Auto-repair invalid OpenClaw config on gateway startup

Co-authored-by: Haze <hazeone@users.noreply.github.com>

electron/gateway/manager.ts

@@ -32,6 +32,7 @@ import {
   type DeviceIdentity,
 } from '../utils/device-identity';
 import { syncGatewayTokenToConfig, syncBrowserConfigToOpenClaw } from '../utils/openclaw-auth';
+import { shouldAttemptConfigAutoRepair } from './startup-recovery';
 
 /**
  * Gateway connection status
@@ -176,6 +177,7 @@ export class GatewayManager extends EventEmitter {
   private shouldReconnect = true;
   private startLock = false;
   private lastSpawnSummary: string | null = null;
+  private recentStartupStderrLines: string[] = [];
   private pendingRequests: Map<string, {
     resolve: (value: unknown) => void;
     reject: (error: Error) => void;
@@ -232,6 +234,16 @@ export class GatewayManager extends EventEmitter {
 
     return { level: 'warn', normalized: msg };
   }
+
+  private recordStartupStderrLine(line: string): void {
+    const normalized = line.trim();
+    if (!normalized) return;
+    this.recentStartupStderrLines.push(normalized);
+    const MAX_STDERR_LINES = 120;
+    if (this.recentStartupStderrLines.length > MAX_STDERR_LINES) {
+      this.recentStartupStderrLines.splice(0, this.recentStartupStderrLines.length - MAX_STDERR_LINES);
+    }
+  }
   
   /**
    * Get current Gateway status
@@ -279,49 +291,71 @@ export class GatewayManager extends EventEmitter {
 
     this.reconnectAttempts = 0;
     this.setStatus({ state: 'starting', reconnectAttempts: 0 });
+    let configRepairAttempted = false;
     
     try {
-      // Check if Python environment is ready (self-healing) asynchronously
-      void isPythonReady().then(pythonReady => {
-        if (!pythonReady) {
-          logger.info('Python environment missing or incomplete, attempting background repair...');
-          // We don't await this to avoid blocking Gateway startup, 
-          // as uv run will handle it if needed, but this pre-warms it.
-          void setupManagedPython().catch(err => {
-            logger.error('Background Python repair failed:', err);
+      while (true) {
+        this.recentStartupStderrLines = [];
+        try {
+          // Check if Python environment is ready (self-healing) asynchronously
+          void isPythonReady().then(pythonReady => {
+            if (!pythonReady) {
+              logger.info('Python environment missing or incomplete, attempting background repair...');
+              // We don't await this to avoid blocking Gateway startup,
+              // as uv run will handle it if needed, but this pre-warms it.
+              void setupManagedPython().catch(err => {
+                logger.error('Background Python repair failed:', err);
+              });
+            }
+          }).catch(err => {
+            logger.error('Failed to check Python environment:', err);
           });
-        }
-      }).catch(err => {
-        logger.error('Failed to check Python environment:', err);
-      });
 
-      // Check if Gateway is already running
-      logger.debug('Checking for existing Gateway...');
-      const existing = await this.findExistingGateway();
-      if (existing) {
-        logger.debug(`Found existing Gateway on port ${existing.port}`);
-        await this.connect(existing.port, existing.externalToken);
-        this.ownsProcess = false;
-        this.setStatus({ pid: undefined });
-        this.startHealthCheck();
-        return;
+          // Check if Gateway is already running
+          logger.debug('Checking for existing Gateway...');
+          const existing = await this.findExistingGateway();
+          if (existing) {
+            logger.debug(`Found existing Gateway on port ${existing.port}`);
+            await this.connect(existing.port, existing.externalToken);
+            this.ownsProcess = false;
+            this.setStatus({ pid: undefined });
+            this.startHealthCheck();
+            return;
+          }
+          
+          logger.debug('No existing Gateway found, starting new process...');
+          
+          // Start new Gateway process
+          await this.startProcess();
+          
+          // Wait for Gateway to be ready
+          await this.waitForReady();
+          
+          // Connect WebSocket
+          await this.connect(this.status.port);
+          
+          // Start health monitoring
+          this.startHealthCheck();
+          logger.debug('Gateway started successfully');
+          return;
+        } catch (error) {
+          if (shouldAttemptConfigAutoRepair(error, this.recentStartupStderrLines, configRepairAttempted)) {
+            configRepairAttempted = true;
+            logger.warn(
+              'Detected invalid OpenClaw config during Gateway startup; running doctor repair before retry'
+            );
+            const repaired = await this.runOpenClawDoctorRepair();
+            if (repaired) {
+              logger.info('OpenClaw doctor repair completed; retrying Gateway startup');
+              this.setStatus({ state: 'starting', error: undefined, reconnectAttempts: 0 });
+              continue;
+            }
+            logger.error('OpenClaw doctor repair failed; not retrying Gateway startup');
+          }
+          throw error;
+        }
       }
       
-      logger.debug('No existing Gateway found, starting new process...');
-      
-      // Start new Gateway process
-      await this.startProcess();
-      
-      // Wait for Gateway to be ready
-      await this.waitForReady();
-      
-      // Connect WebSocket
-      await this.connect(this.status.port);
-      
-      // Start health monitoring
-      this.startHealthCheck();
-      logger.debug('Gateway started successfully');
-      
     } catch (error) {
       logger.error(
         `Gateway start failed (port=${this.status.port}, reconnectAttempts=${this.reconnectAttempts}, spawn=${this.lastSpawnSummary ?? 'n/a'})`,
@@ -709,6 +743,115 @@ export class GatewayManager extends EventEmitter {
     
     return null;
   }
+
+  /**
+   * Attempt to repair invalid OpenClaw config using the built-in doctor command.
+   * Returns true when doctor exits successfully.
+   */
+  private async runOpenClawDoctorRepair(): Promise<boolean> {
+    const openclawDir = getOpenClawDir();
+    const entryScript = getOpenClawEntryPath();
+    if (!existsSync(entryScript)) {
+      logger.error(`Cannot run OpenClaw doctor repair: entry script not found at ${entryScript}`);
+      return false;
+    }
+
+    const platform = process.platform;
+    const arch = process.arch;
+    const target = `${platform}-${arch}`;
+    const binPath = app.isPackaged
+      ? path.join(process.resourcesPath, 'bin')
+      : path.join(process.cwd(), 'resources', 'bin', target);
+    const binPathExists = existsSync(binPath);
+    const finalPath = binPathExists
+      ? `${binPath}${path.delimiter}${process.env.PATH || ''}`
+      : process.env.PATH || '';
+
+    const uvEnv = await getUvMirrorEnv();
+    const command = app.isPackaged ? getNodeExecutablePath() : 'node';
+    const args = [entryScript, 'doctor', '--fix', '--yes', '--non-interactive'];
+    logger.info(
+      `Running OpenClaw doctor repair (command="${command}", args="${args.join(' ')}", cwd="${openclawDir}", bundledBin=${binPathExists ? 'yes' : 'no'})`
+    );
+
+    return new Promise<boolean>((resolve) => {
+      const spawnEnv: Record<string, string | undefined> = {
+        ...process.env,
+        PATH: finalPath,
+        ...uvEnv,
+      };
+
+      if (app.isPackaged) {
+        spawnEnv['ELECTRON_RUN_AS_NODE'] = '1';
+        spawnEnv['OPENCLAW_NO_RESPAWN'] = '1';
+        const existingNodeOpts = spawnEnv['NODE_OPTIONS'] ?? '';
+        if (!existingNodeOpts.includes('--disable-warning=ExperimentalWarning') &&
+            !existingNodeOpts.includes('--no-warnings')) {
+          spawnEnv['NODE_OPTIONS'] = `${existingNodeOpts} --disable-warning=ExperimentalWarning`.trim();
+        }
+      }
+
+      const child = spawn(command, args, {
+        cwd: openclawDir,
+        stdio: ['ignore', 'pipe', 'pipe'],
+        detached: false,
+        shell: false,
+        env: spawnEnv,
+      });
+
+      let settled = false;
+      const finish = (ok: boolean) => {
+        if (settled) return;
+        settled = true;
+        resolve(ok);
+      };
+
+      const timeout = setTimeout(() => {
+        logger.error('OpenClaw doctor repair timed out after 120000ms');
+        try {
+          child.kill('SIGTERM');
+        } catch {
+          // ignore
+        }
+        finish(false);
+      }, 120000);
+
+      child.on('error', (err) => {
+        clearTimeout(timeout);
+        logger.error('Failed to spawn OpenClaw doctor repair process:', err);
+        finish(false);
+      });
+
+      child.stdout?.on('data', (data) => {
+        const raw = data.toString();
+        for (const line of raw.split(/\r?\n/)) {
+          const normalized = line.trim();
+          if (!normalized) continue;
+          logger.debug(`[Gateway doctor stdout] ${normalized}`);
+        }
+      });
+
+      child.stderr?.on('data', (data) => {
+        const raw = data.toString();
+        for (const line of raw.split(/\r?\n/)) {
+          const normalized = line.trim();
+          if (!normalized) continue;
+          logger.warn(`[Gateway doctor stderr] ${normalized}`);
+        }
+      });
+
+      child.on('exit', (code, signal) => {
+        clearTimeout(timeout);
+        if (code === 0) {
+          logger.info('OpenClaw doctor repair completed successfully');
+          finish(true);
+          return;
+        }
+        logger.warn(`OpenClaw doctor repair exited (${this.formatExit(code, signal)})`);
+        finish(false);
+      });
+    });
+  }
   
   /**
    * Start Gateway process
@@ -924,6 +1067,7 @@ export class GatewayManager extends EventEmitter {
       child.stderr?.on('data', (data) => {
         const raw = data.toString();
         for (const line of raw.split(/\r?\n/)) {
+          this.recordStartupStderrLine(line);
           const classified = this.classifyStderrMessage(line);
           if (classified.level === 'drop') continue;
           if (classified.level === 'debug') {

electron/gateway/startup-recovery.ts

@@ -0,0 +1,60 @@
+/**
+ * Gateway startup recovery heuristics.
+ *
+ * This module is intentionally dependency-free so it can be unit-tested
+ * without Electron/runtime mocks.
+ */
+
+const INVALID_CONFIG_PATTERNS: RegExp[] = [
+  /\binvalid config\b/i,
+  /\bconfig invalid\b/i,
+  /\bunrecognized key\b/i,
+  /\brun:\s*openclaw doctor --fix\b/i,
+];
+
+function normalizeLogLine(value: string): string {
+  return value.trim();
+}
+
+/**
+ * Returns true when text appears to indicate OpenClaw config validation failure.
+ */
+export function isInvalidConfigSignal(text: string): boolean {
+  const normalized = normalizeLogLine(text);
+  if (!normalized) return false;
+  return INVALID_CONFIG_PATTERNS.some((pattern) => pattern.test(normalized));
+}
+
+/**
+ * Returns true when either startup stderr lines or startup error message
+ * indicate an OpenClaw config validation failure.
+ */
+export function hasInvalidConfigFailureSignal(
+  startupError: unknown,
+  startupStderrLines: string[],
+): boolean {
+  for (const line of startupStderrLines) {
+    if (isInvalidConfigSignal(line)) {
+      return true;
+    }
+  }
+
+  const errorText = startupError instanceof Error
+    ? `${startupError.name}: ${startupError.message}`
+    : String(startupError ?? '');
+
+  return isInvalidConfigSignal(errorText);
+}
+
+/**
+ * Retry guard for one-time config repair during a single startup flow.
+ */
+export function shouldAttemptConfigAutoRepair(
+  startupError: unknown,
+  startupStderrLines: string[],
+  alreadyAttempted: boolean,
+): boolean {
+  if (alreadyAttempted) return false;
+  return hasInvalidConfigFailureSignal(startupError, startupStderrLines);
+}
+

tests/unit/gateway-startup-recovery.test.ts

@@ -0,0 +1,52 @@
+import { describe, expect, it } from 'vitest';
+import {
+  hasInvalidConfigFailureSignal,
+  isInvalidConfigSignal,
+  shouldAttemptConfigAutoRepair,
+} from '@electron/gateway/startup-recovery';
+
+describe('gateway startup recovery heuristics', () => {
+  it('detects invalid-config signal from stderr lines', () => {
+    const lines = [
+      'Invalid config at C:\\Users\\pc\\.openclaw\\openclaw.json:\\n- skills: Unrecognized key: "enabled"',
+      'Run: openclaw doctor --fix',
+    ];
+    expect(hasInvalidConfigFailureSignal(new Error('gateway start failed'), lines)).toBe(true);
+  });
+
+  it('detects invalid-config signal from error message fallback', () => {
+    expect(
+      hasInvalidConfigFailureSignal(
+        new Error('Config invalid. Run: openclaw doctor --fix'),
+        [],
+      ),
+    ).toBe(true);
+  });
+
+  it('does not treat unrelated startup failures as invalid-config failures', () => {
+    const lines = [
+      'Gateway process exited (code=1, expected=no)',
+      'WebSocket closed before handshake',
+    ];
+    expect(
+      hasInvalidConfigFailureSignal(
+        new Error('Gateway process exited before becoming ready (code=1)'),
+        lines,
+      ),
+    ).toBe(false);
+  });
+
+  it('attempts auto-repair only once per startup flow', () => {
+    const lines = ['Config invalid', '- skills: Unrecognized key: "enabled"'];
+    expect(shouldAttemptConfigAutoRepair(new Error('start failed'), lines, false)).toBe(true);
+    expect(shouldAttemptConfigAutoRepair(new Error('start failed'), lines, true)).toBe(false);
+  });
+
+  it('matches common invalid-config phrases robustly', () => {
+    expect(isInvalidConfigSignal('Config invalid')).toBe(true);
+    expect(isInvalidConfigSignal('skills: Unrecognized key: "enabled"')).toBe(true);
+    expect(isInvalidConfigSignal('Run: openclaw doctor --fix')).toBe(true);
+    expect(isInvalidConfigSignal('Gateway ready after 3 attempts')).toBe(false);
+  });
+});
+