awoooi/verify_telegram_ui.py

"""
verify_telegram_ui.py — Telegram UI 修復驗證腳本
================================================
注入 800 字極端字串，驗證：
1. _smart_truncate 在句子邊界截斷，不破壞括號
2. _parse_debate_summary 正確拆分各欄位（不重複）
3. TYPE-3 requires_human_approval=True → 含批准/拒絕按鈕

2026-04-17 ogt + Claude Sonnet 4.6 (ADR-075 UI 修復驗證)
"""
import sys
sys.path.insert(0, "apps/api")

# ─── 測試 1：smart_truncate ──────────────────────────────────────────────────

from src.services.telegram_gateway import _smart_truncate

# 800 字多層括號測試字串
LONG_REASONING = (
    "診斷：根據告警訊號分析，發現 MoWoooWorkDown 事件導致服務下線，"
    "可能是由於 deployment 配置錯誤或是 pod 問題引起的（信心 90%，系統正常）；"
    "方案：kubectl rollout restart deployment/awoooi-api -n awoooi-prod"
    "（blast_radius=25，rollback_cost=5，降級風險極低）；"
    "安全審查：approve（blast_radius 符合安全閾值 ≤50，靜態規則通過，系統正常）；"
    "質疑：無（通過審查，所有指標在正常範圍內，無需人工干預，建議自動執行）"
    "額外備注：此次分析基於最近 15 分鐘的 Prometheus 指標窗口，"
    "包含 CPU 使用率、記憶體壓力、網路 I/O 三個維度的複合評估（樣本數 N=1440）。"
    "補充說明：若下次相同告警在 30 分鐘內再次出現，建議升級至 P1 並通知值班主管。"
)

print("=" * 60)
print("TEST 1: _smart_truncate")
print("=" * 60)
print(f"原始長度: {len(LONG_REASONING)} 字")
print()

for limit in [100, 200, 300, 500]:
    result = _smart_truncate(LONG_REASONING, limit)
    # 驗證括號平衡
    open_p = result.count("（")
    close_p = result.count("）")
    bracket_ok = open_p == close_p
    print(f"limit={limit}: len={len(result)} 括號平衡={bracket_ok} (（={open_p}, ）={close_p})")
    print(f"  結尾: ...{result[-30:]}")
    print()

# ─── 測試 2：_parse_debate_summary ──────────────────────────────────────────

# 在 decision_manager 中定義（複製相同邏輯做驗證）
def _parse_debate_summary(reasoning: str) -> dict:
    result = {"diagnosis": "", "plan": "", "review": "", "critic": ""}
    for part in reasoning.split("；"):
        part = part.strip()
        if part.startswith("診斷："):
            result["diagnosis"] = part[3:]
        elif part.startswith("方案："):
            result["plan"] = part[3:]
        elif part.startswith("安全審查："):
            result["review"] = part[5:]
        elif part.startswith("質疑："):
            result["critic"] = part[3:]
    return result

print("=" * 60)
print("TEST 2: _parse_debate_summary（各欄位不可重複）")
print("=" * 60)
parsed = _parse_debate_summary(LONG_REASONING)
for key, val in parsed.items():
    print(f"  {key}: {val[:80]}{'...' if len(val) > 80 else ''}")

print()
print("✅ 各欄位均不同（修復重複渲染）:")
vals = [v for v in parsed.values() if v]
all_different = len(vals) == len(set(vals))
print(f"  all_different = {all_different}")

# 模擬 TYPE-8M 卡片渲染
print()
print("── TYPE-8M 卡片預覽 ──")
_diag = _smart_truncate(parsed["diagnosis"] or "（無診斷）", 120)
_impact = _smart_truncate(parsed["plan"] or "", 150)
_cause = _smart_truncate(parsed["critic"] or parsed["review"] or "", 100)
print(f"🎯 診斷結果：{_diag}")
if _impact:
    print(f"🧠 系統影響")
    print(f"   {_impact}")
if _cause:
    print(f"└─ 可能根因：{_cause}")

# ─── 測試 3：requires_human_approval 按鈕邏輯 ───────────────────────────────

print()
print("=" * 60)
print("TEST 3: requires_human_approval → 動態按鈕含批准/拒絕")
print("=" * 60)

# 模擬 callback_dispatcher 回傳 k8s 動態按鈕
MOCK_K8S_BUTTONS = [
    ("🔄 重啟", "restart:INC-001"),
    ("⬆️ 擴容", "scale_up:INC-001"),
    ("⬇️ 縮容", "scale_down:INC-001"),
    ("🔙 回滾", "rollback:INC-001"),
]

def simulate_keyboard(dynamic_buttons: list, requires_human_approval: bool) -> list:
    is_type3 = True
    approve_nonce = "approve-nonce-xxx"
    reject_nonce = "reject-nonce-xxx"
    silence_nonce = "silence-nonce-xxx"

    if is_type3 and dynamic_buttons:
        btns = [{"text": t, "callback_data": cb} for t, cb in dynamic_buttons]
        rows = [btns[i:i+3] for i in range(0, len(btns), 3)]
        if requires_human_approval:
            rows.append([
                {"text": "✅ 批准", "callback_data": approve_nonce},
                {"text": "❌ 拒絕", "callback_data": reject_nonce},
            ])
        rows.append([
            {"text": "📋 詳情", "callback_data": "detail:INC-001"},
            {"text": "🔕 忽略", "callback_data": silence_nonce},
        ])
        return rows
    return [[
        {"text": "✅ 批准", "callback_data": approve_nonce},
        {"text": "❌ 拒絕", "callback_data": reject_nonce},
        {"text": "🔕 靜默", "callback_data": silence_nonce},
    ]]

print()
print("場景 A: requires_human_approval=False（無動態按鈕卡）")
kb_a = simulate_keyboard([], False)
for row in kb_a:
    print("  " + " | ".join(b["text"] for b in row))

print()
print("場景 B: requires_human_approval=False + k8s 動態按鈕（舊 bug：死卡）")
kb_b = simulate_keyboard(MOCK_K8S_BUTTONS, False)
for row in kb_b:
    print("  " + " | ".join(b["text"] for b in row))
has_approve_b = any(b["text"] == "✅ 批准" for row in kb_b for b in row)
print(f"  含批准按鈕: {has_approve_b} ← 舊 bug = False（死卡）")

print()
print("場景 C: requires_human_approval=True + k8s 動態按鈕（新修復）")
kb_c = simulate_keyboard(MOCK_K8S_BUTTONS, True)
for row in kb_c:
    print("  " + " | ".join(b["text"] for b in row))
has_approve_c = any(b["text"] == "✅ 批准" for row in kb_c for b in row)
print(f"  含批准按鈕: {has_approve_c} ← 修復後 = True ✅")

print()
print("=" * 60)
print("SUMMARY")
print("=" * 60)
t1 = not any("（" in _smart_truncate(LONG_REASONING, l) and "）" not in _smart_truncate(LONG_REASONING, l)
             for l in [100, 200, 300, 500])
t2 = all_different
t3 = has_approve_c and not has_approve_b
print(f"TEST 1 smart_truncate 括號不破壞: {'✅' if t1 else '❌'}")
print(f"TEST 2 parse_debate 各欄位不重複: {'✅' if t2 else '❌'}")
print(f"TEST 3 requires_human→批准按鈕:  {'✅' if t3 else '❌'}")
if all([t1, t2, t3]):
    print("\n🎉 全部通過！UI 修復驗證完成。")
else:
    print("\n❌ 有測試未通過，請檢查。")
    sys.exit(1)