New
JSON Reliability Tester
See how often a model returns the JSON your code depends on, measured over several runs.
If your app relies on a model returning clean JSON, one good answer is not enough. This runs the same request several times with your own key and tells you how many came back as valid JSON that matched your format, a plain pass rate like "7 of 10". Comes pre-loaded with a loose instruction next to a strict one, so your first run shows the strict one coming back clean far more often. It works with any model your key can reach, so it never goes out of date.
This free tool is a slice of the full product. With a paid plan you also get:
- score structured-output reliability across your whole workload
- catch a silent model swap with a canary
- see which part of your setup caused a failure
- track reliability over time
- let your AI agent run it for you
Free tool, provided as-is, no warranty. Uses your own API key. You are responsible for your usage and costs. Keys never leave your machine and are never sent to pysolvr. Outputs are informational, not advice. pysolvr is not liable for decisions made from these results.