Verified evidence from recorded runs. These examples deliberately test one bounded behavior at a time.
1. Create a Stripe OpenAPI baseline before claiming change detection
API Breaking-Change Radar run jW0uxs1wQyafDZjEX fetched the Stripe OpenAPI specification from:
https://raw.githubusercontent.com/stripe/openapi/master/openapi/spec3.json
Recorded output:
- API:
stripe. - Change class:
baseline. - Severity:
info. - Title:
Baseline recorded for stripe — future runs will diff against this spec version. - Recommended action: no immediate action; retain the baseline.
- Actor event charges: zero in dry-run; platform usage remained separate.
Public task: https://apify.com/travelmonitorlab/api-breaking-change-radar/examples/stripe-openapi-breaking-change-watch
Detailed proof: https://adagasolutions.lu/apis/api-breaking-change-radar.html
Why it matters: the first successful run establishes the reference used by later comparisons. It does not prove that Stripe has no breaking changes and must not be reported as a completed change-detection cycle. A second run against a changed specification is required to test that behavior.
2. Filter official French fuel data by two departments
Fuel Prices France API run 63iU7PabbEbQLw0vD used departement=69,75, carburant=Gazole and dry-run billing.
Recorded output:
- 226 station records returned after filtering to departments 69 and 75.
- Sample station: ID
69480002, Anse, postal code69480. - Sample Gazole price in that snapshot: 2.309.
- Actor event charges: zero; platform usage remained separate.
Public task: https://apify.com/travelmonitorlab/hermes-revenu-api/examples/compare-diesel-prices-rhone-paris
Detailed proof: https://adagasolutions.lu/apis/hermes-revenu-api.html
Why it matters: the example proves filtering and output structure without presenting the snapshot as today's cheapest station or a national average. A production integration should preserve the fetch time and query filters alongside each price.
3. Inspect the run, not only the HTTP success status
Two additional runs show why dataset inspection matters:
- Brand Tracker run
dk1Hi70ELwgSUp01Iretained three real model answers across two models and one provider error. The provider failure remains visible instead of being silently counted as a measurement. - Hotel Rate Monitoring run
7ObEuoyCma8msaGhqreturned two source-backed hotel cards with HTTP provenance and content hashes. Its suite anomaly, minimum 201 and maximum 195, was preserved rather than “fixed” into a plausible-looking value.
Public tasks:
Integration checklist
- Start from the public task and keep the smallest bounded input.
- Inspect the actual dataset item count and a sample item.
- Distinguish technical dry-run, real billed extraction and provider errors.
- Preserve timestamps, source URLs, query filters and content hashes where available.
- Do not infer live availability, universal AI visibility, absence of breaking changes or current market prices from a one-time acceptance run.
- Check Actor event charges separately from compute, proxy and external-provider costs.
The public catalogue contains 30 prefilled task links, including clearly labelled technical dry-runs:
https://adagasolutions.lu/toolbox/
Source repository: https://github.com/jeffreyturov-dev/apify-scraping-toolbox