Firecrawl Observability

Overview

Monitor Firecrawl web scraping pipelines for success rates, credit consumption, content quality, and latency. Key signals: scrape success rate, crawl job completion, credit burn velocity, extraction quality (did markdown actually contain useful content vs error pages), and webhook delivery health.

Key Metrics

Instructions

Step 1: Instrumented Firecrawl Wrapper

import FirecrawlApp from "@mendable/firecrawl-js";

const firecrawl = new FirecrawlApp({
  apiKey: process.env.FIRECRAWL_API_KEY!,
});

// Counters (use your metrics library: prom-client, statsd, datadog, etc.)
function emit(metric: string, value: number, tags?: Record<string, string>) {
  console.log(JSON.stringify({ metric, value, tags, timestamp: Date.now() }));
}

export async function instrumentedScrape(url: string) {
  const start = Date.now();
  try {
    const result = await firecrawl.scrapeUrl(url, {
      formats: ["markdown"],
      onlyMainContent: true,
    });

    const duration = Date.now() - start;
    const quality = evaluateQuality(result);

    emit("firecrawl_scrapes_total", 1, { status: "success" });
    emit("firecrawl_scrape_duration_ms", duration);
    emit("firecrawl_credits_used", 1);
    emit("firecrawl_content_quality", 1, { quality });

    return result;
  } catch (error: any) {
    emit("firecrawl_scrapes_total", 1, {
      status: "error",
      error_code: String(error.statusCode || "unknown"),
    });
    emit("firecrawl_scrape_duration_ms", Date.now() - start);
    throw error;
  }
}

function evaluateQuality(result: any): string {
  const md = result.markdown || "";
  if (md.length < 100) return "empty";
  if (/404|not found|access denied|captcha/i.test(md)) return "error_page";
  if (!/^#{1,3}\s/m.test(md)) return "no_structure";
  return "good";
}

Step 2: Credit Consumption Monitor

async function checkCreditHealth() {
  const response = await fetch("https://api.firecrawl.dev/v1/team/credits", {
    headers: { Authorization: `Bearer ${process.env.FIRECRAWL_API_KEY}` },
  });
  const data = await response.json();

  emit("firecrawl_credits_remaining", data.credits_remaining || 0);

  if (data.credits_remaining < 1000) {
    console.warn(`LOW CREDITS: ${data.credits_remaining} remaining`);
    emit("firecrawl_credit_alert", 1, { level: "warning" });
  }
  if (data.credits_remaining < 100) {
    emit("firecrawl_credit_alert", 1, { level: "critical" });
  }

  return data;
}

// Run every hour
setInterval(checkCreditHealth, 3600000);

Step 3: Crawl Job Tracking

export async function monitoredCrawl(url: string, limit: number) {
  const start = Date.now();

  const job = await firecrawl.asyncCrawlUrl(url, {
    limit,
    scrapeOptions: { formats: ["markdown"] },
  });

  emit("firecrawl_crawl_jobs_total", 1, { status: "started" });

  // Poll with metrics
  let status = await firecrawl.checkCrawlStatus(job.id);
  while (status.status === "scraping") {
    emit("firecrawl_crawl_progress", status.completed || 0, { jobId: job.id });
    await new Promise(r => setTimeout(r, 5000));
    status = await firecrawl.checkCrawlStatus(job.id);
  }

  const duration = Date.now() - start;
  emit("firecrawl_crawl_jobs_total", 1, { status: status.status });
  emit("firecrawl_crawl_duration_ms", duration);
  emit("firecrawl_crawl_pages", status.data?.length || 0);
  emit("firecrawl_credits_used", status.data?.length || 0);

  return status;
}

Step 4: Prometheus Alert Rules

groups:
  - name: firecrawl
    rules:
      - alert: FirecrawlHighFailureRate
        expr: rate(firecrawl_scrapes_total{status="error"}[1h]) / rate(firecrawl_scrapes_total[1h]) > 0.1
        annotations:
          summary: "Firecrawl error rate exceeds 10%"

      - alert: FirecrawlCreditLow
        expr: firecrawl_credits_remaining < 500
        annotations:
          summary: "Firecrawl credits below 500 — refill soon"

      - alert: FirecrawlHighLatency
        expr: histogram_quantile(0.95, firecrawl_scrape_duration_ms) > 15000
        annotations:
          summary: "Firecrawl p95 latency exceeds 15 seconds"

      - alert: FirecrawlPoorQuality
        expr: rate(firecrawl_content_quality{quality="empty"}[1h]) / rate(firecrawl_content_quality[1h]) > 0.2
        annotations:
          summary: "Over 20% of scrapes returning empty content"

Step 5: Dashboard Panels

Track these in Grafana/Datadog:

• Scrape volume: sum(rate(firecrawl_scrapes_total[5m])) by status
• Credit burn rate: sum(rate(firecrawl_credits_used[1h])) — credits/hour
• Latency p50/p95: histogram_quantile(0.5, firecrawl_scrape_duration_ms)
• Content quality: Pie chart of firecrawl_content_quality by quality label
• Credits remaining: Single stat with thresholds (green > 1000, yellow > 100, red < 100)

Error Handling

Resources

• Firecrawl API Reference
• prom-client (Prometheus for Node.js)
• Grafana Dashboards

Next Steps

For incident response, see firecrawl-incident-runbook.