app.py

import os
import nest_asyncio
nest_asyncio.apply()
import streamlit as st
from transformers import pipeline, AutoTokenizer
from huggingface_hub import login
from streamlit.components.v1 import html
import pandas as pd
import torch
import random
import gc
import time

# Retrieve the token from environment variables
hf_token = os.environ.get("HF_TOKEN")
if not hf_token:
    st.error("Hugging Face token not found. Please set the HF_TOKEN environment variable.")
    st.stop()

# Login with the token
login(token=hf_token)

# Timer component using HTML and JavaScript
def timer():
    return """
    <div id="timer" style="font-size:16px;color:#666;margin-bottom:10px;">⏱️ Elapsed: 00:00</div>
    <script>
    (function() {
        var start = Date.now();
        var timerElement = document.getElementById('timer');
        localStorage.removeItem("freezeTimer");
        var interval = setInterval(function() {
            if(localStorage.getItem("freezeTimer") === "true"){
                clearInterval(interval);
                timerElement.style.color = '#00cc00';
                return;
            }
            var elapsed = Date.now() - start;
            var minutes = Math.floor(elapsed / 60000);
            var seconds = Math.floor((elapsed % 60000) / 1000);
            timerElement.innerHTML = '⏱️ Elapsed: ' +
            (minutes < 10 ? '0' : '') + minutes + ':' +
            (seconds < 10 ? '0' : '') + seconds;
        }, 1000);
    })();
    </script>
    """

st.set_page_config(page_title="Review Scorer & Report Generator", page_icon="📝")
st.header("Review Scorer & Report Generator")

# Concise introduction
st.write("This model will score your reviews in your CSV file and generate a report based on your query and those results.")

# Cache the model loading functions
@st.cache_resource
def load_llama_model():
    """Load and cache the Llama 3.2 model"""
    return pipeline("text-generation", 
                    model="meta-llama/Llama-3.2-1B-Instruct",
                    device=0,  # Use GPU if available
                    torch_dtype=torch.bfloat16)  # Use FP16 for efficiency

@st.cache_resource
def load_sentiment_model():
    """Load and cache the sentiment analysis model"""
    return pipeline("text-classification", 
                    model="cardiffnlp/twitter-roberta-base-sentiment-latest", 
                    device=0 if torch.cuda.is_available() else -1)

# Load Llama 3.2 model
loading_llama_placeholder = st.empty()
loading_llama_placeholder.info("Loading Llama 3.2 summarization model...")

try:
    llama_pipe = load_llama_model()
    
    # Clear loading message
    loading_llama_placeholder.empty()
    
    # Display success message in a placeholder
    success_llama_placeholder = st.empty()
    success_llama_placeholder.success("Llama 3.2 summarization model loaded successfully!")
    
    # Use st.session_state to track when to clear the message
    if "clear_llama_success_time" not in st.session_state:
        st.session_state.clear_llama_success_time = time.time() + 5
    
    # Check if it's time to clear the message
    if time.time() > st.session_state.clear_llama_success_time:
        success_llama_placeholder.empty()
    
except Exception as e:
    # Clear loading message
    loading_llama_placeholder.empty()
    
    st.error(f"Error loading Llama 3.2 summarization model: {e}")
    st.error(f"Detailed error: {type(e).__name__}: {str(e)}")

# Load sentiment analysis model
loading_sentiment_placeholder = st.empty()
loading_sentiment_placeholder.info("Loading sentiment analysis model...")

try:
    score_pipe = load_sentiment_model()
    
    # Clear loading message
    loading_sentiment_placeholder.empty()
    
    # Display success message in a placeholder
    success_sentiment_placeholder = st.empty()
    success_sentiment_placeholder.success("Sentiment analysis model loaded successfully!")
    
    # Use st.session_state to track when to clear the message
    if "clear_sentiment_success_time" not in st.session_state:
        st.session_state.clear_sentiment_success_time = time.time() + 5
    
    # Check if it's time to clear the message
    if time.time() > st.session_state.clear_sentiment_success_time:
        success_sentiment_placeholder.empty()
    
except Exception as e:
    # Clear loading message
    loading_sentiment_placeholder.empty()
    
    st.error(f"Error loading sentiment analysis model: {e}")

def extract_assistant_content(raw_response):
    """Extract only the assistant's content from the Gemma-3 response."""
    # Convert to string and work with it directly
    response_str = str(raw_response)

    # Look for the assistant's content marker
    assistant_marker = "'role': 'assistant', 'content': '"
    if assistant_marker in response_str:
        start_idx = response_str.find(assistant_marker) + len(assistant_marker)
        # Extract everything after the marker until the end or closing quote
        content = response_str[start_idx:]

        # Find the end of the content (last single quote before the end of the string or before closing curly brace)
        end_markers = ["'}", "'}]"]
        end_idx = len(content)
        for marker in end_markers:
            pos = content.rfind(marker)
            if pos != -1 and pos < end_idx:
                end_idx = pos

        return content[:end_idx]

    # Fallback - return the original response
    return response_str

# Input: Query text for scoring and CSV file upload for candidate reviews
query_input = st.text_area("Enter your query text for analysis (this does not need to be part of the CSV):")
uploaded_file = st.file_uploader("Upload Reviews CSV File (must contain a 'reviewText' column)", type=["csv"])

if score_pipe is None or llama_pipe is None:
    st.error("Model loading failed. Please check your model names, token permissions, and GPU configuration.")
else:
    candidate_docs = []
    if uploaded_file is not None:
        try:
            df = pd.read_csv(uploaded_file)
            if 'reviewText' not in df.columns:
                st.error("CSV must contain a 'reviewText' column.")
            else:
                candidate_docs = df['reviewText'].dropna().astype(str).tolist()
        except Exception as e:
            st.error(f"Error reading CSV file: {e}")
    
    if st.button("Generate Report"):
        # Reset timer state so that the timer always shows up
        st.session_state.timer_started = False
        st.session_state.timer_frozen = False
        if uploaded_file is None:
            st.error("Please upload a CSV file.")
        elif not candidate_docs:
            st.error("CSV must contain a 'reviewText' column.")
        elif not query_input.strip():
            st.error("Please enter a query text!")
        else:
            if not st.session_state.timer_started and not st.session_state.timer_frozen:
                st.session_state.timer_started = True
                html(timer(), height=50)
            status_text = st.empty()
            progress_bar = st.progress(0)
            
            
            # Stage 1: Process, summarize (if needed), and score candidate documents
            status_text.markdown("**🔍 Processing and scoring candidate documents...**")
            
            # Process each review individually with summarization for long documents
            processed_docs = []  # Store processed (original or summarized) documents
            scored_results = []  # Store sentiment scores

            for i, doc in enumerate(candidate_docs):
                # Update progress based on current document
                progress = int((i / len(candidate_docs)) * 50)  # First half of progress bar (0-50%)
                progress_bar.progress(progress)
            
                try:
                    # Check if document exceeds the length limit for sentiment analysis
                    if len(doc) > 1500:  # Approximate limit for sentiment model
                        # Use Llama 3.2 to summarize the document
                        summary_prompt = [
                            {"role": "user", "content": f"Summarize the following text into a shorter version that preserves the sentiment and key points: {doc[:2000]}..."}
                        ]
                        
                        summary_result = llama_pipe(
                            summary_prompt, 
                            max_new_tokens=30,  # Limit summary length
                            do_sample=True,
                            temperature=0.3,  # Lower temperature for more factual summaries
                            return_full_text=False  # Return only the generated text
                        )
                        
                        # Extract the summary from the result
                        processed_doc = summary_result[0]['generated_text']
                        status_text.markdown(f"**📝 Summarized document {i+1}/{len(candidate_docs)}**")
                    else:
                        # Use the original document if it's short enough
                        processed_doc = doc
                    
                    # Store the processed document (original or summary)
                    processed_docs.append(processed_doc)
                    
                    # Process the document with sentiment analysis
                    result = score_pipe(processed_doc)
                    
                    # If it's a list, get the first element
                    if isinstance(result, list):
                        result = result[0]
                        
                    scored_results.append(result)
            
                    # Free memory
                    torch.cuda.empty_cache()
                    
                except Exception as e:
                    st.warning(f"Error processing document {i}: {str(e)}")
                    # Add a placeholder result to maintain indexing
                    processed_docs.append("Error processing this document")
                    scored_results.append({"label": "NEUTRAL", "score": 0.5})
    
                
                # Display occasional status updates for large datasets
                if i % max(1, len(candidate_docs) // 10) == 0:
                    status_text.markdown(f"**🔍 Scoring documents... ({i}/{len(candidate_docs)})**")
                    
            # Pair each review with its score assuming the output order matches the input order.
            scored_docs = list(zip(processed_docs, [result.get("score", 0.5) for result in scored_results]))
            
            progress_bar.progress(67)
            
            # Stage 2: Generate Report using Gemma in the new messages format.
            status_text.markdown("**📝 Generating report with Gemma...**")

            # After using score_pipe
            del score_pipe
            gc.collect()
            torch.cuda.empty_cache()
            
            # After using summarization_pipe
            del llama_pipe 
            gc.collect()
            torch.cuda.empty_cache()
            
            # Then reload Gemma specifically for the final step
            tokenizer = AutoTokenizer.from_pretrained("google/gemma-3-1b-it")
            gemma_pipe = pipeline("text-generation", 
                                  model="google/gemma-3-1b-it",
                                  tokenizer=tokenizer,
                                  device=0, 
                                  torch_dtype=torch.bfloat16)

            # Sample or summarize the data for Gemma to avoid memory issues
            import random
            max_reviews = 50  # Adjust based on your GPU memory
            if len(scored_docs) > max_reviews:
                sampled_docs = random.sample(scored_docs, max_reviews)
                st.info(f"Sampling {max_reviews} out of {len(scored_docs)} reviews for report generation")
            else:
                sampled_docs = scored_docs
            
            # Build the user content with query, sentiment results, and original review data.
            # Format the prompt as chat for Gemma
            messages = [
                {"role": "user", "content": f"""
Generate a concise 300-word report based on the following analysis without repeating what's in the analysis.
Query:
"{query_input}"
Candidate Reviews with their scores:
{scored_docs}
        """}
            ]
            
            raw_result = gemma_pipe(messages, max_new_tokens=150)
            report = extract_assistant_content(raw_result)
            progress_bar.progress(100)
            status_text.success("**✅ Generation complete!**")
            html("<script>localStorage.setItem('freezeTimer', 'true');</script>", height=0)
            st.session_state.timer_frozen = True
            #st.write("**Scored Candidate Reviews:**", scored_docs)
            st.write("**Generated Report:**", report)