import gradio as gr
import requests
import pandas as pd
from transformers import pipeline
import torch
import time
import threading
import os
from datetime import datetime, timezone
from huggingface_hub import HfApi
import plotly.graph_objects as go

# --- KONFIGŪRACIJA ---
MODEL_NAME = "ProsusAI/finbert"
TARGET_DATASET = "Vycka12/Base"
HF_TOKEN = os.environ.get("HF_TOKEN")

# API Raktas ir URL (Griežtai nustatyti)
CRYPTOPANIC_API_KEY = "6c0f988f9e33170ccd183c6a14b34e8c2ad0867f"
CRYPTOPANIC_URL = "https://cryptopanic.com/api/developer/v2/posts/"

# Globalūs kintamieji
news_buffer = []
stats = {"bullish": 0, "bearish": 0, "neutral": 0, "overall": "Inicijuojama...", "status": "Startuoja..."}

# Krauname AI
print("⌛ Kraunamas AI modelis...")
try:
    sentiment_pipeline = pipeline("sentiment-analysis", model=MODEL_NAME)
    print("✅ AI paruoštas.")
except Exception as e:
    print(f"❌ AI ERROR: {e}")
    sentiment_pipeline = None

class SentimentSystem:
    def __init__(self):
        self.api = HfApi(token=HF_TOKEN)

    def fetch_and_analyze(self):
        global news_buffer, stats
        stats["status"] = "Gaunamos naujienos..."
        
        # Pataisyti parametrai - tik būtiniausi
        params = {
            "auth_token": CRYPTOPANIC_API_KEY,
            "kind": "news"
        }

        try:
            resp = requests.get(CRYPTOPANIC_URL, params=params, timeout=20)
            
            if resp.status_code == 200:
                data = resp.json()
                raw_news = data.get("results", [])
                
                if not raw_news:
                    stats["status"] = "⚠️ API grąžino tuščią sąrašą."
                    return
                
                temp_news = []
                pos, neg, neut = 0, 0, 0
                
                for item in raw_news:
                    title = item.get("title", "")
                    if not title: continue
                    
                    # AI Analizė
                    if sentiment_pipeline:
                        result = sentiment_pipeline(title[:512])[0]
                        label = result['label']
                        score = result['score']
                    else:
                        label, score = 'neutral', 0.5

                    # Emocijos
                    if label == 'positive': 
                        status, emo = "🟢 BULLISH", "🚀"
                        pos += 1
                    elif label == 'negative': 
                        status, emo = "🔴 BEARISH", "📉"
                        neg += 1
                    else: 
                        status, emo = "⚪ NEUTRAL", "➖"
                        neut += 1

                    # Laikas
                    pub_time = item.get("published_at", "")[:16].replace("T", " ")
                    
                    temp_news.append([emo, status, title, f"{round(score*100)}%", pub_time])

                stats["bullish"] = pos
                stats["bearish"] = neg
                stats["neutral"] = neut
                
                total = pos + neg + neut
                if total > 0:
                    ratio = (pos - neg) / total
                    if ratio > 0.2: stats["overall"] = "OPTIMIZMAS 📈"
                    elif ratio < -0.2: stats["overall"] = "BAIMĖ 📉"
                    else: stats["overall"] = "NEUTRALU ⚖️"
                
                news_buffer = temp_news
                stats["status"] = f"Atnaujinta: {datetime.now().strftime('%H:%M:%S')}"
            else:
                stats["status"] = f"❌ API Klaida: {resp.status_code}"
        except Exception as e:
            stats["status"] = f"❌ Klaida: {str(e)}"

    def create_gauge(self):
        total = stats["bullish"] + stats["bearish"] + stats["neutral"]
        val = 50
        if total > 0:
            # Formulė: 50 + (teigiami - neigiami) * koeficientas
            # Jei visi teigiami -> 100, visi neigiami -> 0
            net_sentiment = (stats["bullish"] - stats["bearish"]) / total
            val = 50 + (net_sentiment * 50)
        
        fig = go.Figure(go.Indicator(
            mode = "gauge+number",
            value = val,
            title = {'text': f"Rinkos Emocija: {stats['overall']}"},
            gauge = {
                'axis': {'range': [0, 100]},
                'bar': {'color': "black"},
                'steps': [
                    {'range': [0, 40], 'color': "#ff4b4b"},
                    {'range': [40, 60], 'color': "#ffffb2"},
                    {'range': [60, 100], 'color': "#00cc96"}
                ],
            }
        ))
        fig.update_layout(height=300, margin=dict(l=20, r=20, t=50, b=20))
        return fig

sys_analyzer = SentimentSystem()

def update_loop():
    # Pirmas paleidimas po 10 sek
    time.sleep(10)
    sys_analyzer.fetch_and_analyze()
    
    while True:
        # Atnaujiname kas 4 valandas (taupome limitą)
        time.sleep(14400)
        sys_analyzer.fetch_and_analyze()

def get_ui_data():
    gauge = sys_analyzer.create_gauge()
    cols = ["Emoji", "Verdiktas", "Antraštė", "Pasitikėjimas", "Laikas"]
    
    if not news_buffer:
        df = pd.DataFrame([["-", "-", "Kraunama...", "-", "-"]], columns=cols)
    else:
        df = pd.DataFrame(news_buffer, columns=cols)
    
    status_text = f"### 📊 Statistika\n**Būsena:** {stats['status']}\n**Geros:** {stats['bullish']} | **Blogos:** {stats['bearish']} | **Viso:** {stats['bullish']+stats['bearish']+stats['neutral']}"
    return gauge, df, status_text

# --- GRADIO UI ---
with gr.Blocks(title="Sentiment AI", theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 🧠 Sentiment AI Analyzer")
    gr.Markdown("Analizuoja realias crypto naujienas naudodamas ProsusAI FinBERT modelį.")
    
    with gr.Row():
        with gr.Column():
            gauge_output = gr.Plot(label="Nuotaika")
        with gr.Column():
            status_output = gr.Markdown("Laukiama duomenų...")
            refresh_btn = gr.Button("🔄 Atnaujinti Dabar", variant="primary")

    gr.Markdown("### 📰 Naujienų srautas ir AI vertinimas")
    table_output = gr.Dataframe(interactive=False)

    # Eventai
    refresh_btn.click(fn=sys_analyzer.fetch_and_analyze).then(
        fn=get_ui_data, outputs=[gauge_output, table_output, status_output]
    )
    
    # Auto-start
    threading.Thread(target=update_loop, daemon=True).start()
    
    # UI atnaujinimas (tik vaizdo, ne duomenų siurbimo)
    demo.load(get_ui_data, outputs=[gauge_output, table_output, status_output])
    gr.Timer(5).tick(get_ui_data, outputs=[gauge_output, table_output, status_output])

if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)