ci: 增强CI失败通知，自动分类失败原因并给出修复建议

scripts/ci_notify_failure.py

@@ -1,17 +1,53 @@
 #!/usr/bin/env python3
-"""发送 CI 失败通知到飞书/项目群 webhook。"""
+"""发送 CI 失败通知到飞书/项目群 webhook（增强版：带失败诊断）。
+
+诊断功能：自动分析失败原因，给出分类和修复建议。
+"""
 
 import json
 import os
 import sys
 import urllib.request
+import subprocess
+
+
+def run_diagnosis() -> dict:
+    """运行失败诊断脚本，返回诊断结果"""
+    diag_script = os.path.join(os.path.dirname(os.path.abspath(__file__)), "ci/ci_failure_diagnosis.py")
+    if not os.path.exists(diag_script):
+        diag_script = "scripts/ci/ci_failure_diagnosis.py"
+    
+    result = {"category": "unknown", "category_cn": "未知", "severity": "medium",
+              "summary": "", "error_lines": [], "suggestions": [], "auto_fixable": False}
+    
+    try:
+        # 运行诊断脚本
+        env = os.environ.copy()
+        env["DIAGNOSIS_OUTPUT"] = "/tmp/ci_diagnosis_result.json"
+        
+        proc = subprocess.run(
+            [sys.executable, diag_script],
+            capture_output=True, text=True, timeout=30, env=env
+        )
+        
+        # 尝试读取结果文件
+        output_file = "/tmp/ci_diagnosis_result.json"
+        if os.path.exists(output_file):
+            with open(output_file) as f:
+                result = json.load(f)
+        elif proc.stdout:
+            # 从stdout解析
+            pass
+    except Exception as e:
+        print(f"诊断脚本执行失败: {e}", file=sys.stderr)
+    
+    return result
 
 
 def main() -> int:
     webhook = os.environ.get("CI_NOTIFY_WEBHOOK", "")
     if not webhook:
         print("未配置 CI_NOTIFY_WEBHOOK，跳过通知")
-        print("如需启用，请在仓库 Settings -> Secrets and variables -> Actions 中添加 CI_NOTIFY_WEBHOOK")
         return 0
 
     failed_job = os.environ.get("FAILED_JOB", "Unknown Job")
@@ -21,6 +57,88 @@ def main() -> int:
     run_id = os.environ.get("GITHUB_RUN_ID", "unknown")
     repo = os.environ.get("GITHUB_REPOSITORY", "unknown")
     run_url = f"https://git.xiaoxiajianji.com/{repo}/actions/runs/{run_id}"
+    pr_number = os.environ.get("PR_NUMBER", "")
+
+    # 运行诊断
+    diagnosis = run_diagnosis()
+    
+    # 构建卡片内容
+    severity_color = {"high": "red", "medium": "orange", "low": "blue"}
+    card_status = severity_color.get(diagnosis.get("severity", "medium"), "red")
+    
+    # 标题
+    title = f"❌ CI失败 - {diagnosis.get('category_cn', '未知')}"
+    
+    # 诊断部分
+    diag_lines = []
+    diag_lines.append(f"**任务**: {failed_job}")
+    diag_lines.append(f"**分类**: {diagnosis.get('category_cn', '未知')}")
+    if diagnosis.get("summary"):
+        diag_lines.append(f"**问题**: {diagnosis['summary']}")
+    
+    # 错误行
+    error_lines = diagnosis.get("error_lines", [])
+    if error_lines:
+        diag_lines.append("")
+        diag_lines.append("**关键错误**:")
+        for err in error_lines[:3]:
+            if len(err) > 100:
+                err = err[:97] + "..."
+            diag_lines.append(f"`{err}`")
+    
+    # 修复建议
+    suggestions = diagnosis.get("suggestions", [])
+    if suggestions:
+        diag_lines.append("")
+        diag_lines.append("**修复建议**:")
+        for i, s in enumerate(suggestions[:3], 1):
+            diag_lines.append(f"{i}. {s}")
+    
+    if diagnosis.get("auto_fixable"):
+        diag_lines.append("")
+        diag_lines.append("💡 *可自动修复的问题，试试Rerun*")
+
+    # 基本信息
+    info_lines = [
+        f"**分支**: {branch}",
+        f"**提交**: `{commit}`",
+        f"**提交者**: {actor}",
+    ]
+    if pr_number:
+        info_lines.append(f"**PR**: #{pr_number}")
+    
+    elements = [
+        {
+            "tag": "div",
+            "text": {
+                "tag": "lark_md",
+                "content": "
+".join(diag_lines),
+            },
+        },
+        {
+            "tag": "hr",
+        },
+        {
+            "tag": "div",
+            "text": {
+                "tag": "lark_md",
+                "content": "
+".join(info_lines),
+            },
+        },
+        {
+            "tag": "action",
+            "actions": [
+                {
+                    "tag": "button",
+                    "text": {"tag": "plain_text", "content": "查看失败日志"},
+                    "url": run_url,
+                    "type": "danger",
+                },
+            ],
+        },
+    ]
 
     payload = {
         "msg_type": "interactive",
@@ -28,36 +146,11 @@ def main() -> int:
             "header": {
                 "title": {
                     "tag": "plain_text",
-                    "content": "❌ CI 构建失败",
+                    "content": title,
                 },
-                "status": "red",
+                "status": card_status,
             },
-            "elements": [
-                {
-                    "tag": "div",
-                    "text": {
-                        "tag": "lark_md",
-                        "content": (
-                            f"**任务**: {failed_job}\n"
-                            f"**分支**: {branch}\n"
-                            f"**提交**: {commit}\n"
-                            f"**提交者**: {actor}\n"
-                            f"**Run ID**: {run_id}"
-                        ),
-                    },
-                },
-                {
-                    "tag": "action",
-                    "actions": [
-                        {
-                            "tag": "button",
-                            "text": {"tag": "plain_text", "content": "查看失败日志"},
-                            "url": run_url,
-                            "type": "danger",
-                        }
-                    ],
-                },
-            ],
+            "elements": elements,
         },
     }
 
@@ -71,7 +164,7 @@ def main() -> int:
     try:
         with urllib.request.urlopen(req, timeout=10) as resp:
             resp.read()
-        print("通知已发送")
+        print("通知已发送（带诊断信息）")
     except Exception as e:
         print(f"通知发送失败: {e}", file=sys.stderr)
         return 1