Resilience Skill

LLM API error tracking, classification, retry, and task recovery for OpenClaw.

Overview

This skill provides visibility into API call health and automated retry management. Use it to:

• Monitor API error rates and patterns
• View per-model performance statistics
• Configure retry strategies
• Generate error reports
• Track task recovery status

Tools

resilience_dashboard

Open the live web dashboard in your browser for real-time error stats and retry strategy management.

Parameters:

• action: "open" (default) | "status" | "stop"

Features:

• Live error overview (today / hour / active retries)
• Model breakdown table
• Recent errors feed
• Retry strategy cards — set default, adjust max retries
• Auto-refresh: 5s, 60s, 5min, 1h, or off

URL: http://127.0.0.1:18765/ (default port, configurable via dashboardPort)

Voice / natural language examples:

• "打开错误统计页面" → resilience_dashboard({ action: "open" })
• "打开监控面板" → resilience_dashboard({ action: "open" })
• "打开 resilience 面板" → resilience_dashboard({ action: "open" })

The dashboard starts automatically when OpenClaw Gateway starts (unless dashboardEnabled: false).

Configuration lives in ~/.openclaw/openclaw.json under plugins.entries.resilience.config (not only api.pluginConfig at hook time). Example:

"resilience": {
  "enabled": true,
  "config": {
    "dashboardPort": 18765,
    "dashboardEnabled": true,
    "instanceLabel": "my-workspace"
  }
}

At gateway_start, config is read from ctx.config + ctx.workspaceDir.

Multi-instance: Use the instance dropdown to view all instances (aggregated) or a single Gateway. Each instance stores data under ~/.openclaw/plugins/resilience/instances/\x3Cid>/. Strategy edits apply only to the local Gateway instance.

resilience_stats

View API error statistics by time period or model.

Parameters:

• query (optional): Natural language query
  • "today" or empty — today's full summary
  • "hour" — current hour stats
  • "week" — current week stats
  • Any model name (e.g., "mimo-v2.5") — model-specific stats

Examples:

• "查看今天报错统计" → resilience_stats({ query: "today" })
• "查看 mimo-v2.5 的错误率" → resilience_stats({ query: "mimo-v2.5" })
• "查看本周错误率" → resilience_stats({ query: "week" })

resilience_strategies

View, add, update, or reset retry strategies.

Parameters:

• action: "list" (default) | "add" | "update" | "reset"
• strategyName: Strategy name (required for add/update)
• updates: Fields to update (for add/update)

Examples:

• "查看当前所有策略配置" → resilience_strategies({ action: "list" })
• "修改超时重试策略为指数退避" → resilience_strategies({ action: "update", strategyName: "default-exponential", updates: { type: "exponential" } })
• "添加一个自定义重试策略" → resilience_strategies({ action: "add", strategyName: "my-strategy", updates: { type: "custom", maxRetries: 3, intervals: [60000, 300000, 600000] } })
• "重置策略为默认" → resilience_strategies({ action: "reset" })

resilience_report

Generate detailed error reports.

Parameters:

• reportType: "daily" (default) | "model" | "recovery" | "full"
• target: Model name or date (YYYY-MM-DD)

Examples:

• "生成今日错误日报" → resilience_report({ reportType: "daily" })
• "查看 mimo-v2.5 的详细报告" → resilience_report({ reportType: "model", target: "mimo-v2.5" })
• "查看任务恢复状态" → resilience_report({ reportType: "recovery" })
• "生成完整状态报告" → resilience_report({ reportType: "full" })

Error Categories

Retry Strategies

Strategy Types

• fixed: Fixed interval between retries (e.g., every 30s)
• exponential: Exponential backoff (1min → 2min → 4min → 8min...)
• custom: User-defined interval schedule (e.g., [1min, 3min, 5min, 15min])

Default Strategies

Data Storage

Per-instance data: ~/.openclaw/plugins/resilience/instances/\x3Cinstance-id>/ (stats, logs, strategies, tasks). Legacy root layout is still read as default.

~/.openclaw/plugins/resilience/instances/\x3Cinstance-id>/
├── meta.json
├── stats.json
├── strategies.json
├── active-retries.json
├── logs/YYYY-MM-DD.jsonl
└── tasks/