Code Vault — 7 Công Cụ Mã Nguồn Mở Cho Radar và Giao Dịch Crypto

#!/usr/bin/env python3
"""
Vitalik Sell Radar — Event-Driven Edition
WebSocket subscription to ERC-20 Transfer events, real-time detection of
Vitalik's sell activity, instant push to Telegram.

Architecture:
1. WebSocket subscribes to Transfer(from=vitalik) events → sub-second detection
2. Classifies sell behavior (transfers to DEX Router / CEX / LP Pool)
3. Queries token info + price via DexScreener
4. Pushes alert to Telegram
5. Auto-reconnect + multi-RPC failover
"""

import asyncio
import json
import logging
import os
import signal
import sys
import time
from datetime import datetime, timezone
from pathlib import Path

import aiohttp
import websockets

# Load .env file
def load_env():
    env_path = Path(__file__).parent / ".env"
    if env_path.exists():
        for line in env_path.read_text().splitlines():
            line = line.strip()
            if line and not line.startswith("#") and "=" in line:
                k, v = line.split("=", 1)
                os.environ.setdefault(k.strip(), v.strip())

load_env()

# ============================================================
# Configuration
# ============================================================

# Vitalik's main wallet
VITALIK_ADDRESS = "0xd8dA6BF26964aF9D7eEd9e03E53415D37aA96045"
VITALIK_PADDED = "0x" + VITALIK_ADDRESS[2:].lower().zfill(64)

# ERC-20 Transfer event topic
TRANSFER_TOPIC = "0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef"

# WebSocket RPC endpoints (free, support eth_subscribe)
WS_ENDPOINTS = [
    "wss://ethereum-rpc.publicnode.com",
    "wss://eth.drpc.org",
    "wss://ethereum.publicnode.com",
]

# HTTP RPC for querying token info
HTTP_RPC = os.environ.get("HTTP_RPC", "https://eth.drpc.org")

# Telegram
TG_BOT_TOKEN = os.environ.get("TG_BOT_TOKEN", "")
TG_CHAT_ID = os.environ.get("TG_CHAT_ID", "")

# Known DEX Router addresses (sell destinations)
KNOWN_DEX_ROUTERS = {
    # Uniswap
    "0x7a250d5630b4cf539739df2c5dacb4c659f2488d": "Uniswap V2 Router",
    "0xe592427a0aece92de3edee1f18e0157c05861564": "Uniswap V3 Router",
    "0x68b3465833fb72a70ecdf485e0e4c7bd8665fc45": "Uniswap V3 Router2",
    "0x3fc91a3afd70395cd496c647d5a6cc9d4b2b7fad": "Uniswap Universal Router",
    "0xef1c6e67703c7bd7107eed8303fbe6ec2554bf6b": "Uniswap Universal Router (old)",
    # 1inch
    "0x1111111254eeb25477b68fb85ed929f73a960582": "1inch V5",
    "0x111111125421ca6dc452d289314280a0f8842a65": "1inch V6",
    # SushiSwap
    "0xd9e1ce17f2641f24ae83637ab66a2cca9c378b9f": "SushiSwap Router",
    # CoW Protocol
    "0x9008d19f58aabd9ed0d60971565aa8510560ab41": "CoW Settlement",
    # 0x
    "0xdef1c0ded9bec7f1a1670819833240f027b25eff": "0x Exchange Proxy",
    # Curve
    "0x99a58482bd75cbab83b27ec03ca68ff489b5788f": "Curve Router",
}

# Known CEX hot wallets (partial list)
KNOWN_CEX = {
    "0x28c6c06298d514db089934071355e5743bf21d60": "Binance Hot Wallet",
    "0x21a31ee1afc51d94c2efccaa2092ad1028285549": "Binance Hot Wallet 2",
    "0xdfd5293d8e347dfe59e90efd55b2956a1343963d": "Binance Hot Wallet 3",
    "0x56eddb7aa87536c09ccc2793473599fd21a8b17f": "Binance Hot Wallet 4",
    "0x71660c4005ba85c37ccec55d0c4493e66fe775d3": "Coinbase",
    "0xa9d1e08c7793af67e9d92fe308d5697fb81d3e43": "Coinbase 10",
    "0x503828976d22510aad0201ac7ec88293211d23da": "Coinbase 2",
    "0x2faf487a4414fe77e2327f0bf4ae2a264a776ad2": "FTX (defunct)",
    "0x267be1c1d684f78cb4f6a176c4911b741e4ffdc0": "Kraken",
    "0xae2d4617c862309a3d75a0ffb358c7a5009c673f": "Kraken 10",
}

# Minimum notification amount (USD), 0 = notify all
MIN_NOTIFY_USD = int(os.environ.get("MIN_NOTIFY_USD", "0"))

# Reconnect settings
RECONNECT_DELAY = 5
MAX_RECONNECT_DELAY = 60

# ============================================================
# Logging
# ============================================================

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(message)s",
    datefmt="%Y-%m-%d %H:%M:%S",
)
log = logging.getLogger("vitalik-radar")

# ============================================================
# Global state
# ============================================================

# Dedup set for processed tx hashes (last 1000)
seen_txs: set = set()
seen_txs_list: list = []

# HTTP session
http_session: aiohttp.ClientSession | None = None

# Token info cache: {address: {symbol, name, decimals}}
token_cache: dict = {}

# Running flag
running = True


# ============================================================
# Utility functions
# ============================================================

def shorten_addr(addr: str) -> str:
    """Shorten address for display"""
    return f"{addr[:6]}...{addr[-4:]}"


def decode_transfer_value(data_hex: str, decimals: int) -> float:
    """Decode Transfer event value"""
    try:
        raw = int(data_hex, 16)
        return raw / (10 ** decimals)
    except:
        return 0.0


def topic_to_address(topic: str) -> str:
    """Extract 20-byte address from 32-byte topic"""
    return "0x" + topic[-40:]


def classify_recipient(to_addr: str) -> tuple[str, str]:
    """
    Classify recipient address.
    Returns (type, name)
    Type: "dex" / "cex" / "pool" / "unknown"
    """
    to_lower = to_addr.lower()

    if to_lower in KNOWN_DEX_ROUTERS:
        return "dex", KNOWN_DEX_ROUTERS[to_lower]

    if to_lower in KNOWN_CEX:
        return "cex", KNOWN_CEX[to_lower]

    return "unknown", ""


async def get_http_session() -> aiohttp.ClientSession:
    global http_session
    if http_session is None or http_session.closed:
        http_session = aiohttp.ClientSession()
    return http_session


async def rpc_call(method: str, params: list) -> dict | None:
    """HTTP JSON-RPC call"""
    session = await get_http_session()
    try:
        async with session.post(
            HTTP_RPC,
            json={"jsonrpc": "2.0", "id": 1, "method": method, "params": params},
            timeout=aiohttp.ClientTimeout(total=10),
        ) as resp:
            data = await resp.json()
            return data.get("result")
    except Exception as e:
        log.error(f"RPC call {method} failed: {e}")
        return None


async def get_token_info(token_addr: str) -> dict:
    """Query ERC-20 token info (symbol, name, decimals)"""
    addr_lower = token_addr.lower()
    if addr_lower in token_cache:
        return token_cache[addr_lower]

    info = {"symbol": "???", "name": "Unknown", "decimals": 18, "address": token_addr}

    # symbol()
    result = await rpc_call("eth_call", [
        {"to": token_addr, "data": "0x95d89b41"}, "latest"
    ])
    if result and len(result) > 2:
        try:
            hex_str = result[2:]
            if len(hex_str) >= 128:
                offset = int(hex_str[:64], 16) * 2
                length = int(hex_str[offset:offset+64], 16)
                symbol_hex = hex_str[offset+64:offset+64+length*2]
                info["symbol"] = bytes.fromhex(symbol_hex).decode("utf-8", errors="replace").strip('\x00')
            elif len(hex_str) == 64:
                info["symbol"] = bytes.fromhex(hex_str).decode("utf-8", errors="replace").strip('\x00')
        except:
            pass

    # decimals()
    result = await rpc_call("eth_call", [
        {"to": token_addr, "data": "0x313ce567"}, "latest"
    ])
    if result and len(result) > 2:
        try:
            info["decimals"] = int(result, 16)
        except:
            pass

    # name()
    result = await rpc_call("eth_call", [
        {"to": token_addr, "data": "0x06fdde03"}, "latest"
    ])
    if result and len(result) > 2:
        try:
            hex_str = result[2:]
            if len(hex_str) >= 128:
                offset = int(hex_str[:64], 16) * 2
                length = int(hex_str[offset:offset+64], 16)
                name_hex = hex_str[offset+64:offset+64+length*2]
                info["name"] = bytes.fromhex(name_hex).decode("utf-8", errors="replace").strip('\x00')
            elif len(hex_str) == 64:
                info["name"] = bytes.fromhex(hex_str).decode("utf-8", errors="replace").strip('\x00')
        except:
            pass

    token_cache[addr_lower] = info
    return info


async def check_if_pool(addr: str) -> bool:
    """Check if address is a Uniswap V2/V3 pool (has token0 method)"""
    result = await rpc_call("eth_call", [
        {"to": addr, "data": "0x0dfe1681"}, "latest"  # token0()
    ])
    if result and len(result) == 66:  # 0x + 64 hex chars
        return True
    return False


async def get_eth_price() -> float:
    """Get ETH price from CoinGecko"""
    session = await get_http_session()
    try:
        async with session.get(
            "https://api.coingecko.com/api/v3/simple/price?ids=ethereum&vs_currencies=usd",
            timeout=aiohttp.ClientTimeout(total=5),
        ) as resp:
            data = await resp.json()
            return data.get("ethereum", {}).get("usd", 0)
    except:
        return 2300  # fallback


async def get_token_price_usd(token_addr: str) -> float | None:
    """Get token USD price from DexScreener (free, no API key needed)"""
    session = await get_http_session()
    try:
        async with session.get(
            f"https://api.dexscreener.com/latest/dex/tokens/{token_addr}",
            timeout=aiohttp.ClientTimeout(total=5),
        ) as resp:
            data = await resp.json()
            pairs = data.get("pairs", [])
            if pairs:
                # Pick pair with highest liquidity
                pairs.sort(key=lambda p: float(p.get("liquidity", {}).get("usd", 0) or 0), reverse=True)
                price = float(pairs[0].get("priceUsd", 0) or 0)
                return price if price > 0 else None
    except:
        pass
    return None


# ============================================================
# Telegram notifications
# ============================================================

async def send_telegram(text: str):
    """Send Telegram message"""
    if not TG_BOT_TOKEN:
        log.warning("[TG] No bot token configured, skipping notification")
        return

    session = await get_http_session()
    url = f"https://api.telegram.org/bot{TG_BOT_TOKEN}/sendMessage"
    payload = {
        "chat_id": TG_CHAT_ID,
        "text": text,
        "parse_mode": "HTML",
        "disable_web_page_preview": True,
    }

    try:
        async with session.post(url, json=payload, timeout=aiohttp.ClientTimeout(total=10)) as resp:
            result = await resp.json()
            if not result.get("ok"):
                log.error(f"[TG] Send failed: {result.get('description', '')}")
                # Retry with plain text if HTML parse fails
                if "parse" in result.get("description", "").lower():
                    payload["parse_mode"] = None
                    async with session.post(url, json=payload) as resp2:
                        pass
            else:
                log.info("[TG] Message sent")
    except Exception as e:
        log.error(f"[TG] Error: {e}")


# ============================================================
# Event handling
# ============================================================

async def handle_transfer_event(log_entry: dict):
    """Process a single Transfer event"""
    tx_hash = log_entry.get("transactionHash", "")
    token_addr = log_entry.get("address", "")
    topics = log_entry.get("topics", [])
    data = log_entry.get("data", "0x0")

    # Dedup
    dedup_key = f"{tx_hash}:{token_addr}"
    if dedup_key in seen_txs:
        return
    seen_txs.add(dedup_key)
    seen_txs_list.append(dedup_key)
    # Cleanup (keep last 1000)
    while len(seen_txs_list) > 1000:
        old = seen_txs_list.pop(0)
        seen_txs.discard(old)

    # Parse topics: [Transfer, from, to]
    if len(topics) < 3:
        return

    from_addr = topic_to_address(topics[1])
    to_addr = topic_to_address(topics[2])

    # Confirm it's from Vitalik
    if from_addr.lower() != VITALIK_ADDRESS.lower():
        return

    # Get token info
    token_info = await get_token_info(token_addr)
    symbol = token_info["symbol"]
    decimals = token_info["decimals"]
    amount = decode_transfer_value(data, decimals)

    if amount == 0:
        return

    # Classify recipient
    recipient_type, recipient_name = classify_recipient(to_addr)

    # If unknown, check if it's an LP pool
    is_pool = False
    if recipient_type == "unknown":
        is_pool = await check_if_pool(to_addr)
        if is_pool:
            recipient_type = "pool"
            recipient_name = "LP Pool"

    # Determine if this is a "sell" action
    is_sell = recipient_type in ("dex", "cex", "pool")

    # If not a sell, just a regular transfer — log silently
    if not is_sell:
        log.info(f"[Transfer] {symbol} {amount:,.2f} → {shorten_addr(to_addr)} (regular transfer, not notifying)")
        return

    # Get price
    token_price = await get_token_price_usd(token_addr)
    usd_value = amount * token_price if token_price else None

    # Below minimum notification amount — skip
    if usd_value is not None and usd_value < MIN_NOTIFY_USD:
        log.info(f"[Sell] {symbol} ${usd_value:.0f} < ${MIN_NOTIFY_USD} minimum, skipping")
        return

    # Build notification message
    timestamp = datetime.now(timezone.utc).strftime("%H:%M:%S UTC")

    sell_type_label = {
        "dex": "🔄 DEX Sell",
        "cex": "🏦 CEX Transfer",
        "pool": "🌊 Pool Sell",
    }

    msg_lines = [
        f"<b>🚨 Vitalik Sell Signal</b>",
        f"",
        f"Token: <b>{symbol}</b>",
        f"Amount: {amount:,.4f}",
    ]

    if usd_value is not None:
        msg_lines.append(f"Value: <b>${usd_value:,.0f}</b>")

    if token_price is not None:
        msg_lines.append(f"Price: ${token_price:,.8f}")

    msg_lines.extend([
        f"",
        f"Type: {sell_type_label.get(recipient_type, 'Unknown')}",
        f"Destination: {recipient_name or shorten_addr(to_addr)}",
        f"Time: {timestamp}",
        f"",
        f"TX: https://etherscan.io/tx/{tx_hash}",
        f"Token: https://dexscreener.com/ethereum/{token_addr}",
    ])

    msg = "\n".join(msg_lines)
    log.info(f"[SELL DETECTED] {symbol} {amount:,.4f} → {recipient_name or to_addr} | ${usd_value or '?'}")
    await send_telegram(msg)


# ============================================================
# WebSocket listener main loop
# ============================================================

async def subscribe_and_listen(ws_url: str):
    """Connect to WebSocket and subscribe to Vitalik Transfer events"""
    log.info(f"Connecting to {ws_url}...")

    async with websockets.connect(
        ws_url,
        ping_interval=20,
        ping_timeout=30,
        close_timeout=10,
        max_size=2**20,  # 1MB
    ) as ws:
        # Subscribe to Transfer FROM Vitalik
        sub_from = {
            "jsonrpc": "2.0", "id": 1,
            "method": "eth_subscribe",
            "params": ["logs", {
                "topics": [TRANSFER_TOPIC, VITALIK_PADDED]
            }]
        }
        await ws.send(json.dumps(sub_from))
        resp = await asyncio.wait_for(ws.recv(), timeout=10)
        data = json.loads(resp)
        if "error" in data:
            raise Exception(f"Subscribe failed: {data['error']}")
        sub_id_from = data.get("result", "")
        log.info(f"✅ Subscribed to Transfer FROM Vitalik (id={sub_id_from[:10]}...)")

        # Also subscribe to Transfer TO Vitalik (monitor buys/receives)
        sub_to = {
            "jsonrpc": "2.0", "id": 2,
            "method": "eth_subscribe",
            "params": ["logs", {
                "topics": [TRANSFER_TOPIC, None, VITALIK_PADDED]
            }]
        }
        await ws.send(json.dumps(sub_to))
        resp2 = await asyncio.wait_for(ws.recv(), timeout=10)
        data2 = json.loads(resp2)
        sub_id_to = data2.get("result", "")
        if sub_id_to:
            log.info(f"✅ Subscribed to Transfer TO Vitalik (id={sub_id_to[:10]}...)")

        log.info("🔍 Monitoring Vitalik wallet... waiting for events")

        # Listen for events
        async for message in ws:
            if not running:
                break

            try:
                evt = json.loads(message)

                if "params" not in evt:
                    continue

                sub_id = evt["params"].get("subscription", "")
                log_entry = evt["params"].get("result", {})

                if sub_id == sub_id_from:
                    # Vitalik sent tokens → check if it's a sell
                    await handle_transfer_event(log_entry)
                elif sub_id == sub_id_to:
                    # Vitalik received tokens — log silently
                    token_addr = log_entry.get("address", "")
                    token_info = await get_token_info(token_addr)
                    data_hex = log_entry.get("data", "0x0")
                    amount = decode_transfer_value(data_hex, token_info["decimals"])
                    log.debug(f"[Receive] {token_info['symbol']} +{amount:,.4f}")

            except json.JSONDecodeError:
                continue
            except Exception as e:
                log.error(f"Event handling error: {e}", exc_info=True)


async def main():
    """Main loop with auto-reconnect"""
    global running

    # Validate required config
    if not TG_BOT_TOKEN:
        log.warning("TG_BOT_TOKEN not set — Telegram notifications disabled")
    if not TG_CHAT_ID:
        log.warning("TG_CHAT_ID not set — Telegram notifications disabled")

    # Signal handling
    def shutdown(sig, frame):
        global running
        log.info(f"Received signal {sig}, shutting down...")
        running = False

    signal.signal(signal.SIGINT, shutdown)
    signal.signal(signal.SIGTERM, shutdown)

    # Startup notice
    log.info("=" * 50)
    log.info("Vitalik Sell Radar — Started")
    log.info(f"Monitoring: {VITALIK_ADDRESS}")
    log.info(f"Min notify: ${MIN_NOTIFY_USD}")
    log.info(f"Telegram: {'✅ Configured' if TG_BOT_TOKEN else '❌ Not configured'}")
    log.info("=" * 50)

    if TG_BOT_TOKEN and TG_CHAT_ID:
        await send_telegram(
            "🟢 Vitalik Sell Radar — Started\n\n"
            f"Monitoring: {shorten_addr(VITALIK_ADDRESS)}\n"
            f"Min notify: ${MIN_NOTIFY_USD}\n"
            "Mode: WebSocket event-driven (sub-second latency)"
        )

    endpoint_idx = 0
    reconnect_delay = RECONNECT_DELAY

    while running:
        ws_url = WS_ENDPOINTS[endpoint_idx % len(WS_ENDPOINTS)]
        try:
            await subscribe_and_listen(ws_url)
            reconnect_delay = RECONNECT_DELAY  # reset on success
        except websockets.exceptions.ConnectionClosed as e:
            log.warning(f"WebSocket closed: {e}")
        except asyncio.TimeoutError:
            log.warning("WebSocket timeout")
        except Exception as e:
            log.error(f"WebSocket error: {e}")

        if not running:
            break

        # Switch endpoint and retry
        endpoint_idx += 1
        log.info(f"Reconnecting in {reconnect_delay}s... (next: {WS_ENDPOINTS[endpoint_idx % len(WS_ENDPOINTS)]})")
        await asyncio.sleep(reconnect_delay)
        reconnect_delay = min(reconnect_delay * 1.5, MAX_RECONNECT_DELAY)

    # Cleanup
    if http_session and not http_session.closed:
        await http_session.close()

    log.info("Vitalik Sell Radar — Stopped")


if __name__ == "__main__":
    asyncio.run(main())

#!/usr/bin/env python3
"""
叙事雷达 → 链上雷达 v1
纯Python，零AI成本（关键词匹配 + 叙事去重）

三条推送通道：
1. 全新叙事 — 从未见过的概念/故事，全链推
2. 马斯克/川普相关 — 重点ETH+SOL，BSC也推
3. 币安/CZ相关 — 只推BSC

数据源：GMGN新币 + DEXScreener搜索
叙事历史：SQLite去重
"""

import requests
import json
import time
import os
import re
import sqlite3
import hashlib
from datetime import datetime, timedelta
from pathlib import Path
from difflib import SequenceMatcher

# === 配置 ===
DATA_DIR = os.path.expanduser("~/crypto-trading")
DB_FILE = os.path.join(DATA_DIR, "narrative_history.db")
LOG_FILE = os.path.join(DATA_DIR, "narrative_radar.log")
SEEN_FILE = os.path.join(DATA_DIR, "narrative_seen.json")
FLAP_SEEN_FILE = os.path.join(DATA_DIR, "flap_seen.json")

# 扫描间隔
SCAN_INTERVAL = 30  # 30秒（GMGN数据约1-5分钟刷新一次，10秒太频繁且数据不变）

# 动量追踪器 — 内存中记录每个币的价格/市值快照
# {address: [{'ts': timestamp, 'mc': market_cap, 'vol': volume, 'price': price}, ...]}
MOMENTUM_TRACKER = {}
MOMENTUM_PUSHED = {}  # {address: {'count': N, 'last_ts': ts, 'last_mc': mc}} 推送计数
MOMENTUM_CONSECUTIVE_UP = 3  # 连续涨3轮（数据实际变化时才算一轮）

# 从.env读取TG配置
def load_env():
    env = {}
    env_file = os.path.expanduser("~/.env")
    if os.path.exists(env_file):
        with open(env_file) as f:
            for line in f:
                line = line.strip()
                if '=' in line and not line.startswith('#'):
                    k, v = line.split('=', 1)
                    env[k] = v
    return env

ENV = load_env()
TG_TOKEN = ENV.get('TELEGRAM_BOT_TOKEN', '')
TG_CHAT_ID = int(os.environ.get('TG_CHAT_ID', '0'))

GMGN_HEADERS = {
    'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36',
    'Accept': 'application/json',
    'Referer': 'https://gmgn.ai/',
}

# ============================================================
# 马斯克/川普关键词库（大小写不敏感）
# ============================================================
MUSK_TRUMP_KEYWORDS = {
    # 马斯克核心
    'musk', 'elon', 'elonmusk',
    # SpaceX/Tesla/X
    'spacex', 'starship', 'tesla', 'cybertruck', 'roadster',
    'neuralink', 'boring', 'hyperloop', 'xai', 'grok',
    # 马斯克相关人物/宠物/梗
    'floki', 'shiba',  # 只在新币上下文中用
    'doge father', 'dogefather', 'technoking',
    'mars colony', 'mars',
    # 川普核心
    'trump', 'donald', 'maga', 'potus', 'trump47',
    'melania', 'barron', 'ivanka',
    # 川普相关
    'dark maga', 'darkmaga', 'ultra maga', 'save america',
    'truth social', 'covfefe',
    # 马斯克+川普联动
    'doge department', 'd.o.g.e', 'government efficiency',
}

# 马斯克/川普正则（捕捉变体）
MUSK_TRUMP_PATTERNS = [
    r'\belon\b', r'\bmusk\b', r'\btrump\b', r'\bmaga\b',
    r'\bspacex\b', r'\bstarship\b', r'\btesla\b', r'\bgrok\b',
    r'\bmelania\b', r'\bbarron\b', r'\bdoge\s*department\b',
    r'\bd\.?o\.?g\.?e\b',  # D.O.G.E变体
    r'\bx\s*ai\b', r'\bneuralink\b',
]

# ============================================================
# 币安/CZ关键词库
# ============================================================
BINANCE_CZ_KEYWORDS = {
    # CZ核心
    'cz', 'changpeng', 'zhao', 'czb', 'czbinance',
    # 何一（BSC现在的核心推手！）
    'heyi', 'yi he', 'he yi', '何一', 'yihe',
    'sister yi', 'yi jie', '一姐', '何一姐',
    # 币安品牌
    'binance', 'bnb', 'pancake', 'pancakeswap',
    # CZ相关动态词（书、活动、推特高频词）
    'giggle academy', 'binance life', 'bnb chain',
    'principles', 'cz book',
    # YZi Labs (原Binance Labs)
    'yzi', 'yzi labs',
    # 中文关键词（BSC上常见）
    '赵长鹏', '币安', '长鹏', 'cz的', '何一的',
    # Four.meme平台相关
    'fourmeme', 'four meme', '4meme',
    # CZ/何一推特互动高频词
    'czs dog', 'cz dog', 'bnb dog',
    'build on bnb', 'bnb ecosystem',
}

BINANCE_CZ_PATTERNS = [
    r'\bcz\b', r'\bbinance\b', r'\bbnb\b',
    r'\bheyi\b', r'\byi\s*he\b', r'\bhe\s*yi\b',
    r'\b何一\b', r'\b一姐\b',
    r'\bpancake\b', r'\bgiggle\b', r'\byzi\b',
    r'\bfourmeme\b', r'\b4meme\b',
]

# ============================================================
# 推特热点/名人关键词库（★★级别）
# ============================================================
CELEBRITY_VIRAL_KEYWORDS = {
    # 科技名人
    'vitalik', 'buterin', 'sam altman', 'satoshi',
    'michael saylor', 'saylor', 'cathie wood',
    'jack dorsey', 'zuckerberg', 'bezos',
    'jensen huang', 'nvidia', 'tim cook',
    # 币圈名人
    'justin sun', 'sun yuchen', '孙宇晨', 'tron',
    'arthur hayes', 'su zhu', '3ac',
    'brian armstrong', 'coinbase',
    'larry fink', 'blackrock',
    'gary gensler', 'sec',
    'michael novogratz', 'galaxy',
    # 政治/社会名人
    'biden', 'obama', 'putin', 'xi jinping',
    'kanye', 'drake', 'snoop dogg', 'paris hilton',
    'mark cuban', 'mr beast', 'mrbeast',
    # 病毒式传播热词（龙虾级别的梗）
    'lobster', '龙虾', 'lobsta',
    'hawk tuah', 'griddy', 'skibidi',
    'rizz', 'sigma', 'gyatt',
    # 重大事件关键词
    'etf', 'halving', '减半',
    'world war', 'wwiii',
    'fed', 'rate cut', '降息',
    'tiktok ban', 'tiktok',
}

CELEBRITY_VIRAL_PATTERNS = [
    r'\bvitalik\b', r'\bsaylor\b', r'\bblackrock\b',
    r'\bcoinbase\b', r'\bjustin\s*sun\b', r'\blobster\b',
    r'\betf\b', r'\bhalving\b', r'\bmrbeast\b',
    r'\bsnoop\b', r'\bkanye\b', r'\bdrake\b',
]

# ============================================================
# 通用垃圾词（过滤明显的骗局/低质量币）
# ============================================================
SPAM_PATTERNS = [
    r'airdrop', r'presale', r'pre\s*sale',
    r'1000x', r'100x guaranteed',
    r'safe\s*moon', r'baby\s*\w+',  # babydoge等仿盘
    r'pornhub', r'porn', r'xxx', r'nsfw',
    r'nigga', r'nigger', r'faggot',
    r'scam', r'rugpull', r'rug\s*pull',
    r'official\s*token', r'official\s*coin',
]

# 常见无叙事意义的单词（过滤单词名币）
COMMON_NOISE_WORDS = {
    'nice', 'good', 'bad', 'cool', 'hot', 'big', 'small',
    'life', 'love', 'hate', 'happy', 'sad', 'fun', 'lol',
    'cat', 'dog', 'moon', 'sun', 'star', 'king', 'queen',
    'gold', 'rich', 'cash', 'money', 'pay', 'buy', 'sell',
    'pump', 'dump', 'bull', 'bear', 'green', 'red',
    'hello', 'world', 'yes', 'no', 'wow', 'omg', 'lmao',
    'simp', 'chad', 'based', 'cope', 'seethe',
    'test', 'new', 'old', 'real', 'fake',
    # 垃圾币名常见词
    'shit', 'shitcoin', 'fuck', 'fart', 'poop', 'pee',
    'cum', 'dick', 'ass', 'boob', 'tit',
    'nigga', 'retard', 'slop',
    # 超通用币名
    'the', 'and', 'for', 'from', 'with', 'this', 'that',
    'coin', 'token', 'meme', 'pepe', 'wojak',
    'peg', 'usd', 'usdt', 'usdc', 'dai',
}

# ============================================================
# 工具函数
# ============================================================
def log(msg):
    ts = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
    line = f"[{ts}] {msg}"
    print(line)
    os.makedirs(DATA_DIR, exist_ok=True)
    with open(LOG_FILE, 'a') as f:
        f.write(line + '\n')

def load_flap_seen():
    if os.path.exists(FLAP_SEEN_FILE):
        try:
            with open(FLAP_SEEN_FILE) as f:
                return json.load(f)
        except:
            pass
    return {}

def save_flap_seen(data):
    # 只保留7天内的
    cutoff = int(time.time()) - 86400 * 7
    data = {k: v for k, v in data.items() if v > cutoff}
    with open(FLAP_SEEN_FILE, 'w') as f:
        json.dump(data, f)

def tg_send(text, parse_mode='Markdown'):
    if not TG_TOKEN:
        log(f"[TG] No token, skip: {text[:80]}")
        return False
    try:
        resp = requests.post(
            f'https://api.telegram.org/bot{TG_TOKEN}/sendMessage',
            json={'chat_id': TG_CHAT_ID, 'text': text, 'parse_mode': parse_mode},
            timeout=10
        )
        result = resp.json()
        if not result.get('ok'):
            # Markdown失败时降级到纯文本
            if 'can\'t parse' in str(result.get('description', '')).lower():
                resp = requests.post(
                    f'https://api.telegram.org/bot{TG_TOKEN}/sendMessage',
                    json={'chat_id': TG_CHAT_ID, 'text': text},
                    timeout=10
                )
            else:
                log(f"[TG] Error: {result.get('description', '')}")
                return False
        return True
    except Exception as e:
        log(f"[TG] Send error: {e}")
        return False

# ============================================================
# 叙事历史数据库
# ============================================================
def init_db():
    """初始化SQLite叙事历史库"""
    conn = sqlite3.connect(DB_FILE)
    c = conn.cursor()
    
    # 所有见过的叙事主题
    c.execute('''CREATE TABLE IF NOT EXISTS narratives (
        id INTEGER PRIMARY KEY AUTOINCREMENT,
        theme TEXT NOT NULL,           -- 归一化的叙事主题（小写）
        first_token_name TEXT,         -- 第一次出现时的代币名
        first_token_address TEXT,      -- 第一次出现时的地址
        first_chain TEXT,              -- 第一次出现的链
        first_seen_at INTEGER,         -- 第一次看到的时间戳
        token_count INTEGER DEFAULT 1, -- 出现过多少次
        last_seen_at INTEGER           -- 最近一次看到
    )''')
    
    # 所有扫描过的代币
    c.execute('''CREATE TABLE IF NOT EXISTS tokens_seen (
        address TEXT PRIMARY KEY,
        chain TEXT,
        name TEXT,
        symbol TEXT,
        narrative_theme TEXT,
        category TEXT,                 -- 'musk_trump' / 'binance_cz' / 'novel' / 'common'
        first_seen_at INTEGER,
        market_cap REAL,
        pushed INTEGER DEFAULT 0,      -- 是否已推送
        seen_count INTEGER DEFAULT 1   -- 出现次数
    )''')
    
    # 索引
    c.execute('CREATE INDEX IF NOT EXISTS idx_theme ON narratives(theme)')
    c.execute('CREATE INDEX IF NOT EXISTS idx_addr ON tokens_seen(address)')
    
    conn.commit()
    return conn

def normalize_theme(name, symbol):
    """
    从代币名称+符号提取归一化的叙事主题
    例如：'Elon Mars Colony' → 'elon mars colony'
          'TRUMP2028' → 'trump'
          'PancakeBunny' → 'pancake bunny'
    """
    # 合并name和symbol
    text = f"{name} {symbol}".lower().strip()
    
    # 去除常见后缀/前缀
    noise = ['token', 'coin', 'inu', 'swap', 'finance', 'protocol',
             'dao', 'defi', 'nft', 'meta', 'verse', 'fi', 'ai',
             'pepe', 'wojak', 'chad', 'based']
    
    # 分割camelCase
    text = re.sub(r'([a-z])([A-Z])', r'\1 \2', text)
    # 去除数字（如2028、1000x）
    text = re.sub(r'\d+x?', '', text)
    # 只保留字母和空格
    text = re.sub(r'[^a-z\s]', ' ', text)
    # 去噪
    words = [w for w in text.split() if w and len(w) > 1 and w not in noise]
    
    if not words:
        return name.lower().strip()
    
    return ' '.join(sorted(set(words)))

def is_similar_theme(theme1, theme2, threshold=0.7):
    """模糊匹配两个叙事主题"""
    if theme1 == theme2:
        return True
    # 子串匹配
    if theme1 in theme2 or theme2 in theme1:
        return True
    # 词重叠
    words1 = set(theme1.split())
    words2 = set(theme2.split())
    if words1 and words2:
        overlap = len(words1 & words2) / min(len(words1), len(words2))
        if overlap >= 0.6:
            return True
    # 序列匹配
    return SequenceMatcher(None, theme1, theme2).ratio() >= threshold

def check_narrative_novelty(conn, theme, name, symbol, address, chain):
    """
    检查叙事状态
    返回：
      ('novel', None)                    — 第一次见到
      ('heating', narrative_row)         — 短时间内持续出现新币！热点信号！
      ('existing', existing_theme_row)   — 已有叙事，不热
    
    核心逻辑：同一主题在30分钟内出现2+个不同的币 = 热点
    """
    c = conn.cursor()
    now = int(time.time())
    HEAT_WINDOW = 1800  # 30分钟窗口
    HEAT_THRESHOLD = 2  # 窗口内出现2个以上同主题币就是热点
    
    # 精确匹配
    c.execute('SELECT id, theme, first_token_name, first_token_address, first_chain, first_seen_at, token_count, last_seen_at FROM narratives WHERE theme = ?', (theme,))
    exact = c.fetchone()
    if exact:
        row_id, _, _, _, _, first_seen, count, last_seen = exact
        # 更新计数
        new_count = count + 1
        c.execute('UPDATE narratives SET token_count = ?, last_seen_at = ? WHERE theme = ?',
                  (new_count, now, theme))
        conn.commit()
        
        # 热点判断：在HEAT_WINDOW内出现了多个币
        if now - first_seen < HEAT_WINDOW and new_count >= HEAT_THRESHOLD:
            return ('heating', exact)
        # 或者：最近一次和这次间隔很短（说明持续在冒）
        if now - last_seen < HEAT_WINDOW and new_count >= HEAT_THRESHOLD:
            return ('heating', exact)
        
        return ('existing', exact)
    
    # 模糊匹配 — 取最近1000个主题比对
    c.execute('SELECT id, theme, first_token_name, first_token_address, first_chain, first_seen_at, token_count, last_seen_at FROM narratives ORDER BY last_seen_at DESC LIMIT 1000')
    for row in c.fetchall():
        if is_similar_theme(theme, row[1]):
            row_id, _, _, _, _, first_seen, count, last_seen = row
            new_count = count + 1
            c.execute('UPDATE narratives SET token_count = ?, last_seen_at = ? WHERE id = ?',
                      (new_count, now, row[0]))
            conn.commit()
            
            # 热点判断
            if now - last_seen < HEAT_WINDOW and new_count >= HEAT_THRESHOLD:
                return ('heating', row)
            
            return ('existing', row)
    
    # 第一次见到 — 记录
    c.execute('''INSERT INTO narratives (theme, first_token_name, first_token_address, first_chain, first_seen_at, last_seen_at)
                 VALUES (?, ?, ?, ?, ?, ?)''',
              (theme, name, address, chain, now, now))
    conn.commit()
    return ('novel', None)

def get_token_seen_count(conn, address):
    """获取代币出现次数"""
    c = conn.cursor()
    c.execute('SELECT seen_count FROM tokens_seen WHERE address = ?', (address,))
    row = c.fetchone()
    return row[0] if row else 0

def is_token_seen(conn, address):
    """检查代币是否已经扫描过"""
    c = conn.cursor()
    c.execute('SELECT address FROM tokens_seen WHERE address = ?', (address,))
    return c.fetchone() is not None

def record_token(conn, address, chain, name, symbol, theme, category, mc, pushed=False):
    """记录已扫描的代币 — 重复出现时计数+1"""
    c = conn.cursor()
    # 检查是否已存在
    c.execute('SELECT seen_count FROM tokens_seen WHERE address = ?', (address,))
    existing = c.fetchone()
    if existing:
        # 已存在：计数+1，更新市值
        new_count = existing[0] + 1
        c.execute('''UPDATE tokens_seen SET seen_count = ?, market_cap = ?, category = ?
                     WHERE address = ?''', (new_count, mc, category, address))
    else:
        # 新记录
        c.execute('''INSERT INTO tokens_seen 
                     (address, chain, name, symbol, narrative_theme, category, first_seen_at, market_cap, pushed, seen_count)
                     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, 1)''',
                  (address, chain, name, symbol, theme, category, int(time.time()), mc, 1 if pushed else 0))
    conn.commit()

# ============================================================
# 叙事分类引擎
# ============================================================
def classify_narrative(name, symbol, chain):
    """
    分类代币叙事
    返回：('musk_trump', matched_keywords) / ('binance_cz', matched_keywords) / ('novel', None) / ('common', None)
    """
    text = f"{name} {symbol}".lower()
    
    # 1. 检查是否是垃圾币
    for pat in SPAM_PATTERNS:
        if re.search(pat, text, re.IGNORECASE):
            return ('spam', None)
    
    # 2. 马斯克/川普检测
    matched_mt = []
    for kw in MUSK_TRUMP_KEYWORDS:
        if kw.lower() in text:
            matched_mt.append(kw)
    if not matched_mt:
        for pat in MUSK_TRUMP_PATTERNS:
            m = re.search(pat, text, re.IGNORECASE)
            if m:
                matched_mt.append(m.group())
    
    if matched_mt:
        # 马斯克/川普：重点ETH+SOL，BSC也可以
        chain_lower = chain.lower()
        if chain_lower in ('eth', 'ethereum', 'sol', 'solana', 'bsc', 'base'):
            return ('musk_trump', matched_mt)
    
    # 3. 币安/CZ检测 — 只在BSC上推
    matched_bc = []
    for kw in BINANCE_CZ_KEYWORDS:
        if kw.lower() in text:
            matched_bc.append(kw)
    if not matched_bc:
        for pat in BINANCE_CZ_PATTERNS:
            m = re.search(pat, text, re.IGNORECASE)
            if m:
                matched_bc.append(m.group())
    
    if matched_bc:
        chain_lower = chain.lower()
        if chain_lower in ('bsc',):
            return ('binance_cz', matched_bc)
        else:
            return ('binance_cz_wrong_chain', matched_bc)
    
    # 4. 名人/推特热点检测（★★级别）
    matched_cv = []
    for kw in CELEBRITY_VIRAL_KEYWORDS:
        if kw.lower() in text:
            matched_cv.append(kw)
    if not matched_cv:
        for pat in CELEBRITY_VIRAL_PATTERNS:
            m = re.search(pat, text, re.IGNORECASE)
            if m:
                matched_cv.append(m.group())
    
    if matched_cv:
        return ('celebrity_viral', matched_cv)
    
    # 5. 都不匹配 → 需要进一步检查是否全新叙事
    return ('check_novelty', None)

# ============================================================
# 安全检查（复用现有逻辑）
# ============================================================
def check_token_safety(chain, address):
    """快速安全检查 — 只拦硬伤（蜜罐/可增发），卖税不作为否决条件"""
    if chain in ('sol', 'solana'):
        try:
            r = requests.get(f'https://api.rugcheck.xyz/v1/tokens/{address}/report', timeout=10)
            if r.status_code == 200:
                data = r.json()
                score = data.get('score', 999)
                mint = data.get('mintAuthority')
                freeze = data.get('freezeAuthority')
                return {
                    'safe': not mint and not freeze,
                    'score': score, 'mint': mint is not None,
                    'freeze': freeze is not None
                }
        except:
            pass
    else:
        chain_map = {'ethereum': '1', 'eth': '1', 'bsc': '56', 'base': '8453'}
        cid = chain_map.get(chain, '1')
        try:
            r = requests.get(f'https://api.gopluslabs.io/api/v1/token_security/{cid}?contract_addresses={address}', timeout=10)
            if r.status_code == 200:
                result = r.json().get('result', {})
                data = result.get(address.lower(), {})
                if data:
                    honeypot = data.get('is_honeypot', '0') == '1'
                    mintable = data.get('is_mintable', '0') == '1'
                    sell_tax = float(data.get('sell_tax', '0') or '0')
                    buy_tax = float(data.get('buy_tax', '0') or '0')
                    return {
                        'safe': not honeypot and not mintable,  # 卖税不作为否决
                        'honeypot': honeypot, 'mintable': mintable,
                        'sell_tax': sell_tax, 'buy_tax': buy_tax
                    }
        except:
            pass
    return {'safe': False, 'reason': '无法检查'}  # 无法检查时不推，宁可错过不踩坑

# ============================================================
# GMGN数据获取
# ============================================================
def gmgn_get(url):
    try:
        resp = requests.get(url, headers=GMGN_HEADERS, timeout=15)
        if resp.status_code == 200:
            return resp.json().get('data', {})
    except:
        pass
    return {}

def fetch_token_description(chain, address):
    """获取代币描述/故事 — 叙事雷达核心信息"""
    desc = ''
    
    # SOL链：Pump.fun有最完整的description
    if chain in ('sol', 'solana'):
        try:
            r = requests.get(f'https://frontend-api-v3.pump.fun/coins/{address}', timeout=8)
            if r.status_code == 200:
                data = r.json()
                desc = data.get('description', '') or ''
                twitter = data.get('twitter', '') or ''
                telegram = data.get('telegram', '') or ''
                website = data.get('website', '') or ''
                return {
                    'description': desc.strip(),
                    'twitter': twitter,
                    'telegram': telegram,
                    'website': website,
                }
        except:
            pass
    
    # 所有链：DEXScreener info字段（网站+社交链接）
    try:
        chain_dex = {'sol': 'solana', 'eth': 'ethereum', 'bsc': 'bsc', 'base': 'base',
                     'solana': 'solana', 'ethereum': 'ethereum'}.get(chain, chain)
        r = requests.get(f'https://api.dexscreener.com/latest/dex/tokens/{address}', timeout=8)
        if r.status_code == 200:
            pairs = r.json().get('pairs', [])
            if pairs:
                info = pairs[0].get('info', {})
                websites = info.get('websites', [])
                socials = info.get('socials', [])
                twitter = ''
                telegram = ''
                website = ''
                for s in socials:
                    if s.get('type') == 'twitter':
                        twitter = s.get('url', '')
                    elif s.get('type') == 'telegram':
                        telegram = s.get('url', '')
                for w in websites:
                    if w.get('label', '').lower() == 'website':
                        website = w.get('url', '')
                if not desc:
                    # DEXScreener没有description但有社交信息
                    return {
                        'description': desc,
                        'twitter': twitter,
                        'telegram': telegram,
                        'website': website,
                    }
    except:
        pass
    
    return {'description': desc, 'twitter': '', 'telegram': '', 'website': ''}

def fetch_new_tokens():
    """从GMGN获取各链新币 + 多维度覆盖"""
    all_tokens = []
    seen_addrs = set()
    
    for chain in ['eth', 'bsc', 'base']:
        # 多维度拉数据，避免漏掉
        urls = [
            # 按创建时间 — 最新的币
            f'https://gmgn.ai/defi/quotation/v1/rank/{chain}/swaps/1h?orderby=open_timestamp&direction=desc&limit=100',
            # 按交易量 — 最活跃的币
            f'https://gmgn.ai/defi/quotation/v1/rank/{chain}/swaps/1h?orderby=swaps&direction=desc&limit=50',
        ]
        
        for url in urls:
            data = gmgn_get(url)
            tokens = data.get('rank', [])
            
            for t in tokens:
                addr = t.get('address', '')
                if not addr or addr in seen_addrs:
                    continue
                
                mc = t.get('market_cap', 0) or t.get('fdv', 0) or 0
                liq = t.get('liquidity', 0) or 0
                
                # 基本过滤：太小的不看
                if mc < 1000 or liq < 500 or mc > 10000000:
                    continue
                
                age_ts = t.get('open_timestamp', 0)
                age_h = (time.time() - age_ts) / 3600 if age_ts > 0 else 999
                
                # 不限年龄 — 动量追踪核心逻辑：涨就推，不管新旧
                
                seen_addrs.add(addr)
                all_tokens.append({
                    'address': addr,
                    'chain': chain,
                    'name': t.get('name', '?'),
                    'symbol': t.get('symbol', '?'),
                    'mc': mc,
                    'liq': liq,
                    'volume': t.get('volume', 0) or 0,
                    'holders': t.get('holder_count', 0) or 0,
                    'sm': t.get('smart_degen_count', 0) or 0,
                    'chg_1h': t.get('price_change_percent1h', 0) or 0,
                    'chg_24h': t.get('price_change_percent', 0) or 0,
                    'age_h': age_h,
                    'price': t.get('price', 0),
                    'buys_1h': t.get('buys', 0) or 0,
                    'sells_1h': t.get('sells', 0) or 0,
                })
            
            time.sleep(0.3)
    
    return all_tokens

def fetch_flap_tokens():
    """
    FLAP平台扫描 — BSC社区驱动型发射台
    找形态：跌下来但有底部支撑（有庄在低位推）
    特征：24h跌了，但1h企稳/反弹，买入>卖出，holders在涨
    """
    data = gmgn_get(
        'https://gmgn.ai/defi/quotation/v1/rank/bsc/swaps/24h?launchpad=flap&orderby=volume&direction=desc&limit=30'
    )
    tokens = data.get('rank', [])
    
    candidates = []
    for t in tokens:
        addr = t.get('address', '')
        if not addr:
            continue
        
        mc = t.get('market_cap', 0) or 0
        liq = t.get('liquidity', 0) or 0
        vol = t.get('volume', 0) or 0
        holders = t.get('holder_count', 0) or 0
        buys = t.get('buys', 0) or 0
        sells = t.get('sells', 0) or 0
        chg_1h = t.get('price_change_percent1h', 0) or 0
        chg_24h = t.get('price_change_percent', 0) or 0
        age_ts = t.get('open_timestamp', 0)
        age_h = (time.time() - age_ts) / 3600 if age_ts > 0 else 0
        
        # 基本门槛
        if mc < 1000 or liq < 500:
            continue
        if holders < 5:
            continue
        
        # 底部支撑形态判断：
        # 条件1: 24h跌了（或者涨幅有限），说明不是刚拉的
        # 条件2: 1h跌幅小于24h跌幅，说明在企稳
        # 条件3: 买入 > 卖出，有人在接
        buy_ratio = buys / max(sells, 1)
        
        is_support = False
        reason = ''
        
        # 形态A: 24h跌了，1h在企稳/反弹
        if chg_24h < -10 and chg_1h > chg_24h * 0.3:
            is_support = True
            reason = f'24h跌{chg_24h:.0f}%但1h企稳{chg_1h:+.0f}%'
        
        # 形态B: 24h微跌或横盘，1h微涨，买卖比健康
        if -10 <= chg_24h <= 30 and chg_1h > -5 and buy_ratio > 1.1:
            is_support = True
            reason = f'底部横盘 买卖比{buy_ratio:.2f}'
        
        # 形态C: 大跌后强反弹
        if chg_24h < -30 and chg_1h > 10:
            is_support = True
            reason = f'大跌{chg_24h:.0f}%后反弹{chg_1h:+.0f}%'
        
        if is_support and buy_ratio >= 1.0:
            candidates.append({
                'address': addr,
                'chain': 'bsc',
                'name': t.get('name', '?'),
                'symbol': t.get('symbol', '?'),
                'mc': mc,
                'liq': liq,
                'volume': vol,
                'holders': holders,
                'sm': 0,
                'chg_1h': chg_1h,
                'chg_24h': chg_24h,
                'age_h': age_h,
                'price': t.get('price', 0),
                'buys': buys,
                'sells': sells,
                'buy_ratio': buy_ratio,
                'support_reason': reason,
                'launchpad': 'flap',
            })
    
    # 按市值排序
    candidates.sort(key=lambda x: x['mc'], reverse=True)
    return candidates

def format_flap_alert(token, desc_info=None):
    """FLAP低吸信号推送"""
    msg = f"链上雷达 — FLAP低吸信号\n"
    msg += f"链: BSC | 平台: FLAP\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    # 故事描述
    desc = (desc_info or {}).get('description', '')
    if desc:
        if len(desc) > 200:
            desc = desc[:200] + '...'
        msg += f"故事: {desc}\n\n"
    
    msg += f"形态: {token['support_reason']}\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"24h量    ${token['volume']:>12,.0f}\n"
    msg += f"持有人   {token['holders']:>12,d}\n"
    msg += f"买/卖    {token['buys']:>6,d}/{token['sells']:>6,d}\n"
    msg += f"买卖比   {token['buy_ratio']:>12.2f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    msg += f"24h涨幅  {token['chg_24h']:>+11.1f}%\n"
    msg += f"```\n"
    msg += "\nFLAP社区币 — 低吸进场信号"
    
    # 社交链接
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"\nTwitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if (desc_info or {}).get('website'):
        links.append(f"Web: {desc_info['website']}")
    if links:
        msg += '\n'.join(links)
    
    return msg

# ============================================================
# 推送格式
# ============================================================
def format_musk_trump_alert(token, matched_kw, desc_info=None):
    """马斯克/川普叙事推送"""
    chain_map = {'sol': 'SOL', 'eth': 'ETH', 'bsc': 'BSC', 'base': 'BASE'}
    ch = chain_map.get(token['chain'], token['chain'].upper())
    
    msg = f"链上雷达 — 马斯克/川普概念\n"
    msg += f"链: {ch}\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    # 叙事故事（核心！）
    desc = (desc_info or {}).get('description', '')
    if desc:
        # 截取前200字符，避免太长
        if len(desc) > 200:
            desc = desc[:200] + '...'
        msg += f"故事: {desc}\n\n"
    
    msg += f"命中关键词: {', '.join(matched_kw[:5])}\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    if token.get('sm', 0) > 0:
        msg += f"聪明钱   {token['sm']:>12d}\n"
    msg += f"币龄     {token['age_h']:>10.1f}h\n"
    msg += f"```\n"
    
    # 社交链接
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"Twitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if (desc_info or {}).get('website'):
        links.append(f"Web: {desc_info['website']}")
    if links:
        msg += '\n' + '\n'.join(links)
    
    return msg

def format_binance_cz_alert(token, matched_kw, desc_info=None):
    """币安/CZ叙事推送"""
    msg = f"链上雷达 — 币安/CZ概念\n"
    msg += f"链: BSC\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    # 叙事故事
    desc = (desc_info or {}).get('description', '')
    if desc:
        if len(desc) > 200:
            desc = desc[:200] + '...'
        msg += f"故事: {desc}\n\n"
    
    msg += f"命中关键词: {', '.join(matched_kw[:5])}\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    msg += f"币龄     {token['age_h']:>10.1f}h\n"
    msg += f"```\n"
    
    # 社交链接
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"Twitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if (desc_info or {}).get('website'):
        links.append(f"Web: {desc_info['website']}")
    if links:
        msg += '\n' + '\n'.join(links)
    
    return msg

def format_novel_narrative_alert(token, theme, desc_info=None):
    """全新叙事推送 — 保留备用"""
    return format_heating_narrative_alert(token, theme, 1, desc_info)

def format_heating_narrative_alert(token, theme, count, desc_info=None):
    """叙事热点推送 — 同主题持续冒新币"""
    chain_map = {'sol': 'SOL', 'eth': 'ETH', 'bsc': 'BSC', 'base': 'BASE'}
    ch = chain_map.get(token['chain'], token['chain'].upper())
    
    msg = f"链上雷达 — 叙事热点\n"
    msg += f"链: {ch}\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    # 叙事故事
    desc = (desc_info or {}).get('description', '')
    if desc:
        if len(desc) > 300:
            desc = desc[:300] + '...'
        msg += f"故事: {desc}\n\n"
    else:
        msg += f"叙事主题: {theme}\n\n"
    
    msg += f"同类概念已出现{count}个币 — 持续有人做\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    if token.get('sm', 0) > 0:
        msg += f"聪明钱   {token['sm']:>12d}\n"
    msg += f"持有人   {token['holders']:>12d}\n"
    msg += f"币龄     {token['age_h']:>10.1f}h\n"
    msg += f"```"
    
    # 社交链接
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"\nTwitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if (desc_info or {}).get('website'):
        links.append(f"Web: {desc_info['website']}")
    if links:
        msg += '\n'.join(links)
    
    return msg

# ============================================================
# 动量追踪器 — 持续上涨+放量检测
# ============================================================
def track_momentum(tokens):
    """
    每轮扫描更新币的快照。
    连续多轮市值上涨+成交量增加 = 动量信号，直接推。
    """
    global MOMENTUM_TRACKER, MOMENTUM_PUSHED
    now = time.time()
    alerts = []
    
    # 当前轮所有地址
    current_addrs = set()
    
    for token in tokens:
        addr = token['address']
        mc = token['mc']
        vol = token.get('volume', 0) or 0
        price = token.get('price', 0) or 0
        buys = token.get('buys_1h', 0) or token.get('buys', 0) or 0
        
        current_addrs.add(addr)
        
        # 基本门槛
        if mc < 1000 or token.get('liq', 0) < 500 or mc > 10000000:
            continue
        
        # 记录快照 — 只有数据真正变化时才记录（GMGN有缓存）
        if addr not in MOMENTUM_TRACKER:
            MOMENTUM_TRACKER[addr] = []
        
        snapshots = MOMENTUM_TRACKER[addr]
        
        # 跳过重复数据（跟上一次完全一样就不记录）
        if snapshots and snapshots[-1]['mc'] == mc and snapshots[-1]['vol'] == vol:
            continue  # 数据没变，跳过
        
        snapshots.append({
            'ts': now,
            'mc': mc,
            'vol': vol,
            'price': price,
            'buys': buys,
        })
        
        # 只保留最近20个快照（约200秒）
        if len(snapshots) > 20:
            snapshots[:] = snapshots[-20:]
        
        # 至少需要3个快照才能判断
        if len(snapshots) < MOMENTUM_CONSECUTIVE_UP:
            continue
        
        # 检测最近N轮是否持续涨
        recent = snapshots[-MOMENTUM_CONSECUTIVE_UP:]
        consecutive_up = True
        total_gain = 0
        
        for i in range(1, len(recent)):
            prev_mc = recent[i-1]['mc']
            curr_mc = recent[i]['mc']
            if prev_mc <= 0:
                consecutive_up = False
                break
            gain = (curr_mc - prev_mc) / prev_mc
            if gain <= 0:  # 任何一轮没涨就不算
                consecutive_up = False
                break
            total_gain += gain
        
        if not consecutive_up:
            continue
        
        # 连续涨了！检查放量（成交量在增）
        vol_increasing = True
        for i in range(1, len(recent)):
            if recent[i]['buys'] < recent[i-1]['buys'] * 0.8:  # 允许小幅波动
                vol_increasing = False
                break
        
        # 计算总涨幅
        first_mc = recent[0]['mc']
        last_mc = recent[-1]['mc']
        pct_gain = ((last_mc - first_mc) / first_mc * 100) if first_mc > 0 else 0
        
        # 推送条件：连续涨 + 涨幅>5%
        if pct_gain < 5:
            continue
        
        # 信号计数：同一个币每次触发信号，计数+1
        push_info = MOMENTUM_PUSHED.get(addr, {'count': 0, 'last_ts': 0, 'last_mc': 0})
        
        # 必须比上次推送时市值还高才推（真的还在涨）
        if push_info['count'] > 0 and last_mc <= push_info['last_mc']:
            continue
        
        push_info['count'] += 1
        push_info['last_ts'] = now
        push_info['last_mc'] = last_mc
        signal_count = push_info['count']
        
        # 安全检查
        safety = check_token_safety(token['chain'], addr)
        if not safety.get('safe'):
            continue
        
        # 叙事分类 → 星级评分
        category, matched_kw = classify_narrative(token['name'], token['symbol'], token['chain'])
        is_flap = token.get('launchpad') == 'flap'
        
        if category == 'musk_trump':
            stars = 3
            narrative_tag = f"马斯克/川普概念 ({', '.join(matched_kw[:3])})"
        elif category == 'binance_cz':
            stars = 3
            narrative_tag = f"币安/CZ概念 ({', '.join(matched_kw[:3])})"
        elif category == 'celebrity_viral':
            stars = 2
            narrative_tag = f"名人/热点 ({', '.join(matched_kw[:3])})"
        elif is_flap:
            stars = 2
            narrative_tag = "FLAP社区币"
        else:
            # 检查是否全新叙事
            theme = normalize_theme(token['name'], token['symbol'])
            theme_words = [w for w in theme.split() if w not in COMMON_NOISE_WORDS and len(w) > 2]
            if len(theme_words) >= 2:
                stars = 2
                narrative_tag = f"叙事: {theme}"
            else:
                stars = 1
                narrative_tag = "无明确叙事"
        
        # 生成推送
        desc_info = fetch_token_description(token['chain'], addr)
        
        # FLAP币额外标注社区/CTO信息
        if is_flap:
            has_twitter = bool(desc_info.get('twitter'))
            has_tg = bool(desc_info.get('telegram'))
            has_web = bool(desc_info.get('website'))
            community_tags = []
            if has_twitter:
                community_tags.append("有推特")
            if has_tg:
                community_tags.append("有TG群")
            if has_web:
                community_tags.append("有官网")
            if community_tags:
                narrative_tag += f" | {' '.join(community_tags)}"
                stars = min(3, stars + 1)  # 有社区加一星
            else:
                narrative_tag += " | 无社区链接"
        
        msg = format_momentum_alert(token, pct_gain, len(recent), vol_increasing, stars, narrative_tag, desc_info, signal_count)
        alerts.append({'msg': msg, 'token': token})
        MOMENTUM_PUSHED[addr] = push_info
        
        log(f"[动量信号{signal_count}] {token['name']} ({token['symbol']}) on {token['chain']} — 连涨{len(recent)}轮 +{pct_gain:.1f}%")
    
    # 清理不再出现的币
    stale = [a for a in MOMENTUM_TRACKER if a not in current_addrs]
    for a in stale:
        if now - MOMENTUM_TRACKER[a][-1]['ts'] > 600:  # 10分钟没出现就清理
            del MOMENTUM_TRACKER[a]
    
    # 清理推送记录 — 1小时没出现的清掉
    MOMENTUM_PUSHED = {k: v for k, v in MOMENTUM_PUSHED.items() if now - v.get('last_ts', 0) < 3600}
    
    return alerts

def format_momentum_alert(token, pct_gain, rounds, vol_up, stars, narrative_tag, desc_info=None, seen_count=0):
    """持续上涨动量推送 — 带叙事星级"""
    chain_map = {'sol': 'SOL', 'eth': 'ETH', 'bsc': 'BSC', 'base': 'BASE'}
    ch = chain_map.get(token['chain'], token['chain'].upper())
    
    vol_tag = "放量" if vol_up else ""
    star_str = "★" * stars + "☆" * (3 - stars)
    
    # 信号编号从标题移到下面
    msg = f"链上雷达\n"
    msg += f"链: {ch}\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    # 故事描述
    desc = (desc_info or {}).get('description', '')
    if desc:
        if len(desc) > 200:
            desc = desc[:200] + '...'
        msg += f"故事: {desc}\n\n"
    
    msg += f"叙事: {narrative_tag}\n"
    msg += f"连涨{rounds}轮 +{pct_gain:.1f}% {vol_tag}\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    if token.get('sm', 0) > 0:
        msg += f"聪明钱   {token['sm']:>12d}\n"
    msg += f"币龄     {token['age_h']:>10.1f}h\n"
    msg += f"```\n"
    msg += f"评星: {star_str}  出现次数: {seen_count}"
    
    # 社交链接
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"\nTwitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if (desc_info or {}).get('website'):
        links.append(f"Web: {desc_info['website']}")
    if links:
        msg += '\n'.join(links)
    
    return msg

def format_celebrity_alert(token, matched_kw, desc_info=None):
    """名人/推特热点推送 ★★"""
    chain_map = {'sol': 'SOL', 'eth': 'ETH', 'bsc': 'BSC', 'base': 'BASE'}
    ch = chain_map.get(token['chain'], token['chain'].upper())
    
    msg = f"链上雷达 — 名人/热点 ★★\n"
    msg += f"链: {ch}\n\n"
    msg += f"{token['name']} ({token['symbol']})\n"
    msg += f"`{token['address']}`\n\n"
    
    desc = (desc_info or {}).get('description', '')
    if desc:
        if len(desc) > 200:
            desc = desc[:200] + '...'
        msg += f"故事: {desc}\n\n"
    
    msg += f"命中关键词: {', '.join(matched_kw[:5])}\n\n"
    msg += f"```\n"
    msg += f"市值     ${token['mc']:>12,.0f}\n"
    msg += f"流动性   ${token['liq']:>12,.0f}\n"
    msg += f"1h涨幅   {token['chg_1h']:>+11.1f}%\n"
    if token.get('sm', 0) > 0:
        msg += f"聪明钱   {token['sm']:>12d}\n"
    msg += f"币龄     {token['age_h']:>10.1f}h\n"
    msg += f"```"
    
    links = []
    if (desc_info or {}).get('twitter'):
        links.append(f"\nTwitter: {desc_info['twitter']}")
    if (desc_info or {}).get('telegram'):
        links.append(f"TG: {desc_info['telegram']}")
    if links:
        msg += '\n'.join(links)
    
    return msg

# ============================================================
# 核心扫描逻辑
# ============================================================
def scan_narratives():
    """主扫描函数"""
    conn = init_db()
    tokens = fetch_new_tokens()
    
    log(f"扫描 {len(tokens)} 个新币...")
    
    # === 动量追踪 — 每轮更新所有币的快照，检测持续上涨 ===
    # 拉FLAP币一起喂进动量追踪器
    flap_tokens = []
    try:
        flap_tokens = fetch_flap_tokens()
    except:
        pass
    all_momentum_tokens = tokens + flap_tokens
    momentum_alerts = track_momentum(all_momentum_tokens)
    
    for token in tokens:
        addr = token['address']
        chain = token['chain']
        name = token['name']
        symbol = token['symbol']
        
        # 已扫描过的 — 更新seen_count和narratives的token_count，但不重复推
        if is_token_seen(conn, addr):
            # 更新seen_count
            c = conn.cursor()
            c.execute('UPDATE tokens_seen SET seen_count = seen_count + 1, market_cap = ? WHERE address = ?', (token['mc'], addr))
            # 更新narratives表的token_count（按主题）
            theme_tmp = normalize_theme(name, symbol)
            if theme_tmp:
                c.execute('UPDATE narratives SET token_count = token_count + 1, last_seen_at = ? WHERE theme = ?', (int(time.time()), theme_tmp))
            conn.commit()
            continue
        
        # 分类叙事
        category, matched_kw = classify_narrative(name, symbol, chain)
        
        if category == 'spam':
            record_token(conn, addr, chain, name, symbol, '', 'spam', token['mc'])
            continue
        
        # 基本质量门槛（防止推太多垃圾）
        min_mc = 1000
        min_liq = 500
        if token['mc'] < min_mc or token['liq'] < min_liq:
            record_token(conn, addr, chain, name, symbol, '', 'too_small', token['mc'])
            continue
        
        theme = normalize_theme(name, symbol)
        
        # 所有分类只记录，不直接推送 — 推送统一走动量引擎
        record_token(conn, addr, chain, name, symbol, theme, category, token['mc'])
        check_narrative_novelty(conn, theme, name, symbol, addr, chain)
    
    conn.close()
    
    # === 推送动量信号 ===
    pushed = 0
    for ma in momentum_alerts[:8]:  # 单轮最多推8个
        if tg_send(ma['msg']):
            pushed += 1
            time.sleep(1)  # 避免TG限流
    
    return pushed, len(momentum_alerts)

# ============================================================
# 主循环
# ============================================================
def main():
    log("=" * 50)
    log("链上雷达 v1 启动")
    log(f"扫描间隔: {SCAN_INTERVAL}s")
    log(f"推送逻辑: 动量优先 — 连涨才推，叙事只做分类标签")
    log("=" * 50)
    
    # 初始化DB
    init_db()
    
    # 启动通知
    tg_send(
        "链上雷达 v1 已启动\n\n"
        "核心逻辑: 动量优先\n"
        "连涨3轮+涨幅>5%才推送\n"
        "叙事只做分类标签:\n"
        "★★★ 马斯克/川普 | 币安/CZ | FLAP有社区\n"
        "★★ 名人热点 | FLAP无社区 | 有叙事\n"
        "★ 无明确叙事\n\n"
        f"扫描频率: 每{SCAN_INTERVAL}秒"
    )
    
    scan_count = 0
    total_pushed = 0
    
    while True:
        try:
            scan_count += 1
            pushed, found = scan_narratives()
            total_pushed += pushed
            
            if pushed > 0:
                log(f"第{scan_count}轮: 发现{found}个, 推送{pushed}个 (累计推送{total_pushed})")
            else:
                if scan_count % 20 == 0:  # 每20轮报一次无信号
                    log(f"第{scan_count}轮: 无新信号 (累计推送{total_pushed})")
            
        except Exception as e:
            log(f"扫描异常: {e}")
        
        time.sleep(SCAN_INTERVAL)

if __name__ == '__main__':
    main()