zhaolei
2f9f8da48e
Some checks failed
AI automation / Route event (push) Has been cancelled
AI automation / Hand reopened issue to maintainers (push) Has been cancelled
AI automation / Clean source issue state (push) Has been cancelled
AI automation / Reconcile handoffs (push) Has been cancelled
AI automation / Classify issue (push) Has been cancelled
AI automation / Claude Code smoke (push) Has been cancelled
AI automation / Review issue follow-up (push) Has been cancelled
AI automation / Publish issue follow-up (push) Has been cancelled
AI automation / Implement with Claude Code (push) Has been cancelled
AI automation / Publish implement PR (push) Has been cancelled
AI automation / Continue queued issue comments (push) Has been cancelled
AI automation / Codex review loop (push) Has been cancelled
AI automation / Publish Codex fix (push) Has been cancelled
AI automation / Clear Codex dispatch marker (push) Has been cancelled
AI automation / Own PR re-request Codex (push) Has been cancelled
AI automation / External PR re-request Codex (push) Has been cancelled
AI automation / Poll Codex reaction / retry (push) Has been cancelled
330 lines
12 KiB
TypeScript
330 lines
12 KiB
TypeScript
"use client";
|
||
|
||
/**
|
||
* 巡检报告 AI 汇总
|
||
*
|
||
* 复用 RCA runner 的 AI provider 读取（localStorage netcatty_ai_providers_v1）
|
||
* 通过 window.netcatty.aiFetch bridge 发 HTTP 请求到用户配置的 LLM endpoint。
|
||
*
|
||
* Prompt 四段式输出，JSON 结构化：
|
||
* { health_summary, anomalies[], trend_notes[], recommendations[] }
|
||
*/
|
||
|
||
type ProviderInfo = {
|
||
providerId: string;
|
||
apiKey: string;
|
||
baseUrl: string;
|
||
model: string;
|
||
style: "openai" | "google" | "anthropic";
|
||
};
|
||
|
||
const AI_PROVIDERS_KEY = "netcatty_ai_providers_v1";
|
||
const AI_ACTIVE_PROVIDER_KEY = "netcatty_ai_active_provider_v1";
|
||
|
||
async function resolveProvider(): Promise<ProviderInfo | null> {
|
||
let providers: any[] | null = null;
|
||
let activeId: string | null = null;
|
||
try {
|
||
const raw = localStorage.getItem(AI_PROVIDERS_KEY);
|
||
if (raw) providers = JSON.parse(raw);
|
||
activeId = localStorage.getItem(AI_ACTIVE_PROVIDER_KEY);
|
||
} catch (_) {}
|
||
if (!Array.isArray(providers) || providers.length === 0) return null;
|
||
|
||
let primary: any = null;
|
||
if (activeId) primary = providers.find((p) => p.id === activeId);
|
||
if (!primary) primary = providers.find((p) => p.enabled !== false);
|
||
if (!primary?.apiKey) return null;
|
||
|
||
// 同步 providers 到主进程（RCA 路径也需要）
|
||
const api = (window as any).netcatty;
|
||
if (api?.aiSyncProviders) {
|
||
try { await api.aiSyncProviders(providers); } catch (_) {}
|
||
}
|
||
|
||
const lower = (primary.baseURL || primary.providerId || "").toLowerCase();
|
||
const style = /anthropic|claude/.test(lower) ? "anthropic"
|
||
: /google|gemini/.test(lower) ? "google"
|
||
: "openai";
|
||
|
||
const suffix = style === "anthropic" ? "/v1/messages"
|
||
: style === "google" ? "/v1beta/models"
|
||
: "/chat/completions";
|
||
|
||
return {
|
||
providerId: primary.id || "default",
|
||
apiKey: primary.apiKey,
|
||
baseUrl: (primary.baseURL || "").replace(/\/$/, "") + suffix,
|
||
model: primary.defaultModel || "gpt-4o-mini",
|
||
style,
|
||
};
|
||
}
|
||
|
||
async function aiChatCompletion(
|
||
provider: ProviderInfo,
|
||
systemPrompt: string,
|
||
userContent: string,
|
||
): Promise<string> {
|
||
const api = (window as any).netcatty;
|
||
if (!api?.aiFetch) throw new Error("aiFetch bridge unavailable");
|
||
|
||
const b = (window as any).netcatty;
|
||
if (b?.aiAllowlistAddHost) {
|
||
try { await b.aiAllowlistAddHost(provider.baseUrl); } catch (_) {}
|
||
}
|
||
|
||
let url = provider.baseUrl;
|
||
const body: any = {
|
||
model: provider.model,
|
||
temperature: 0.3,
|
||
};
|
||
|
||
// 用 placeholder — 主进程 injectApiKeyIntoRequest 会根据 providerId 解密注入真实 key
|
||
const headers: Record<string, string> = {
|
||
"Content-Type": "application/json",
|
||
"Authorization": "Bearer __IPC_SECURED__",
|
||
};
|
||
|
||
if (provider.style === "google") {
|
||
url = `${provider.baseUrl}/${provider.model}:generateContent`;
|
||
delete headers.Authorization;
|
||
headers["x-goog-api-key"] = "__IPC_SECURED__";
|
||
body.contents = [{
|
||
role: "user",
|
||
parts: [{ text: `${systemPrompt}\n\n${userContent}` }],
|
||
}];
|
||
} else if (provider.style === "anthropic") {
|
||
headers["x-api-key"] = "__IPC_SECURED__";
|
||
headers["anthropic-version"] = "2023-06-01";
|
||
delete headers.Authorization;
|
||
body.model = provider.model;
|
||
body.system = systemPrompt;
|
||
body.messages = [{ role: "user", content: userContent }];
|
||
body.max_tokens = 1024;
|
||
} else {
|
||
body.messages = [
|
||
{ role: "system", content: systemPrompt },
|
||
{ role: "user", content: userContent },
|
||
];
|
||
}
|
||
|
||
// timeoutMs=60s — prompt 已压缩（只给失败+resource/log 输出），正常 30s 内返回
|
||
const resp: any = await api.aiFetch(url, "POST", headers, JSON.stringify(body), provider.providerId, true, false, false, 60000);
|
||
if (!resp?.ok) {
|
||
const msg = resp?.error ? `${resp.error} (HTTP ${resp.status})` : `HTTP ${resp?.status || "?"}`;
|
||
throw new Error(`AI 调用失败: ${msg}`);
|
||
}
|
||
|
||
const data = resp.data ? JSON.parse(resp.data) : {};
|
||
if (provider.style === "google") {
|
||
return data?.candidates?.[0]?.content?.parts?.[0]?.text || "";
|
||
} else if (provider.style === "anthropic") {
|
||
return data?.content?.[0]?.text || "";
|
||
}
|
||
return data?.choices?.[0]?.message?.content || "";
|
||
}
|
||
|
||
function extractJson(text: string): any {
|
||
// 去掉 ```json ... ``` 包裹
|
||
let t = text.trim();
|
||
t = t.replace(/^```json\s*/i, "").replace(/\s*```\s*$/, "");
|
||
// 找最外层 { ... }
|
||
const m = t.match(/\{[\s\S]*\}/);
|
||
if (!m) return null;
|
||
try { return JSON.parse(m[0]); } catch { return null; }
|
||
}
|
||
|
||
// ─── Prompt ────────────────────────────────────
|
||
|
||
const SYSTEM_PROMPT = `你是一名资深运维工程师，负责汇总服务器/交换机/防火墙的巡检报告。
|
||
数据中提供了每台设备执行各类巡检命令的原始输出，请观察这些细节数据，识别隐藏风险和运行状态。
|
||
|
||
输出 JSON：
|
||
|
||
{
|
||
"health_summary": "一句话整体评估（X 台在线 / Y 台离线，命令成功率 %，关键发现一句话）",
|
||
"offline_devices": [ { "device": "hostname", "reason": "SSH连接超时|认证失败" } ],
|
||
"device_status": [
|
||
{
|
||
"device": "hostname",
|
||
"catalog": "huawei_h3c_switch",
|
||
"running_ok": true,
|
||
"highlights": ["关键观察1", "关键观察2"],
|
||
"issues": ["失败的命令或隐藏风险"]
|
||
}
|
||
],
|
||
"hidden_risks": [
|
||
{ "device": "hostname", "observation": "具体数值和异常点", "severity": "critical|warning|info" }
|
||
],
|
||
"recommendations": ["按优先级排序的处置建议，必须具体到设备和操作"]
|
||
}
|
||
|
||
分析要点：
|
||
- 观察温度数值：正常交换机 40-55°C，超过 60°C 值得关注
|
||
- 观察 CPU/内存使用率：单台持续 >75% 要标 warning
|
||
- 观察风扇转速：异常偏高（如 4500 RPM 但正常 3000）可能预示轴承磨损
|
||
- 观察端口状态：CRC error、input error 非零 = 物理链路问题
|
||
- 观察 NTP：stratum 变化、peer 切换频繁 = 时钟源不稳定
|
||
- 观察 display logbuffer：找出 WARNING/CRITICAL 级别的日志条目作为隐藏风险
|
||
- vrrp/stp 角色异常（非预期的 Master/Backup 切换、指定 STP 阻塞端口）= 网络拓扑变更
|
||
- Linux 服务器的 load average / /proc/meminfo / ip link 状态也要关注
|
||
- Linux journalctl -p warning 输出的是 warning 以上日志，逐条解读
|
||
|
||
hidden_risks 用于放那些"命令没失败但数值异常"的情况（如 display temperature 成功但读到 65°C）
|
||
device_status 用于列出每台设备的简要运行概览（亮点 + 问题）
|
||
严格输出合法 JSON，不要额外解释`;
|
||
|
||
/**
|
||
* 按命令类型分层保留 stdout：
|
||
* 基础信息/资源指标 → 完整保留（短，是核心）
|
||
* 网络状态 → 前 800 字
|
||
* 日志类 → 最后 30 行（tail）
|
||
* 失败命令 → 完整保留错误输出
|
||
*/
|
||
function _classifyCmd(cmdName: string): "base" | "resource" | "network" | "log" {
|
||
const n = cmdName.toLowerCase();
|
||
if (/version|device|hostname|uptime|uname/.test(n)) return "base";
|
||
if (/memory|cpu|temperature|fan|power|free -m|vmstat|top -bn1|journalctl -p/.test(n)) return "resource";
|
||
if (/interface|arp|vlan|stp|vrrp|ntp|route|neigh|ip -br|ip link|ip addr/.test(n)) return "network";
|
||
if (/logbuffer|journalctl|dmesg|messages|syslog|show log/.test(n)) return "log";
|
||
return "network"; // 未知默认中等长度
|
||
}
|
||
|
||
function _truncateStdout(cmdName: string, stdout: string): string {
|
||
if (!stdout) return "";
|
||
const cls = _classifyCmd(cmdName);
|
||
if (cls === "log") {
|
||
// 日志取最后 30 行
|
||
const lines = stdout.split("\n");
|
||
if (lines.length > 30) return lines.slice(-30).join("\n");
|
||
return stdout;
|
||
}
|
||
if (cls === "network") {
|
||
return stdout.slice(0, 800);
|
||
}
|
||
// base / resource 完整保留（一般 < 30 行）
|
||
return stdout.slice(0, 2000); // 兜底防极端
|
||
}
|
||
|
||
function buildSnapshotUserContext(snapshot: any): string {
|
||
const offlineDevices: any[] = [];
|
||
const onlineDevices: any[] = [];
|
||
|
||
for (const d of snapshot.devices) {
|
||
const allFail = d.commands.every((c: any) => !c.ok);
|
||
const isConnectionErr = d.commands.every((c: any) =>
|
||
/Connection refused|ETIMEDOUT|ECONNREFUSED|connect failed|authentication failed|SSH.*failed/i.test(c.stderr || c.stdout || "")
|
||
);
|
||
const isOffline = !d.ok || (allFail && isConnectionErr);
|
||
|
||
if (isOffline) {
|
||
const reason = d.commands[0]?.stderr?.slice(0, 200)
|
||
|| d.commands[0]?.stdout?.slice(0, 200)
|
||
|| "SSH连接失败";
|
||
offlineDevices.push({ device: d.hostname, reason });
|
||
} else {
|
||
// 在线设备：成功设备只给概览，只给失败命令 + 资源/日志关键命令的截断 stdout
|
||
const detailCmds: any[] = [];
|
||
let hasFail = false;
|
||
for (const c of d.commands) {
|
||
if (!c.ok) {
|
||
hasFail = true;
|
||
detailCmds.push({
|
||
cmd: c.cmd, purpose: c.purpose, status: "fail",
|
||
output: (c.stdout || c.stderr || "").slice(0, 600),
|
||
});
|
||
continue;
|
||
}
|
||
// 成功命令：只保留 resource（CPU/温度/风扇）和 log 类的截断输出
|
||
const cls = _classifyCmd(c.cmd);
|
||
if (cls === "resource") {
|
||
detailCmds.push({
|
||
cmd: c.cmd, purpose: c.purpose, status: "ok",
|
||
output: (c.stdout || "").slice(0, 400),
|
||
});
|
||
} else if (cls === "log") {
|
||
detailCmds.push({
|
||
cmd: c.cmd, purpose: c.purpose, status: "ok",
|
||
output: _truncateStdout(c.cmd, c.stdout || ""),
|
||
});
|
||
}
|
||
// base / network 类成功命令 → **完全不塞**
|
||
}
|
||
|
||
onlineDevices.push({
|
||
device: d.hostname,
|
||
catalog: d.catalog_id,
|
||
status: hasFail ? "partial_fail" : "ok",
|
||
command_total: d.commands.length,
|
||
command_failed: d.commands.filter((c: any) => !c.ok).length,
|
||
commands: detailCmds, // 只有失败命令 + resource/log 成功命令
|
||
});
|
||
}
|
||
}
|
||
|
||
// 总体统计
|
||
const totalOnline = onlineDevices.length;
|
||
const totalOffline = offlineDevices.length;
|
||
const totalCmds = onlineDevices.reduce((a, d) => a + (d as any).command_total, 0);
|
||
const failCmds = onlineDevices.reduce((a, d) => a + (d as any).command_failed, 0);
|
||
|
||
return JSON.stringify({
|
||
timestamp: new Date(snapshot.timestamp).toISOString(),
|
||
trigger: snapshot.trigger,
|
||
duration_ms: snapshot.duration_ms,
|
||
offline_count: totalOffline,
|
||
online_count: totalOnline,
|
||
total_commands: totalCmds,
|
||
failed_commands: failCmds,
|
||
offline_devices: offlineDevices,
|
||
online_devices: onlineDevices,
|
||
}, null, 2);
|
||
}
|
||
|
||
function buildDailyUserContext(report: any): string {
|
||
const s = report.summary;
|
||
const failDevices = Object.entries(s.deviceSuccessRate)
|
||
.filter(([, r]) => (r as number) < 0.95)
|
||
.map(([id, r]) => ({ device: id, successRate: r }));
|
||
return JSON.stringify({
|
||
date: report.date,
|
||
mode: report.mode,
|
||
snapshotCount: s.snapshotCount,
|
||
deviceCount: s.deviceCount,
|
||
totalCmds: s.totalCmds,
|
||
failRate: s.failRate,
|
||
failingDevices: failDevices,
|
||
topFailedCommands: s.topFailedCommands.slice(0, 10),
|
||
}, null, 2);
|
||
}
|
||
|
||
// ─── 主入口 ────────────────────────────────────
|
||
|
||
export async function summarizeSnapshot(snapshot: any): Promise<any> {
|
||
const provider = await resolveProvider();
|
||
if (!provider) {
|
||
throw new Error("AI provider 未配置，请在设置→AI 页面添加 provider");
|
||
}
|
||
const content = buildSnapshotUserContext(snapshot);
|
||
console.log("[ReportSummary] summarizeSnapshot:", {
|
||
providerId: provider.providerId,
|
||
model: provider.model,
|
||
devices: snapshot?.devices?.length,
|
||
contentLen: content.length,
|
||
});
|
||
const t0 = Date.now();
|
||
const raw = await aiChatCompletion(provider, SYSTEM_PROMPT, content);
|
||
console.log("[ReportSummary] AI 返回，耗时：", (Date.now() - t0) / 1000, "s，返回长度：", raw.length);
|
||
return extractJson(raw) || { health_summary: raw, anomalies: [], trend_notes: [], recommendations: [] };
|
||
}
|
||
|
||
export async function summarizeDailyReport(report: any): Promise<any> {
|
||
const provider = await resolveProvider();
|
||
if (!provider) {
|
||
throw new Error("AI provider 未配置，请在设置→AI 页面添加 provider");
|
||
}
|
||
const content = buildDailyUserContext(report);
|
||
const raw = await aiChatCompletion(provider, SYSTEM_PROMPT, content);
|
||
return extractJson(raw) || { health_summary: raw, anomalies: [], trend_notes: [], recommendations: [] };
|
||
}
|