2e2f5a1881
Some checks failed
CD Pipeline / build-and-deploy (push) Has been cancelled
統帥批准 Phase 5 全 Sprint，Sprint 5.0 產出:
1. callback_action_spec.yaml (24 actions)
- 10 查類 (info 2-part callback, 無副作用): check_process, check_port,
check_log_*, check_health, check_pod_logs, describe_pod, open_signoz,
open_flywheel
- 10 寫類 (nonce 4-part, 有副作用): k8s_restart/scale_up/scale_down/rollback,
host_restart_service/clear_log, docker_restart, minio_restart,
reload_nginx, renew_cert
- 4 secops (Multi-Sig CRITICAL): secops_isolate/block_ip/evict/authorize
2. callback_dispatcher.py
- Registry pattern (lru_cache): get_action_spec / list_actions_for_category
- 模板變數替換: {incident_id} / {labels.xxx} / {signals[0].xxx}
- dispatch_action() 骨架 (Sprint 5.2+ 接 MCP)
- _format_reply: text/code/truncated/url 4 種格式
3. test_callback_dispatcher.py (22 tests全過)
- Registry loading 正確性
- Category filtering
- Template resolution (含 nested list index)
- dispatch stub 返回正確 spec 提示
下一步 Sprint 5.1: 接入 MCP registry + telegram callback_handler 整合
MCP 底層能力已有 (k8s 10+ tools, ssh 15 tools)
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
325 lines
10 KiB
Python
325 lines
10 KiB
Python
"""
|
||
Telegram Callback Dispatcher — 分類按鈕統一調度
|
||
================================================
|
||
Phase 5 Sprint 5.0-5.1 — 2026-04-14 Claude Sonnet 4.6
|
||
相關: docs/superpowers/plans/2026-04-14-PHASE-5-category-buttons-completion.md
|
||
ADR-079 分類按鈕完整化
|
||
|
||
職責:
|
||
1. 從 callback_action_spec.yaml 載入 action registry
|
||
2. 接收 Telegram callback_data (action:incident_id or action:id:ts:rand)
|
||
3. 驗證 nonce（寫類按鈕）或 allow info（查類按鈕）
|
||
4. 依 spec 呼叫對應 MCP tool
|
||
5. Reply 執行結果到原告警卡片（reply_to_message_id）
|
||
|
||
設計原則:
|
||
- Registry pattern — 新增按鈕只需 yaml 一行，無需改 dispatcher code
|
||
- 模板變數: {incident_id} / {labels.xxx} / {signals[0].xxx} / {callback.user_id}
|
||
- 所有 action 都有 audit log（寫類額外 nonce 驗證 log）
|
||
- reply_to 原告警 message_id（從 Redis tg_msg:{incident_id}）
|
||
|
||
遵守「禁止 Mock 測試鐵律」: 純邏輯 + MCP dispatch，測試用真實 registry。
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import time
|
||
from dataclasses import dataclass
|
||
from functools import lru_cache
|
||
from pathlib import Path
|
||
from typing import Any
|
||
|
||
import structlog
|
||
import yaml
|
||
|
||
logger = structlog.get_logger(__name__)
|
||
|
||
|
||
# =============================================================================
|
||
# Data Types
|
||
# =============================================================================
|
||
|
||
|
||
@dataclass
|
||
class ActionSpec:
|
||
"""從 callback_action_spec.yaml 載入的單一 action 規格"""
|
||
|
||
name: str
|
||
label: str
|
||
emoji: str
|
||
risk: str # low | medium | high | critical
|
||
callback_format: str # info | nonce
|
||
category: str
|
||
mcp_provider: str # k8s | ssh | prometheus | signoz | database | internal
|
||
mcp_tool: str
|
||
mcp_params: dict[str, Any]
|
||
reply_format: str # text | code | url | truncated
|
||
timeout_sec: int
|
||
description: str
|
||
requires_multi_sig: bool = False
|
||
|
||
|
||
@dataclass
|
||
class DispatchResult:
|
||
"""Dispatcher 執行結果"""
|
||
|
||
success: bool
|
||
action: str
|
||
incident_id: str
|
||
user_id: int | None
|
||
result_text: str
|
||
error: str | None = None
|
||
duration_ms: float = 0.0
|
||
|
||
|
||
# =============================================================================
|
||
# Spec Registry
|
||
# =============================================================================
|
||
|
||
|
||
@lru_cache(maxsize=1)
|
||
def load_action_registry() -> dict[str, ActionSpec]:
|
||
"""
|
||
載入 callback_action_spec.yaml 並快取（進程內不重載，重啟 Pod 才更新）
|
||
|
||
Returns:
|
||
{action_name: ActionSpec}
|
||
"""
|
||
spec_path = Path(__file__).parent / "callback_action_spec.yaml"
|
||
if not spec_path.exists():
|
||
logger.warning("callback_action_spec_not_found", path=str(spec_path))
|
||
return {}
|
||
|
||
with spec_path.open("r", encoding="utf-8") as f:
|
||
data = yaml.safe_load(f)
|
||
|
||
registry: dict[str, ActionSpec] = {}
|
||
for name, spec_dict in (data.get("actions") or {}).items():
|
||
mcp = spec_dict.get("mcp", {}) or {}
|
||
registry[name] = ActionSpec(
|
||
name=name,
|
||
label=spec_dict.get("label", name),
|
||
emoji=spec_dict.get("emoji", ""),
|
||
risk=spec_dict.get("risk", "medium"),
|
||
callback_format=spec_dict.get("callback_format", "info"),
|
||
category=spec_dict.get("category", ""),
|
||
mcp_provider=mcp.get("provider", ""),
|
||
mcp_tool=mcp.get("tool", ""),
|
||
mcp_params=mcp.get("params") or {},
|
||
reply_format=spec_dict.get("reply_format", "text"),
|
||
timeout_sec=int(spec_dict.get("timeout_sec", 10)),
|
||
description=spec_dict.get("description", ""),
|
||
requires_multi_sig=bool(spec_dict.get("requires_multi_sig", False)),
|
||
)
|
||
logger.info("callback_action_registry_loaded", count=len(registry))
|
||
return registry
|
||
|
||
|
||
def get_action_spec(action_name: str) -> ActionSpec | None:
|
||
"""查找單一 action 規格"""
|
||
return load_action_registry().get(action_name)
|
||
|
||
|
||
def list_actions_for_category(alert_category: str) -> list[ActionSpec]:
|
||
"""列出特定分類的所有可用 action（供 _build_inline_keyboard 使用）"""
|
||
return [
|
||
spec for spec in load_action_registry().values()
|
||
if spec.category == alert_category
|
||
]
|
||
|
||
|
||
# =============================================================================
|
||
# Template Variable Substitution
|
||
# =============================================================================
|
||
|
||
|
||
def _resolve_template(template: Any, context: dict) -> Any:
|
||
"""
|
||
遞迴替換模板變數。
|
||
|
||
支援:
|
||
- {incident_id}
|
||
- {labels.xxx} / {labels.xxx.yyy}
|
||
- {signals[0].xxx}
|
||
- {callback.user_id}
|
||
|
||
Example:
|
||
template = {"host": "{labels.instance}", "lines": 50}
|
||
context = {"labels": {"instance": "192.168.0.110"}, "incident_id": "INC-123"}
|
||
→ {"host": "192.168.0.110", "lines": 50}
|
||
"""
|
||
if isinstance(template, dict):
|
||
return {k: _resolve_template(v, context) for k, v in template.items()}
|
||
if isinstance(template, list):
|
||
return [_resolve_template(v, context) for v in template]
|
||
if isinstance(template, str) and "{" in template:
|
||
# 找出所有 {xxx} placeholder 並替換
|
||
import re
|
||
def _repl(m: re.Match) -> str:
|
||
key = m.group(1)
|
||
val = _lookup_context(key, context)
|
||
return str(val) if val is not None else m.group(0)
|
||
return re.sub(r"\{([a-zA-Z0-9_.\[\]]+)\}", _repl, template)
|
||
return template
|
||
|
||
|
||
def _lookup_context(key: str, context: dict) -> Any:
|
||
"""
|
||
從 context 查表（支援巢狀 key: labels.instance / signals[0].alert_name）
|
||
"""
|
||
parts = key.replace("[", ".").replace("]", "").split(".")
|
||
cur: Any = context
|
||
for part in parts:
|
||
if part == "":
|
||
continue
|
||
if isinstance(cur, dict):
|
||
cur = cur.get(part)
|
||
elif isinstance(cur, list):
|
||
try:
|
||
cur = cur[int(part)]
|
||
except (ValueError, IndexError):
|
||
return None
|
||
else:
|
||
return None
|
||
if cur is None:
|
||
return None
|
||
return cur
|
||
|
||
|
||
# =============================================================================
|
||
# Dispatcher (Sprint 5.1)
|
||
# =============================================================================
|
||
|
||
|
||
async def dispatch_action(
|
||
action_name: str,
|
||
incident_id: str,
|
||
user_id: int | None = None,
|
||
labels: dict | None = None,
|
||
extra_context: dict | None = None,
|
||
) -> DispatchResult:
|
||
"""
|
||
執行 callback action — 依 spec 呼叫 MCP tool
|
||
|
||
Args:
|
||
action_name: action 名稱（對應 spec registry）
|
||
incident_id: 關聯 incident
|
||
user_id: Telegram user id（callback 來源）
|
||
labels: alert labels（供模板替換）
|
||
extra_context: 額外上下文（signals 等）
|
||
|
||
Returns:
|
||
DispatchResult（包含 result_text 供 reply 使用）
|
||
"""
|
||
start = time.perf_counter()
|
||
spec = get_action_spec(action_name)
|
||
if not spec:
|
||
logger.warning("dispatch_action_unknown", action=action_name)
|
||
return DispatchResult(
|
||
success=False,
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
user_id=user_id,
|
||
result_text="",
|
||
error=f"Unknown action: {action_name}",
|
||
duration_ms=(time.perf_counter() - start) * 1000,
|
||
)
|
||
|
||
# 建立模板 context
|
||
context = {
|
||
"incident_id": incident_id,
|
||
"labels": labels or {},
|
||
"callback": {"user_id": user_id or 0},
|
||
**(extra_context or {}),
|
||
}
|
||
resolved_params = _resolve_template(spec.mcp_params, context)
|
||
|
||
# Audit log (all actions)
|
||
logger.info(
|
||
"dispatch_action_start",
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
user_id=user_id,
|
||
risk=spec.risk,
|
||
provider=spec.mcp_provider,
|
||
tool=spec.mcp_tool,
|
||
params=resolved_params,
|
||
)
|
||
|
||
# MCP 呼叫（Sprint 5.1 只實作骨架，實際 MCP dispatch 在 Sprint 5.2/5.3）
|
||
try:
|
||
# Sprint 5.1 TODO: 接入 MCP registry
|
||
# from src.plugins.mcp.registry import get_provider
|
||
# provider = get_provider(spec.mcp_provider)
|
||
# import asyncio
|
||
# mcp_result = await asyncio.wait_for(
|
||
# provider.execute(spec.mcp_tool, resolved_params),
|
||
# timeout=spec.timeout_sec,
|
||
# )
|
||
# result_text = _format_reply(mcp_result, spec.reply_format, spec.label, spec.emoji)
|
||
|
||
# Sprint 5.0 骨架：僅返回「尚未實作」提示
|
||
result_text = (
|
||
f"{spec.emoji} <b>{spec.label}</b>\n"
|
||
f"⚠️ Phase 5 Sprint 5.2+ 實作中（spec 已載入：{spec.mcp_provider}/{spec.mcp_tool}）\n"
|
||
f"參數: <code>{resolved_params}</code>"
|
||
)
|
||
duration = (time.perf_counter() - start) * 1000
|
||
logger.info(
|
||
"dispatch_action_stub",
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
duration_ms=round(duration, 1),
|
||
)
|
||
return DispatchResult(
|
||
success=True,
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
user_id=user_id,
|
||
result_text=result_text,
|
||
duration_ms=duration,
|
||
)
|
||
except Exception as e:
|
||
duration = (time.perf_counter() - start) * 1000
|
||
logger.error(
|
||
"dispatch_action_failed",
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
error=str(e),
|
||
duration_ms=round(duration, 1),
|
||
)
|
||
return DispatchResult(
|
||
success=False,
|
||
action=action_name,
|
||
incident_id=incident_id,
|
||
user_id=user_id,
|
||
result_text=f"{spec.emoji} {spec.label} 執行失敗",
|
||
error=str(e),
|
||
duration_ms=duration,
|
||
)
|
||
|
||
|
||
def _format_reply(
|
||
mcp_result: Any, reply_format: str, label: str, emoji: str
|
||
) -> str:
|
||
"""
|
||
依 spec 格式化 reply 文字。
|
||
|
||
reply_format:
|
||
- text: 單行文字
|
||
- code: <code>...</code>
|
||
- truncated: 截斷到 500 字
|
||
- url: 直接返回 URL
|
||
"""
|
||
header = f"{emoji} <b>{label}</b>"
|
||
if reply_format == "url":
|
||
return f"{header}\n{mcp_result}"
|
||
if reply_format == "code":
|
||
return f"{header}\n<code>{str(mcp_result)[:800]}</code>"
|
||
if reply_format == "truncated":
|
||
text = str(mcp_result)[:500]
|
||
if len(str(mcp_result)) > 500:
|
||
text += "...\n<i>(已截斷)</i>"
|
||
return f"{header}\n<pre>{text}</pre>"
|
||
return f"{header}\n{mcp_result}"
|