fix(gateway): fallback health check to CLI for OpenClaw 2026.3

app/api/gateway-health/route.ts

@@ -1,10 +1,69 @@
 import { NextResponse } from "next/server";
 import fs from "fs";
 import path from "path";
+import { execFile } from "child_process";
+import { promisify } from "util";
 
 const OPENCLAW_HOME = process.env.OPENCLAW_HOME || path.join(process.env.HOME || "", ".openclaw");
 const CONFIG_PATH = path.join(OPENCLAW_HOME, "openclaw.json");
 const DEGRADED_LATENCY_MS = 1500;
+const execFileAsync = promisify(execFile);
+
+function parseJsonFromMixedOutput(output: string): any {
+  for (let i = 0; i < output.length; i++) {
+    if (output[i] !== "{") continue;
+    let depth = 0;
+    let inString = false;
+    let escaped = false;
+    for (let j = i; j < output.length; j++) {
+      const ch = output[j];
+      if (inString) {
+        if (escaped) escaped = false;
+        else if (ch === "\\") escaped = true;
+        else if (ch === "\"") inString = false;
+        continue;
+      }
+      if (ch === "\"") {
+        inString = true;
+        continue;
+      }
+      if (ch === "{") depth++;
+      else if (ch === "}") {
+        depth--;
+        if (depth === 0) {
+          const candidate = output.slice(i, j + 1).trim();
+          try {
+            return JSON.parse(candidate);
+          } catch {
+            break;
+          }
+        }
+      }
+    }
+  }
+  throw new Error("Failed to parse JSON from openclaw gateway status output");
+}
+
+async function probeGatewayViaCli(token: string, timeoutMs = 5000): Promise<{ ok: boolean; error?: string }> {
+  try {
+    const args = ["gateway", "status", "--json", "--timeout", String(timeoutMs)];
+    if (token) args.push("--token", token);
+    const { stdout, stderr } = await execFileAsync("openclaw", args, {
+      maxBuffer: 10 * 1024 * 1024,
+      env: { ...process.env, FORCE_COLOR: "0" },
+    });
+    const parsed = parseJsonFromMixedOutput(`${stdout}\n${stderr || ""}`);
+    const ok = parsed?.rpc?.ok === true;
+    const error =
+      typeof parsed?.rpc?.error === "string" && parsed.rpc.error.trim()
+        ? parsed.rpc.error.trim()
+        : undefined;
+    return { ok, error };
+  } catch (err: any) {
+    const message = err?.message || "Failed to run openclaw gateway status";
+    return { ok: false, error: message };
+  }
+}
 
 export async function GET() {
   const startedAt = Date.now();
@@ -21,41 +80,85 @@ export async function GET() {
     const controller = new AbortController();
     const timeout = setTimeout(() => controller.abort(), 5000);
 
-    const resp = await fetch(url, { headers, signal: controller.signal });
+    const resp = await fetch(url, { headers, signal: controller.signal, cache: "no-store" });
     clearTimeout(timeout);
-    const checkedAt = Date.now();
-    const responseMs = checkedAt - startedAt;
-
-    if (!resp.ok) {
+    if (resp.ok) {
+      const checkedAt = Date.now();
+      const responseMs = checkedAt - startedAt;
+      const data = await resp.json().catch(() => null);
       return NextResponse.json({
-        ok: false,
-        error: `HTTP ${resp.status}`,
-        status: "down",
+        ok: true,
+        data,
+        status: responseMs > DEGRADED_LATENCY_MS ? "degraded" : "healthy",
         checkedAt,
         responseMs,
+        webUrl: `http://localhost:${port}/chat${token ? '?token=' + encodeURIComponent(token) : ''}`,
       });
     }
 
-    const data = await resp.json().catch(() => null);
-    return NextResponse.json({
-      ok: true,
-      data,
-      status: responseMs > DEGRADED_LATENCY_MS ? "degraded" : "healthy",
-      checkedAt,
-      responseMs,
-      webUrl: `http://localhost:${port}/chat${token ? '?token=' + encodeURIComponent(token) : ''}`,
-    });
-  } catch (err: any) {
+    // OpenClaw 2026.3.x no longer serves /api/health; fallback to CLI probe.
+    const cli = await probeGatewayViaCli(token, 5000);
     const checkedAt = Date.now();
     const responseMs = checkedAt - startedAt;
-    const msg = err.cause?.code === "ECONNREFUSED"
+    if (cli.ok) {
+      return NextResponse.json({
+        ok: true,
+        data: null,
+        status: "healthy",
+        checkedAt,
+        responseMs,
+        webUrl: `http://localhost:${port}/chat${token ? '?token=' + encodeURIComponent(token) : ''}`,
+      });
+    }
+
+    return NextResponse.json({
+      ok: false,
+      error: cli.error || `HTTP ${resp.status}`,
+      status: "down",
+      checkedAt,
+      responseMs,
+    });
+  } catch (err: any) {
+    // If HTTP probe fails due transport/runtime issues, attempt CLI probe before declaring down.
+    const raw = err.cause?.code === "ECONNREFUSED"
       ? "Gateway 未运行"
       : err.name === "AbortError"
         ? "请求超时"
         : err.message;
+    const token = (() => {
+      try {
+        const rawCfg = fs.readFileSync(CONFIG_PATH, "utf-8");
+        const cfg = JSON.parse(rawCfg);
+        return cfg.gateway?.auth?.token || "";
+      } catch {
+        return "";
+      }
+    })();
+    const cli = await probeGatewayViaCli(token, 5000);
+    const checkedAt = Date.now();
+    const responseMs = checkedAt - startedAt;
+    if (cli.ok) {
+      const port = (() => {
+        try {
+          const rawCfg = fs.readFileSync(CONFIG_PATH, "utf-8");
+          const cfg = JSON.parse(rawCfg);
+          return cfg.gateway?.port || 18789;
+        } catch {
+          return 18789;
+        }
+      })();
+      return NextResponse.json({
+        ok: true,
+        data: null,
+        status: "healthy",
+        checkedAt,
+        responseMs,
+        webUrl: `http://localhost:${port}/chat${token ? '?token=' + encodeURIComponent(token) : ''}`,
+      });
+    }
     return NextResponse.json({
       ok: false,
-      error: msg,
+      error: cli.error || raw,
       status: "down",
       checkedAt,
       responseMs,