app.py

import streamlit as st
import requests
import json
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# Load model and tokenizer
@st.cache_resource
def load_model():
    model = AutoModelForSeq2SeqLM.from_pretrained("shreyanshjha0709/watch-description-generator")
    tokenizer = AutoTokenizer.from_pretrained("shreyanshjha0709/watch-description-generator")
    return model, tokenizer

model, tokenizer = load_model()

# Load the JSON file from a URL
@st.cache_data
def load_json_from_url(url):
    response = requests.get(url)
    return response.json()

# Provide your JSON URL here
json_url = "https://www.ethoswatches.com/feeds/holbox_ai.json"
data = load_json_from_url(json_url)

# Extract unique brands
brands = sorted(list(set([item["brand"] for item in data])))

# Streamlit UI
st.title("Watch Description Generator")

# Select brand
selected_brand = st.selectbox("Select a Brand", ["Select"] + brands)

# Filter watches and SKUs by the selected brand
if selected_brand != "Select":
    watches = [item["name"] for item in data if item["brand"] == selected_brand]
    skus = [item["sku"] for item in data if item["brand"] == selected_brand]
    selected_watch = st.selectbox("Select Watch Name (Optional)", ["Select"] + watches)
    selected_sku = st.selectbox("Select SKU (Optional)", ["Select"] + skus)

    # Get the selected watch data from the JSON
    watch_data = None
    if selected_watch != "Select":
        watch_data = next((item for item in data if item["name"] == selected_watch), None)
    elif selected_sku != "Select":
        watch_data = next((item for item in data if item["sku"] == selected_sku), None)

    if watch_data:
        # Generation parameters
        max_length = st.slider("Max Length", min_value=100, max_value=500, value=300)  # Increase max length to allow for more text
        temperature = st.slider("Temperature", min_value=0.1, max_value=1.0, value=0.7, step=0.1)
        repetition_penalty = st.slider("Repetition Penalty", min_value=1.0, max_value=2.0, value=1.2, step=0.1)  # Penalize repetitive text

        # Generate description based on attributes
        if st.button("Generate Description"):
            attributes = {
                "brand": watch_data["brand"],
                "name": watch_data.get("name", "Unknown Watch"),
                "sku": watch_data.get("sku", "Unknown SKU"),
                "features": watch_data.get("features", "Unknown Features"),
                "casesize": watch_data.get("casesize", "Unknown Case Size"),
                "movement": watch_data.get("movement", "Unknown Movement"),
                "gender": watch_data.get("gender", "Unknown Gender"),
            }

            # Format input similar to training data (adjust this based on your training data format)
            input_text = f"""Generate a detailed description for the following watch:
Brand: {attributes['brand']}
Name: {attributes['name']}
SKU: {attributes['sku']}
Features: {attributes['features']}
Case Size: {attributes['casesize']}
Movement: {attributes['movement']}
Gender: {attributes['gender']}

Description:"""

            # Tokenize input and generate description
            inputs = tokenizer(input_text, return_tensors="pt", max_length=512, truncation=True)
            outputs = model.generate(
                **inputs,
                max_length=max_length,
                num_return_sequences=1,
                temperature=temperature,
                top_k=50,
                top_p=0.95,
                do_sample=True,
                repetition_penalty=repetition_penalty  # Prevent repetitive descriptions
            )

            # Decode generated text
            description = tokenizer.decode(outputs[0], skip_special_tokens=True)

            # Display the result
            st.write("### Generated Description")
            st.write(description)

            # Display watch details
            st.write("### Watch Details")
            st.json(json.dumps(watch_data, indent=2))

            # Display input text for debugging
            st.write("### Input Text (for debugging)")
            st.text(input_text)
else:
    st.warning("Please select a brand.")

# Add some information about the app
st.sidebar.title("About")
st.sidebar.info(
    "This app uses a fine-tuned AI model to generate descriptions for watches. "
    "Select a brand and a watch to get started. The model will generate a unique "
    "description based on the watch's attributes."
)

# Add a footer
st.markdown(
    """
    <style>
    .footer {
        position: fixed;
        left: 0;
        bottom: 0;
        width: 100%;
        background-color: #f1f1f1;
        color: black;
        text-align: center;
    }
    </style>
    <div class="footer">
        <p>Developed with ❤️ by Shreyansh Jha</p>
    </div>
    """,
    unsafe_allow_html=True
)