app.py

import os
import numpy as np
import pandas as pd
import streamlit as st
from huggingface_hub import login
from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
from keybert import KeyBERT
from openai import AzureOpenAI  # new

# ─── Azure OpenAI Client ─────────────────────────────────────────────────────
openai_client = AzureOpenAI(
  api_key = "fbca46bfd8814334be46a2e5c323904c", # use your key here
  api_version = "2023-05-15", # apparently HKUST uses a deprecated version
  azure_endpoint = "https://hkust.azure-api.net" # per HKUST instructions
)

# ─── (your existing cache decorators) ────────────────────────────────────────
@st.cache_resource
def load_sentiment_pipeline():
    model_name = "mayf/amazon_reviews_bert_ft"
    tok = AutoTokenizer.from_pretrained(model_name, use_auth_token=True)
    mdl = AutoModelForSequenceClassification.from_pretrained(model_name, use_auth_token=True)
    return pipeline(
        "sentiment-analysis",
        model=mdl,
        tokenizer=tok,
        return_all_scores=True
    )

@st.cache_resource
def load_keybert_model():
    return KeyBERT(model="all-MiniLM-L6-v2")


def main():
    st.title("📊 Review Sentiment & Keyword Analyzer + GPT Insights")

    # ─── Inputs & Models ──────────────────────────────────────────────────────
    review = st.text_area("Enter your review:")
    if not st.button("Analyze Review"):
        return

    if not review:
        st.warning("Please enter a review to analyze.")
        return

    # ─── Sentiment & Keywords ─────────────────────────────────────────────────
    sentiment_pipeline = load_sentiment_pipeline()
    kw_model = load_keybert_model()

    scores = sentiment_pipeline(review)[0]
    sentiment_results = {item['label']: float(item['score']) for item in scores}
    st.subheader("Sentiment Scores")
    st.json({k: round(v, 4) for k, v in sentiment_results.items()})

    keywords = kw_model.extract_keywords(
        review,
        keyphrase_ngram_range=(1, 2),
        stop_words="english",
        top_n=3
    )
    st.subheader("Top 3 Keywords")
    for kw, score in keywords:
        st.write(f"- **{kw}** (Score: {score:.4f})")

    # ─── Determine Highest Sentiment ───────────────────────────────────────────
    max_label, max_score = max(sentiment_results.items(), key=lambda x: x[1])
    st.subheader("Highest Sentiment")
    st.write(f"{max_label} (Score: {max_score:.4f})")

    # ─── GPT-Driven Analysis & Suggestions ────────────────────────────────────
    st.subheader("GPT Analysis & Seller Suggestions")

    # build a single text prompt for GPT
    prompt = f"""
You are a helpful assistant for e-commerce sellers. 
Here is a product review, its sentiment breakdown, and the top keywords:

Review:
\"\"\"{review}\"\"\"

Sentiment scores:
{sentiment_results}

Top keywords:
{[kw for kw, _ in keywords]}

First, provide a one-paragraph professional analysis of what the customer feels and why (combine sentiment + keywords). 
Then, give 3 detailed, actionable suggestions the seller can implement to improve future reviews or address the feedback.
"""

    # call Azure OpenAI
    response = openai_client.chat.completions.create(
    model="gpt-35-turbo",                     # ← use your deployed model name here
    messages=[
        {"role": "system", "content": "You are a product-feedback analyst."},
        {"role": "user", "content": prompt}
    ],
    temperature=0.7,
    max_tokens=400
)

    # display GPT’s reply
    gpt_reply = response.choices[0].message.content
    st.markdown(gpt_reply)


if __name__ == "__main__":
    # make sure your env vars are set: AZURE_OPENAI_KEY, AZURE_OPENAI_ENDPOINT
    main()