{"title":"自愈测试流水线","slug":"self-healing-tests","category":"guides","summary":"使用 Synapse 记忆构建能从失败中学习并自动适应的测试流水线。","audience":["human","llm"],"tags":["guide","testing","self-healing","ci"],"difficulty":"advanced","updated":"2026-06-27","word_count":126,"read_minutes":1,"lang":"zh","translated":true,"requested_lang":"zh","content_markdown":"\n# 自愈测试流水线\n\n传统测试套件在 UI 变化时会失败。自愈测试使用 Synapse 记忆从过往失败中学习并适应 — 减少不稳定测试与维护成本。\n\n## 概念\n\n```\n┌─────────┐ 失败 ┌──────────┐ 存储 ┌──────────┐\n│ 测试 │ ───────▶ │ Synapse │ ───────▶ │ 记忆 │\n│ 运行 │ │ 记忆 │ │ (失败) │\n└─────────┘ └──────────┘ └──────────┘\n ▲ │\n │ 回放 │\n │ 下次运行前 │\n └─────────────────────┘\n```\n\n1. 测试运行\n2. 若失败，存储失败信息（出了什么错、为什么、如何修）\n3. 下次运行：执行前回放相关失败\n4. 自动应用已知修复\n\n## 实现\n\n### 第 1 步：测试包装器\n\n为每个测试包装记忆回放/存储：\n\n```python\nimport requests\nfrom datetime import datetime\n\nURL = \"https://synapse.schaefer.zone\"\nMIND_KEY = \"mk_...\"\n\ndef self_healing_test(test_name, test_fn):\n \"\"\"装饰器：用自愈记忆包装测试。\"\"\"\n def wrapper():\n # 1. 回放该测试的过往失败\n past_failures = requests.get(\n f\"{URL}/memory/search?q={test_name}+failure\",\n headers={\"Authorization\": f\"Bearer {MIND_KEY}\"}\n ).json()\n \n # 2. 在失败上下文中运行测试\n try:\n test_fn(known_failures=past_failures)\n except Exception as e:\n # 3. 存储失败\n store_failure(test_name, e, traceback.format_exc())\n raise\n \n return wrapper\n\ndef store_failure(test_name, error, traceback_str):\n requests.post(f\"{URL}/memory\",\n headers={\"Authorization\": f\"Bearer {MIND_KEY}\",\n \"Content-Type\": \"application/json\"},\n json={\n \"category\": \"mistake\",\n \"key\": f\"test_failure_{test_name}_{datetime.now().isoformat()}\",\n \"content\": f\"Test: {test_name}\\nError: {error}\\nTrace:\\n{traceback_str}\",\n \"tags\": [\"test\", \"failure\", test_name],\n \"priority\": \"high\"\n })\n```\n\n### 第 2 步：自适应测试逻辑\n\n在测试内部检查已知失败并应用修复：\n\n```python\n@self_healing_test\ndef test_login_page(browser, known_failures=None):\n browser.goto(\"https://app.com/login\")\n \n # 检查是否之前见过这种页面变化\n if known_failures and known_failures.get(\"results\"):\n for failure in known_failures[\"results\"]:\n if \"button moved\" in failure[\"content\"].lower():\n # 改用 accessibility 标签而非坐标\n browser.click(by_label=\"Login button\")\n return\n \n # 默认：使用坐标\n browser.click(x=150, y=400)\n```\n\n### 第 3 步：恢复策略\n\n把恢复策略作为记忆存储：\n\n```python\ndef store_recovery(failure_type, strategy):\n requests.post(f\"{URL}/memory\",\n headers={\"Authorization\": f\"Bearer {MIND_KEY}\",\n \"Content-Type\": \"application/json\"},\n json={\n \"category\": \"skill\",\n \"key\": f\"recovery_{failure_type}\",\n \"content\": strategy,\n \"tags\": [\"test\", \"recovery\", failure_type],\n \"priority\": \"high\"\n })\n\n# 为常见失败存储恢复策略\nstore_recovery(\"element_not_found\",\n \"When element not found by ID, try by CSS class, then by XPath, \"\n \"then by accessibility label. Take screenshot for debugging.\")\n\nstore_recovery(\"timeout\",\n \"Increase timeout to 30s. If still fails, check if page is loading \"\n \"dynamically — wait for specific element instead of fixed time.\")\n\nstore_recovery(\"stale_element\",\n \"Re-find element before each interaction. Don't cache element references \"\n \"across page transitions.\")\n```\n\n### 第 4 步：CI 集成\n\n```yaml\n# .gitlab-ci.yml\ntest:self-healing:\n script:\n - export SYNAPSE_MIND_KEY=$SYNAPSE_TEST_MIND_KEY\n - pytest tests/ --self-healing\n after_script:\n # 汇总新失败\n - python scripts/synapse_failure_summary.py\n```\n\n### 第 5 步：失败分析仪表板\n\n```python\n# 获取最近一周的所有测试失败\nr = requests.get(\n f\"{URL}/memory/search?q=test+failure\",\n headers={\"Authorization\": f\"Bearer {MIND_KEY}\"}\n)\n\n# 按测试名分组\nfailures = {}\nfor mem in r.json().get(\"results\", []):\n test_name = extract_test_name(mem[\"content\"])\n failures.setdefault(test_name, []).append(mem)\n\n# 报告\nfor test, fails in sorted(failures.items(), key=lambda x: -len(x[1])):\n print(f\"{test}: {len(fails)} failures\")\n```\n\n## 最佳实践\n\n> [!TIP]\n> - **存储 traceback** — 它们包含失败的确切行\n> - **按测试名打标签** — 加速过滤\n> - **使用 `mistake` 分类** — 与常规记忆分离\n> - **设为 `high` 优先级** — 失败绝不能被遗忘\n> - **定期清理** — 删除已解决问题的记忆\n> - **不要存储敏感数据** — 凭据、PII\n\n## 常见失败模式存储\n\n| 失败类型 | 存储内容 |\n|--------------|---------------|\n| 元素未找到 | 尝试的选择器、页面状态、截图 |\n| 超时 | 等待时间、等待的对象 |\n| 断言失败 | 期望值与实际值 |\n| 网络错误 | URL、状态码、响应体 |\n| 权限被拒 | 所需权限、当前用户角色 |\n\n## 下一步\n\n- [iOS 自动化测试](/docs/guides/automated-testing-ios)\n- [记忆最佳实践](/docs/guides/memory-best-practices)\n- [错误恢复 Cookbook](/docs/llm-cookbook/error-recovery)\n","content_html":"

自愈测试流水线

\n

传统测试套件在 UI 变化时会失败。自愈测试使用 Synapse 记忆从过往失败中学习并适应 — 减少不稳定测试与维护成本。

\n

概念

\n
┌─────────┐  失败   ┌──────────┐  存储   ┌──────────┐\n│  测试   │ ───────▶ │ Synapse  │ ───────▶ │  记忆    │\n│  运行   │          │  记忆    │          │ (失败)   │\n└─────────┘          └──────────┘          └──────────┘\n                           ▲                     │\n                           │   回放              │\n                           │  下次运行前         │\n                           └─────────────────────┘
    \n
  1. 测试运行
    2. \n
  2. 若失败，存储失败信息（出了什么错、为什么、如何修）
    3. \n
  3. 下次运行：执行前回放相关失败
    4. \n
  4. 自动应用已知修复
    5. \n
\n

实现

\n

第 1 步：测试包装器

\n

为每个测试包装记忆回放/存储：

\n
import requests\nfrom datetime import datetime\n\nURL = "https://synapse.schaefer.zone"\nMIND_KEY = "mk_..."\n\ndef self_healing_test(test_name, test_fn):\n    """装饰器：用自愈记忆包装测试。"""\n    def wrapper():\n        # 1. 回放该测试的过往失败\n        past_failures = requests.get(\n            f"{URL}/memory/search?q={test_name}+failure",\n            headers={"Authorization": f"Bearer {MIND_KEY}"}\n        ).json()\n        \n        # 2. 在失败上下文中运行测试\n        try:\n            test_fn(known_failures=past_failures)\n        except Exception as e:\n            # 3. 存储失败\n            store_failure(test_name, e, traceback.format_exc())\n            raise\n    \n    return wrapper\n\ndef store_failure(test_name, error, traceback_str):\n    requests.post(f"{URL}/memory",\n        headers={"Authorization": f"Bearer {MIND_KEY}",\n                 "Content-Type": "application/json"},\n        json={\n            "category": "mistake",\n            "key": f"test_failure_{test_name}_{datetime.now().isoformat()}",\n            "content": f"Test: {test_name}\\nError: {error}\\nTrace:\\n{traceback_str}",\n            "tags": ["test", "failure", test_name],\n            "priority": "high"\n        })

第 2 步：自适应测试逻辑

\n

在测试内部检查已知失败并应用修复：

\n
@self_healing_test\ndef test_login_page(browser, known_failures=None):\n    browser.goto("https://app.com/login")\n    \n    # 检查是否之前见过这种页面变化\n    if known_failures and known_failures.get("results"):\n        for failure in known_failures["results"]:\n            if "button moved" in failure["content"].lower():\n                # 改用 accessibility 标签而非坐标\n                browser.click(by_label="Login button")\n                return\n    \n    # 默认：使用坐标\n    browser.click(x=150, y=400)

第 3 步：恢复策略

\n

把恢复策略作为记忆存储：

\n
def store_recovery(failure_type, strategy):\n    requests.post(f"{URL}/memory",\n        headers={"Authorization": f"Bearer {MIND_KEY}",\n                 "Content-Type": "application/json"},\n        json={\n            "category": "skill",\n            "key": f"recovery_{failure_type}",\n            "content": strategy,\n            "tags": ["test", "recovery", failure_type],\n            "priority": "high"\n        })\n\n# 为常见失败存储恢复策略\nstore_recovery("element_not_found",\n    "When element not found by ID, try by CSS class, then by XPath, "\n    "then by accessibility label. Take screenshot for debugging.")\n\nstore_recovery("timeout",\n    "Increase timeout to 30s. If still fails, check if page is loading "\n    "dynamically — wait for specific element instead of fixed time.")\n\nstore_recovery("stale_element",\n    "Re-find element before each interaction. Don't cache element references "\n    "across page transitions.")

第 4 步：CI 集成

\n
# .gitlab-ci.yml\ntest:self-healing:\n  script:\n    - export SYNAPSE_MIND_KEY=$SYNAPSE_TEST_MIND_KEY\n    - pytest tests/ --self-healing\n  after_script:\n    # 汇总新失败\n    - python scripts/synapse_failure_summary.py

第 5 步：失败分析仪表板

\n
# 获取最近一周的所有测试失败\nr = requests.get(\n    f"{URL}/memory/search?q=test+failure",\n    headers={"Authorization": f"Bearer {MIND_KEY}"}\n)\n\n# 按测试名分组\nfailures = {}\nfor mem in r.json().get("results", []):\n    test_name = extract_test_name(mem["content"])\n    failures.setdefault(test_name, []).append(mem)\n\n# 报告\nfor test, fails in sorted(failures.items(), key=lambda x: -len(x[1])):\n    print(f"{test}: {len(fails)} failures")

最佳实践

\n

常见失败模式存储

\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n
失败类型存储内容
元素未找到尝试的选择器、页面状态、截图
超时等待时间、等待的对象
断言失败期望值与实际值
网络错误URL、状态码、响应体
权限被拒所需权限、当前用户角色
\n

下一步

\n\n","urls":{"html":"/docs/guides/self-healing-tests","text":"/docs/guides/self-healing-tests?format=text","json":"/docs/guides/self-healing-tests?format=json","llm":"/docs/guides/self-healing-tests?format=llm"},"translations_available":["en","zh","hi","es","fr","ar","pt","ru","ja","de","it","ko","nl","pl","tr","sv","vi","th","id","uk"]}