0]['message']['content'].strip()
        
        # Append assistant's message to messages list for context
        messages.append({"role": "assistant", "content": output_text})
        
        return output_text
        
    except Exception as e:
        print(f"Error in generating response: {e}")
        return f"Error in generating response: {e}"

def query_model(question):
    """
    Process a question, find relevant information, and generate a response.
    """
    if question == "":
        return "Welcome to GreenGuide! Input the type of trash that you want to sort, and I will tell you which trash bin to put it in."
    relevant_segment = find_relevant_segment(question, segments)
    if not relevant_segment:
        return "Description of trash is too vague. Please refine your question and provide more details."
    response = generate_response(question, relevant_segment)
    return response





# Define the welcome message and specific topics the chatbot can provide information about
welcome_message = """
# ♻️ Welcome to GreenGuide!

## Your AI-driven assistant for all trash sorting related queries. Created by Emma, Laura, and Saahiti of the 2024 Kode With Klossy Seattle Camp. 
"""

topics = """
## Feel free to ask me anything about how to sort your trash! Please provide the material the trash is made of in your question!
- Prescription bottles
- Soft plastic
- Hard plastic
- Glass
- Furniture
- Wood
- Food scraps
- Mercury-containing products
- Aluminum and tin cans
- Metal and appliances
- Electronics
- Paper
- Yard trimmings or yard waste
- Fabric or textiles
- Batteries
- Paint
- **Mixed materials that cannot be separated**
- **Items contaminated with food**
"""
thankyou_message = """
## Thank you so much for visiting our website and learning more about how to sort your trash! According to the EPA, as much as 25% of all recycling is contaminated and cannot be recycled. With your help, we can reduce that percentage so less waste is going into our landfills. This is crucial for the preservation of our environment as the amount of space on this planet is limited, so one day we will run out of places to store our trash! Once again, we are so grateful that you are  helping the planet become a better place by learning about how to recycle and compost correctly 💚🌲!
"""           
##              
# Setup the Gradio Blocks interface with custom layout components
with gr.Blocks(theme='JohnSmith9982/small_and_pretty') as demo:
    gr.Image("GreenGuideLogo2.png", show_label = False, show_share_button = False, show_download_button = False, height = 500) 
    gr.Markdown(welcome_message)  # Display the formatted welcome message
    with gr.Row():
        with gr.Column():
            gr.Markdown(topics)  # Show the topics on the left side
    with gr.Row():
        with gr.Column():
            question = gr.Textbox(label="Your question", placeholder="Which type of trash do you want to ask about?")
            submit_button = gr.Button("Submit")
            answer = gr.Textbox(label="GreenGuide Response", placeholder="GreenGuide will respond here...", interactive=False, lines=10)
            submit_button.click(fn=query_model, inputs=question, outputs=answer)
    gr.Markdown(thankyou_message)

# Launch the Gradio app to allow user interaction
demo.launch(share=True)