import streamlit as st
from transformers import pipeline
import pandas as pd
from datetime import datetime
import time
import plotly.graph_objects as go
import json
import os
import gspread
from google.oauth2.service_account import Credentials
 
# --- PAGE CONFIG ---
st.set_page_config(page_title="Sentiment Analyzer AI | Bilingual Engine", page_icon="🌐", layout="wide")
 
# --- PROFESSIONAL NEUMORPHIC / GLASS CSS ---
st.markdown("""
    <style>
    .stApp { background: linear-gradient(135deg, #12141d 0%, #1a1c2c 100%); color: #ffffff; }
    div[data-baseweb="input"] { background: rgba(255, 255, 255, 0.05) !important; backdrop-filter: blur(10px) !important; border-radius: 15px !important; border: 1px solid rgba(255, 255, 255, 0.1) !important; padding: 5px !important; }
    .glass-card { background: rgba(255, 255, 255, 0.05); backdrop-filter: blur(10px); border-radius: 20px; border: 1px solid rgba(255, 255, 255, 0.1); padding: 30px; margin-top: 20px; margin-bottom: 25px; transition: 0.4s ease; }
    .stButton>button { background: linear-gradient(90deg, #4facfe 0%, #00f2fe 100%); color: white; border: none; border-radius: 12px; font-weight: 600; letter-spacing: 0.5px; height: 3rem; transition: all 0.3s ease; }
    .stButton>button:hover { box-shadow: 0 0 20px rgba(79, 172, 254, 0.4); transform: scale(1.02); }
    [data-testid="stMetricValue"] { color: #00f2fe; font-weight: 800; }
    </style>
    """, unsafe_allow_html=True)
 
# --- GOOGLE SHEETS CONNECTION (gspread - works in Docker) ---
def get_connection():
    try:
        # Docker Spaces inject secrets as ENV VARS
        json_secrets = os.environ.get("GSHEETS_JSON")
        sheet_url = os.environ.get("GSHEETS_URL")
 
        # Fallback for local development
        if not json_secrets:
            json_secrets = st.secrets.get("GSHEETS_JSON")
        if not sheet_url:
            sheet_url = st.secrets.get("GSHEETS_URL")
 
        if not json_secrets or not sheet_url:
            st.error("❌ Secrets not found. Please add GSHEETS_JSON and GSHEETS_URL in Space Settings → Secrets.")
            st.stop()
 
        creds_dict = json.loads(json_secrets)
        scopes = [
            "https://spreadsheets.google.com/feeds",
            "https://www.googleapis.com/auth/drive"
        ]
        creds = Credentials.from_service_account_info(creds_dict, scopes=scopes)
        client = gspread.authorize(creds)
        sheet = client.open_by_url(sheet_url).worksheet("Sheet1")
        return sheet, sheet_url
 
    except json.JSONDecodeError:
        st.error("❌ GSHEETS_JSON is not valid JSON. Please re-paste your service account key.")
        st.stop()
    except Exception as e:
        st.error(f"❌ Connection Failed: {e}")
        st.stop()
 
# --- INITIALIZATION ---
if 'conn' not in st.session_state or 'url' not in st.session_state:
    conn, GSHEETS_URL = get_connection()
    st.session_state.conn = conn
    st.session_state.url = GSHEETS_URL
else:
    conn = st.session_state.conn
    GSHEETS_URL = st.session_state.url
 
# --- SAVE TO GOOGLE SHEETS ---
def save_to_cloud(text, ai_label, ai_score, corrected_label=None):
    try:
        sheet = st.session_state.conn
        new_row = [
            datetime.now().strftime("%Y-%m-%d %H:%M:%S"),
            text,
            ai_label,
            f"{ai_score:.2%}",
            corrected_label if corrected_label else "N/A"
        ]
        sheet.append_row(new_row)
        return True
    except Exception as e:
        st.error(f"Cloud Save Failed: {e}")
        return False
 
# --- MODEL ENGINE ---
MODEL_PATH = "SumedhGajbhiye/Sentiment-Analyzer"
 
@st.cache_resource
def load_engine(path):
    return pipeline("sentiment-analysis", model=path, tokenizer=path)
 
# --- UI LAYOUT ---
col_h1, col_h2 = st.columns([3, 1])
with col_h1:
    st.title("Sentiment Analyzer")
    st.caption("Advanced Bilingual Sentiment Analysis for English, Hindi & Hinglish")
 
# --- SIDEBAR STATS ---
with st.sidebar:
    st.markdown("### 🛠️ ENGINE STATUS")
    try:
        sheet = st.session_state.conn
        all_rows = sheet.get_all_records()
        df_log = pd.DataFrame(all_rows)
        st.metric("Total Ingested", len(df_log))
        st.divider()
        st.download_button("📤 Export Dataset", df_log.to_csv(index=False), "engine_feedback.csv", "text/csv")
    except Exception:
        df_log = pd.DataFrame()
        st.info("Engine is connecting to cloud...")
 
# --- MAIN LOGIC ---
classifier = load_engine(MODEL_PATH)
 
if classifier:
    user_input = st.text_input("QUERY INPUT:", placeholder="Enter sentence...", key="main_input", label_visibility="collapsed")
 
    if user_input:
        with st.status("Initializing Neural Weights...", expanded=False) as status:
            time.sleep(0.4)
            result = classifier(user_input)[0]
            status.update(label="Analysis Complete", state="complete", expanded=False)
 
        label = result['label']
        score = result['score']
 
        emoji_map = {"Positive": "🟢", "Neutral": "🟡", "Negative": "🔴"}
        color = "#00ff88" if "POS" in label.upper() else "#ff4b4b" if "NEG" in label.upper() else "#ffaa00"
 
        st.markdown(f'''
            <div class="glass-card">
                <h4 style="color: #888; margin:0;">CLASSIFICATION RESULT</h4>
                <h1 style="color: {color}; margin:0; font-size: 3.5rem;">{label} {emoji_map.get(label, "")}</h1>
                <p style="color: #aaa; margin-top: 10px;">Deep linguistic scan detected {label.lower()} intent with {score:.1%} confidence.</p>
            </div>
            ''', unsafe_allow_html=True)
 
        col_chart, col_feed = st.columns([1, 2])
        with col_chart:
            fig = go.Figure(go.Indicator(
                mode="gauge+number", value=score * 100,
                gauge={'axis': {'range': [None, 100]}, 'bar': {'color': color}, 'bgcolor': "rgba(0,0,0,0)"}
            ))
            fig.update_layout(height=280, margin=dict(t=50, b=50, l=40, r=40), paper_bgcolor='rgba(0,0,0,0)', font={'color': "#fff"})
            st.plotly_chart(fig, use_container_width=True)
 
        with col_feed:
            st.markdown("### ⚖️ HUMAN VERIFICATION")
            c1, c2 = st.columns(2)
            with c1:
                if st.button("CONFIRM ACCURACY"):
                    if save_to_cloud(user_input, label, score):
                        st.toast("✅ Logic logged to cloud database.")
                        time.sleep(1.0)
                        st.rerun()
            with c2:
                correction = st.selectbox("OVERRIDE LABEL:", ["Positive", "Neutral", "Negative"])
                if st.button("FORCE UPDATE ENGINE"):
                    if save_to_cloud(user_input, label, score, corrected_label=correction):
                        st.toast(f"✅ Engine forced to {correction}")
                        time.sleep(1.0)
                        st.rerun()
 
# --- RECENT LOGS ---
try:
    if not df_log.empty:
        with st.expander("📂 VIEW SYSTEM LOGS"):
            st.dataframe(df_log.tail(10), use_container_width=True)
except Exception:
    pass