Research

Testing a sandbox integration with repeatable response fixtures

Run the documented JavaScript quickstart against seven synthetic responses, including failures, before connecting a production key.

A successful JSON response is only one part of an integration. A client also needs to stop on a failed HTTP request, malformed JSON or missing data. This small download makes those checks reproducible without an account, API key or network connection.

What is checked

The runner contains the JavaScript example from the documentation quickstart. It supplies a local replacement for fetch, then runs that example against the response fixtures. The example checks HTTP success, parses JSON and requires success, data and a non-empty data.prediction before printing it.

The fixed sample represents GET /sandbox/astrology/horoscope/aries. The fixture text, number and colour are synthetic. They are not a live forecast, a customer response or evidence of calculation accuracy.

Run it offline

Use Node.js 18 or later. Download both files into the same folder. Read the runner before executing it; it uses only the standard library and does not install packages.

node run.mjs

The expected final line is:

7/7 expected outcomes. Synthetic offline checks only; no API request made.

Each case prints its expected and actual acceptance. The process exits with code 0 only when all outcomes match, and 1 on a mismatch or an unreadable fixture file. This is a reproducible parser check, not a measured success rate for the API.

The seven cases

The expected behaviour is to accept the valid response and reject each failure below. The malformed JSON case deliberately cannot be parsed.

  1. Valid envelope: accept.
  2. Missing prediction: reject.
  3. Empty prediction: reject.
  4. Malformed JSON: reject.
  5. Success is false: reject.
  6. Absent data: reject.
  7. HTTP 500: reject.

To check that a failure can be detected, change a fixture’s accept value, save the file and run it again. Its checksum changes and the mismatched case must print FAIL with exit code 1.

Version and checksums

Fixture version: 2026-10-29. The runner prints the SHA-256 checksum of the fixture bytes it actually reads. Compare that value with the published value below before changing a case.

Response fixtures SHA-256
d853bbb2366dda28c80e3f5111dd13b59f1793c824fdd02aeab8efc0a585314d
Runner SHA-256
9c75c8448badc5c0dacc2e65dc2c8b0d58c09e0e12c46ea89ad875ac45f72558
Embedded quickstart SHA-256
74c1c3338b6fadb42e987c2848879bf5ebedfd2ea01dde01e30583273d9234ea

The quickstart checksum covers the decoded JavaScript code alone. It lets a reader identify the exact example used in this download. A later documentation change requires a new matching download and an updated note.

What these checks do not show

  • No API request is made, so this does not measure availability, latency, rate limits or the current live sandbox.
  • The seven hand-written fixtures do not cover every possible response or every API operation. They do not validate astronomical calculations or the meaning of a reading.
  • The example checks the fields it reads. It is not a complete JSON Schema validator.
  • Production authentication, authorisation, billing, idempotency and real response variations need separate operation-specific checks. A passing fixture run does not establish production readiness.

Connect the next step

Use the keyless sandbox to inspect supported demonstration responses. Read the operation reference for the inputs and response contract you need. When connecting production, keep the credential on your server and follow the authentication and error-handling guidance. Review the current plan terms before sending paid requests.

Published 29 October 2026 by Vedika Intelligence. Report corrections through the contact page.

More methods and documentation

Browse research notes · Read the quickstarts