bifitnex-trading/check_errors.py

#!/usr/bin/env python3
"""Check API errors in the last hour, attempt auto-fix via Claude, and report to Slack."""

import json
import os
import shutil
import subprocess
import sys
from datetime import datetime, timedelta

sys.path.insert(0, os.path.dirname(__file__))

import slack_notifier

PROJECT_DIR = os.path.dirname(os.path.abspath(__file__))

LOG_FILES = [
    "trading.log",
    "sync_cost_basis.log",
    "cron.log",
]

LOOKBACK_MINUTES = 60


def collect_recent_errors() -> list[str]:
    since = datetime.now() - timedelta(minutes=LOOKBACK_MINUTES)
    since_str = since.strftime("%Y-%m-%d %H:%M")
    errors = []

    for logfile in LOG_FILES:
        path = os.path.join(PROJECT_DIR, logfile)
        if not os.path.exists(path):
            continue
        with open(path) as f:
            for line in f:
                if "[ERROR]" not in line:
                    continue
                if not line[:4].isdigit():
                    continue
                if line[:16] >= since_str:
                    errors.append(line.rstrip())

    return errors


def attempt_auto_fix(errors: list[str]) -> str | None:
    """Ask Claude to diagnose and fix the errors. Returns a summary or None on failure."""
    error_text = "\n".join(errors[:20])
    prompt = f"""你是 bifitnex-trading 專案的維運工程師。
專案目錄: {PROJECT_DIR}

以下是最近一小時的錯誤 log：
```
{error_text}
```

請：
1. 診斷錯誤的根因
2. 如果是程式碼問題，直接修復（編輯檔案）
3. 如果是外部問題（API 暫時不可用、網路問題等），不需要改程式碼，只需說明

最後用以下 JSON 格式回覆：
```json
{{
  "diagnosis": "錯誤根因簡述",
  "fixed": true/false,
  "changes": "修改了什麼（沒改就寫 null）",
  "suggestion": "如需人工介入的建議（不需要就寫 null）"
}}
```
只回傳 JSON，不要其他文字。"""

    claude_bin = shutil.which("claude") or os.path.expanduser("~/.local/bin/claude")

    try:
        result = subprocess.run(
            [claude_bin, "-p", prompt, "--output-format", "json",
             "--no-session-persistence", "--allowedTools", "Read,Edit,Glob,Grep"],
            capture_output=True,
            text=True,
            timeout=180,
            cwd=PROJECT_DIR,
        )
        if result.returncode != 0:
            return None
        return _parse_fix_response(result.stdout)
    except (subprocess.TimeoutExpired, FileNotFoundError, Exception):
        return None


def _parse_fix_response(raw: str) -> dict | None:
    """Parse Claude's JSON response."""
    try:
        wrapper = json.loads(raw)
        text = wrapper.get("result", raw) if isinstance(wrapper, dict) else raw
    except json.JSONDecodeError:
        text = raw

    # Try to extract JSON from the text
    try:
        return json.loads(text)
    except json.JSONDecodeError:
        pass

    # Try finding JSON block in text
    start = text.find("{")
    end = text.rfind("}") + 1
    if start >= 0 and end > start:
        try:
            return json.loads(text[start:end])
        except json.JSONDecodeError:
            pass

    return None


def main():
    errors = collect_recent_errors()
    if not errors:
        return

    unique = list(dict.fromkeys(errors))
    display = unique[:20]
    remaining = len(unique) - len(display)

    # Try auto-fix
    fix_result = attempt_auto_fix(unique)

    # Build Slack message
    lines = [
        "<!channel>",
        f"⚠️ *API 錯誤警報* (最近 {LOOKBACK_MINUTES} 分鐘，共 {len(unique)} 筆)",
        "",
    ]
    for e in display:
        lines.append(f"• `{e[:200]}`")
    if remaining > 0:
        lines.append(f"\n...另有 {remaining} 筆錯誤")

    # Append fix report
    lines.append("")
    if fix_result:
        lines.append("🔧 *自動修復報告：*")
        lines.append(f"  診斷：{fix_result.get('diagnosis', '未知')}")
        if fix_result.get("fixed"):
            lines.append(f"  ✅ 已修復：{fix_result.get('changes', '-')}")
        else:
            lines.append("  ❌ 未自動修復")
        if fix_result.get("suggestion"):
            lines.append(f"  💡 建議：{fix_result.get('suggestion')}")
    else:
        lines.append("🔧 *自動修復：* Claude 分析失敗，請人工檢查")

    slack_notifier._send({"text": "\n".join(lines)})


if __name__ == "__main__":
    main()