|
| 1 | +/** |
| 2 | + * @license |
| 3 | + * Copyright 2026 Google LLC |
| 4 | + * SPDX-License-Identifier: Apache-2.0 |
| 5 | + */ |
| 6 | + |
| 7 | +import { MessageType } from '../types.js'; |
| 8 | +import type { HistoryItemPerfDashboard } from '../types.js'; |
| 9 | +import { formatDuration } from '../utils/formatters.js'; |
| 10 | +import { |
| 11 | + type CommandContext, |
| 12 | + type SlashCommand, |
| 13 | + CommandKind, |
| 14 | +} from './types.js'; |
| 15 | + |
| 16 | +/** |
| 17 | + * Collects a performance snapshot from the current session context. |
| 18 | + * This aggregates metrics from the session's models and tools into |
| 19 | + * a single PerfSnapshot object for the dashboard to display. |
| 20 | + */ |
| 21 | +function collectPerfSnapshot(context: CommandContext) { |
| 22 | + const { stats } = context.session; |
| 23 | + const { metrics } = stats; |
| 24 | + const now = new Date(); |
| 25 | + const uptimeMs = now.getTime() - stats.sessionStartTime.getTime(); |
| 26 | + |
| 27 | + // Aggregate tool performance by name |
| 28 | + const toolPerf: Array<{ |
| 29 | + name: string; |
| 30 | + calls: number; |
| 31 | + avgMs: number; |
| 32 | + totalMs: number; |
| 33 | + successRate: number; |
| 34 | + }> = []; |
| 35 | + |
| 36 | + for (const [name, toolStats] of Object.entries(metrics.tools.byName)) { |
| 37 | + const avgMs = toolStats.count > 0 ? toolStats.durationMs / toolStats.count : 0; |
| 38 | + const successRate = toolStats.count > 0 ? (toolStats.success / toolStats.count) * 100 : 0; |
| 39 | + toolPerf.push({ |
| 40 | + name, |
| 41 | + calls: toolStats.count, |
| 42 | + avgMs: Math.round(avgMs), |
| 43 | + totalMs: Math.round(toolStats.durationMs), |
| 44 | + successRate, |
| 45 | + }); |
| 46 | + } |
| 47 | + |
| 48 | + // Sort tools by total time descending |
| 49 | + toolPerf.sort((a, b) => b.totalMs - a.totalMs); |
| 50 | + |
| 51 | + // Aggregate API performance by model |
| 52 | + const apiPerf: Array<{ |
| 53 | + model: string; |
| 54 | + requests: number; |
| 55 | + avgLatencyMs: number; |
| 56 | + totalLatencyMs: number; |
| 57 | + errorRate: number; |
| 58 | + inputTokens: number; |
| 59 | + outputTokens: number; |
| 60 | + cachedTokens: number; |
| 61 | + }> = []; |
| 62 | + |
| 63 | + for (const [model, modelMetrics] of Object.entries(metrics.models)) { |
| 64 | + const avgLatency = modelMetrics.api.totalRequests > 0 |
| 65 | + ? modelMetrics.api.totalLatencyMs / modelMetrics.api.totalRequests |
| 66 | + : 0; |
| 67 | + const errorRate = modelMetrics.api.totalRequests > 0 |
| 68 | + ? (modelMetrics.api.totalErrors / modelMetrics.api.totalRequests) * 100 |
| 69 | + : 0; |
| 70 | + apiPerf.push({ |
| 71 | + model: model.replace('-001', ''), |
| 72 | + requests: modelMetrics.api.totalRequests, |
| 73 | + avgLatencyMs: Math.round(avgLatency), |
| 74 | + totalLatencyMs: Math.round(modelMetrics.api.totalLatencyMs), |
| 75 | + errorRate, |
| 76 | + inputTokens: modelMetrics.tokens.input, |
| 77 | + outputTokens: modelMetrics.tokens.candidates, |
| 78 | + cachedTokens: modelMetrics.tokens.cached, |
| 79 | + }); |
| 80 | + } |
| 81 | + |
| 82 | + // Memory snapshot |
| 83 | + const memUsage = process.memoryUsage(); |
| 84 | + const memory = { |
| 85 | + heapUsedMB: Math.round((memUsage.heapUsed / 1024 / 1024) * 10) / 10, |
| 86 | + heapTotalMB: Math.round((memUsage.heapTotal / 1024 / 1024) * 10) / 10, |
| 87 | + rssMB: Math.round((memUsage.rss / 1024 / 1024) * 10) / 10, |
| 88 | + externalMB: Math.round((memUsage.external / 1024 / 1024) * 10) / 10, |
| 89 | + }; |
| 90 | + |
| 91 | + // Memory warnings |
| 92 | + const memoryWarnings: string[] = []; |
| 93 | + const heapPercent = memUsage.heapUsed / memUsage.heapTotal; |
| 94 | + if (heapPercent > 0.9) { |
| 95 | + memoryWarnings.push('Critical: Heap usage above 90%!'); |
| 96 | + } else if (heapPercent > 0.75) { |
| 97 | + memoryWarnings.push('Warning: Heap usage above 75%'); |
| 98 | + } |
| 99 | + if (memUsage.rss > 512 * 1024 * 1024) { |
| 100 | + memoryWarnings.push('Warning: RSS memory exceeds 512MB'); |
| 101 | + } |
| 102 | + |
| 103 | + // Total token counts |
| 104 | + const totalInputTokens = Object.values(metrics.models).reduce( |
| 105 | + (acc, m) => acc + m.tokens.input, 0, |
| 106 | + ); |
| 107 | + const totalOutputTokens = Object.values(metrics.models).reduce( |
| 108 | + (acc, m) => acc + m.tokens.candidates, 0, |
| 109 | + ); |
| 110 | + const totalCachedTokens = Object.values(metrics.models).reduce( |
| 111 | + (acc, m) => acc + m.tokens.cached, 0, |
| 112 | + ); |
| 113 | + |
| 114 | + // Total API time and tool time |
| 115 | + const totalApiTime = Object.values(metrics.models).reduce( |
| 116 | + (acc, m) => acc + m.api.totalLatencyMs, 0, |
| 117 | + ); |
| 118 | + const totalToolTime = metrics.tools.totalDurationMs; |
| 119 | + |
| 120 | + return { |
| 121 | + uptimeMs, |
| 122 | + duration: formatDuration(uptimeMs), |
| 123 | + memory, |
| 124 | + memoryWarnings, |
| 125 | + toolPerf, |
| 126 | + apiPerf, |
| 127 | + totalToolCalls: metrics.tools.totalCalls, |
| 128 | + totalToolTime, |
| 129 | + totalApiRequests: Object.values(metrics.models).reduce( |
| 130 | + (acc, m) => acc + m.api.totalRequests, 0, |
| 131 | + ), |
| 132 | + totalApiTime, |
| 133 | + totalInputTokens, |
| 134 | + totalOutputTokens, |
| 135 | + totalCachedTokens, |
| 136 | + totalLinesAdded: metrics.files.totalLinesAdded, |
| 137 | + totalLinesRemoved: metrics.files.totalLinesRemoved, |
| 138 | + }; |
| 139 | +} |
| 140 | + |
| 141 | +export const perfCommand: SlashCommand = { |
| 142 | + name: 'perf', |
| 143 | + altNames: ['performance'], |
| 144 | + description: 'Performance monitoring dashboard. Usage: /perf [overview|memory|tools|api]', |
| 145 | + kind: CommandKind.BUILT_IN, |
| 146 | + autoExecute: false, |
| 147 | + isSafeConcurrent: true, |
| 148 | + action: async (context: CommandContext) => { |
| 149 | + const snapshot = collectPerfSnapshot(context); |
| 150 | + context.ui.addItem({ |
| 151 | + type: MessageType.PERF_DASHBOARD, |
| 152 | + view: 'overview', |
| 153 | + snapshot, |
| 154 | + } as HistoryItemPerfDashboard); |
| 155 | + }, |
| 156 | + subCommands: [ |
| 157 | + { |
| 158 | + name: 'overview', |
| 159 | + description: 'Show performance overview dashboard', |
| 160 | + kind: CommandKind.BUILT_IN, |
| 161 | + autoExecute: true, |
| 162 | + isSafeConcurrent: true, |
| 163 | + action: (context: CommandContext) => { |
| 164 | + const snapshot = collectPerfSnapshot(context); |
| 165 | + context.ui.addItem({ |
| 166 | + type: MessageType.PERF_DASHBOARD, |
| 167 | + view: 'overview', |
| 168 | + snapshot, |
| 169 | + } as HistoryItemPerfDashboard); |
| 170 | + }, |
| 171 | + }, |
| 172 | + { |
| 173 | + name: 'memory', |
| 174 | + description: 'Show detailed memory usage and warnings', |
| 175 | + kind: CommandKind.BUILT_IN, |
| 176 | + autoExecute: true, |
| 177 | + isSafeConcurrent: true, |
| 178 | + action: (context: CommandContext) => { |
| 179 | + const snapshot = collectPerfSnapshot(context); |
| 180 | + context.ui.addItem({ |
| 181 | + type: MessageType.PERF_DASHBOARD, |
| 182 | + view: 'memory', |
| 183 | + snapshot, |
| 184 | + } as HistoryItemPerfDashboard); |
| 185 | + }, |
| 186 | + }, |
| 187 | + { |
| 188 | + name: 'tools', |
| 189 | + description: 'Show tool execution timing and frequency', |
| 190 | + kind: CommandKind.BUILT_IN, |
| 191 | + autoExecute: true, |
| 192 | + isSafeConcurrent: true, |
| 193 | + action: (context: CommandContext) => { |
| 194 | + const snapshot = collectPerfSnapshot(context); |
| 195 | + context.ui.addItem({ |
| 196 | + type: MessageType.PERF_DASHBOARD, |
| 197 | + view: 'tools', |
| 198 | + snapshot, |
| 199 | + } as HistoryItemPerfDashboard); |
| 200 | + }, |
| 201 | + }, |
| 202 | + { |
| 203 | + name: 'api', |
| 204 | + description: 'Show model API latency breakdown', |
| 205 | + kind: CommandKind.BUILT_IN, |
| 206 | + autoExecute: true, |
| 207 | + isSafeConcurrent: true, |
| 208 | + action: (context: CommandContext) => { |
| 209 | + const snapshot = collectPerfSnapshot(context); |
| 210 | + context.ui.addItem({ |
| 211 | + type: MessageType.PERF_DASHBOARD, |
| 212 | + view: 'api', |
| 213 | + snapshot, |
| 214 | + } as HistoryItemPerfDashboard); |
| 215 | + }, |
| 216 | + }, |
| 217 | + { |
| 218 | + name: 'export', |
| 219 | + description: 'Export performance report as JSON', |
| 220 | + kind: CommandKind.BUILT_IN, |
| 221 | + autoExecute: true, |
| 222 | + isSafeConcurrent: true, |
| 223 | + action: (context: CommandContext) => { |
| 224 | + const snapshot = collectPerfSnapshot(context); |
| 225 | + const report = JSON.stringify({ |
| 226 | + timestamp: new Date().toISOString(), |
| 227 | + session_id: context.session.stats.sessionId, |
| 228 | + uptime_ms: snapshot.uptimeMs, |
| 229 | + memory: snapshot.memory, |
| 230 | + tools: { |
| 231 | + total_calls: snapshot.totalToolCalls, |
| 232 | + total_time_ms: snapshot.totalToolTime, |
| 233 | + by_tool: snapshot.toolPerf, |
| 234 | + }, |
| 235 | + api: { |
| 236 | + total_requests: snapshot.totalApiRequests, |
| 237 | + total_time_ms: snapshot.totalApiTime, |
| 238 | + by_model: snapshot.apiPerf, |
| 239 | + }, |
| 240 | + tokens: { |
| 241 | + input: snapshot.totalInputTokens, |
| 242 | + output: snapshot.totalOutputTokens, |
| 243 | + cached: snapshot.totalCachedTokens, |
| 244 | + }, |
| 245 | + files: { |
| 246 | + lines_added: snapshot.totalLinesAdded, |
| 247 | + lines_removed: snapshot.totalLinesRemoved, |
| 248 | + }, |
| 249 | + }, null, 2); |
| 250 | + |
| 251 | + context.ui.addItem({ |
| 252 | + type: MessageType.INFO, |
| 253 | + text: `📊 Performance Report (JSON):\n\n${report}`, |
| 254 | + }); |
| 255 | + }, |
| 256 | + }, |
| 257 | + ], |
| 258 | +}; |
0 commit comments