MCP AI Bug Helper

#!/usr/bin/env tsx import { assertConfig, config } from '../src/config.js'; import { OpenRouterClient } from '../src/openrouterClient.js'; import { CodingAdvisorCoordinator } from '../src/codingAdvisors.js'; interface Scenario { name: string; question: string; context?: string; } const scenarios: Scenario[] = [ { name: 'React hydration + suspense deadlock', question: 'Next.js 15 app using React 19 RC occasionally hangs with a hydration error "Text content does not match server-rendered HTML" when Suspense + dynamic import are combined. How to stabilise hydration?', context: 'Key details: components use experimental `use()` with a data fetch; devtools shows server output includes `Loading…` placeholders that never resolve on client. There is also a custom `AppShell` that wraps everything in `<Suspense fallback={<Spinner/>}>`. Need actionable mitigation steps.', }, { name: 'Node.js worker threads crash under load', question: 'A worker-thread pool processing image thumbnails dies after 2-3k jobs with ERR_WORKER_UNSERIALIZABLE_ERROR. How can we keep the pool alive?', context: 'Implementation: parent sends `{ filePath, width, height }` objects to workers via `worker.postMessage`. Crash logs show `Cannot clone object, FileHandle` when streams are passed. Need suggestions for redesign + instrumentation.', }, { name: 'Go gRPC memory spike', question: 'Go 1.23 microservice uses gRPC streaming and memory climbs steadily until OOM. Profiling shows `runtime.malg` + `google.golang.org/grpc/internal/transport/recvBuffer` dominating. How to stop the leak?', context: 'Service streams protobuf chunks (~2MB) to clients with slow downstream. GOGC=100, but queue grows. We already set `grpc.MaxConcurrentStreams(10)` yet leak persists.', }, ]; async function main() { const filter = process.env.SCENARIO?.toLowerCase(); const selected = filter ? scenarios.filter((scenario) => scenario.name.toLowerCase().includes(filter)) : scenarios; if (selected.length === 0) { console.error('No scenarios matched filter:', filter); process.exit(1); } assertConfig(); const client = new OpenRouterClient(config.apiKey, { appName: config.appName, referer: config.referer, }); const coordinator = new CodingAdvisorCoordinator(client); for (const scenario of selected) { console.log(`\n=== Scenario: ${scenario.name} ===`); try { const batch = await coordinator.advise({ question: scenario.question, context: scenario.context, }); batch.answers.forEach((answer, idx) => { console.log(`\nAdvisor ${idx + 1}: ${answer.modelLabel}`); console.log( `Tier: ${answer.isFree ? 'free' : 'paid'} | Latency: ${answer.latencyMs}ms | Tag: ${answer.scenarioTag}` ); if (answer.usage) { console.log( `Usage: prompt=${answer.usage.promptTokens ?? 'n/a'}, completion=${answer.usage.completionTokens ?? 'n/a'}` ); } console.log('---'); console.log(answer.responseText); console.log('---\n'); }); } catch (error) { console.error('Scenario failed:', scenario.name, error); } } } main().catch((error) => { console.error('Fatal scenario runner error', error); process.exit(1); });