Skip to main content
Glama

usage_stats

Monitor model usage statistics including call counts, success rates, and average response times to track performance and optimize resource allocation.

Instructions

查看各模型的使用统计(调用次数、成功率、平均耗时)

Input Schema

TableJSON Schema
NameRequiredDescriptionDefault

No arguments

Implementation Reference

  • The handler for the 'usage_stats' tool. It retrieves formatted usage statistics from the globalStats instance and returns it as text content.
    case 'usage_stats': { return { content: [ { type: 'text', text: globalStats.formatStats(), }, ], }; }
  • src/server.ts:259-266 (registration)
    Registration of the 'usage_stats' tool in the ListTools response, including its name, description, and empty input schema (no parameters required).
    { name: 'usage_stats', description: '查看各模型的使用统计(调用次数、成功率、平均耗时)', inputSchema: { type: 'object', properties: {}, }, },
  • The core helper method formatStats() in UsageStats class that computes and formats the usage statistics into a Markdown table for display.
    formatStats(): string { const stats = this.getGlobalStats(); const uptime = Math.floor((Date.now() - stats.since) / 1000 / 60); let output = `📊 **使用统计** (运行 ${uptime} 分钟)\n\n`; output += `| 指标 | 数值 |\n|------|------|\n`; output += `| 总调用 | ${stats.totalCalls} 次 |\n`; output += `| 成功 | ${stats.totalSuccess} 次 |\n`; output += `| 失败 | ${stats.totalFailed} 次 |\n`; output += `| 平均耗时 | ${(stats.avgDuration / 1000).toFixed(2)}s |\n`; output += `| 总耗时 | ${(stats.totalDuration / 1000).toFixed(1)}s |\n\n`; if (stats.models.length > 0) { output += `### 各模型统计\n\n`; output += `| 模型 | 调用 | 成功率 | 平均耗时 |\n|------|------|--------|----------|\n`; for (const model of stats.models) { output += `| ${model.model} | ${model.totalCalls} | ${(model.successRate * 100).toFixed(0)}% | ${(model.avgDuration / 1000).toFixed(2)}s |\n`; } } return output; }

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/7836246/claude-team-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server