Blog · Datasource Pulse
Field notes from the data supply chain.
What we learn running health checks on APIs, keys, and scrapers: staged failures, agents that pay their own way, and the quiet ways data pipelines break.
August 6, 2026 · Case study
Two scrapers changed their output on the same week. Triage took one question.
Our own monitoring flagged schema drift on a Reddit scraper and the Instagram scraper within the same week. Both runs exited green. The full incident with timestamps, the one triage question that closed it in minutes, and why drift severity is defined by the consumer, not the vendor.
Read the post →August 2, 2026 · Field guide
How to Detect When a Scraper’s Output Schema Changes
A renamed field never throws an error: the run exits green and a downstream column quietly fills with nulls. The fingerprint-and-diff pattern, the three traps that make homemade detectors page you at 3 a.m., and the automated schema_drift check at $0.02 per run.
Read the post →August 2, 2026 · Apify
How to Monitor an Apify Actor for Failures, Even One You Don’t Own
A failed run is the easy case. The costly failures keep the run status green: successful runs that deliver zero items, field names that quietly changed, record counts that collapsed. Six alarms for any Store actor you depend on but did not build, read-only, $0.02 per check.
Read the post →August 1, 2026 · Quota watch
How to Detect When an API Key Expires or Silently Fails
A dead key never announces itself: the job exits green and the vendor emailed an inbox nobody reads. The scheduled canary-call pattern with the four outcomes your checker must tell apart, real field names for GitHub, SerpAPI, and OpenAI, and the automated route at $0.02 per check.
Read the post →August 1, 2026 · Quota watch
How to get an alert when a data feed goes stale
The nastiest data failure has no red anywhere: the feed is up, the data is old. The cron-plus-timestamp pattern and its three traps, then supply-side freshness detection: zero-yield streaks, volume drops, and abandoned-scraper warnings.
Read the post →August 1, 2026 · Quota watch
How to monitor GitHub personal access token expiration
GitHub emails one inbox before a PAT expires, then your pipelines hit 401s. Read the expiry header from the free /rate_limit endpoint for the heads-up, and run a daily liveness check that also catches revocations and permission downgrades no reminder will.
Read the post →August 1, 2026 · Quota watch
How to get alerted before your SerpAPI quota runs out
The manual route with SerpAPI’s own account endpoint, the field trap that makes naive scripts report 0% used forever, and the automated route: one JSON block, $0.02 per check, with a heads-up at 80% and escalation at your own budget line.
Read the post →July 24, 2026 · Featured
An AI agent just paid our API two cents. No account. No card. No human.
An AI agent hit our x402 endpoint, paid $0.02 in USDC on Base, and got a live connector health verdict back. The payment settled on a public blockchain, so the payer, the receiver, and the amount are all checkable. On-chain proof inside.
Read the post →July 25, 2026 · n8n
We broke an endpoint on purpose
We pointed one of our n8n templates at a deliberately broken endpoint and watched the Slack alert land in about a minute. Every check is JSON you design: any endpoint, key, or actor, with 35+ vendors pre-wired as a head start.
Read the post →