What is Tessary's malformed_output classifier?

It validates each structured output against the schema its call site declared: a pass or a fail, no judgment involved. When a call site is set up to return structured data, Tessary captures that schema and checks the outputs against it.

It catches exactly what your calling code would hit in production: output that doesn’t parse as valid JSON, or output that parses fine but violates the declared shape, a missing field, the wrong type, an extra wrapper around the object. Because the check is mechanical rather than judged, every firing is high confidence, and it carries the actual violation message as evidence.

Two limits. Where a call site declares no schema there’s nothing to check against, so it stays quiet. And a well-formed answer can still be a wrong one, which this says nothing about. A violation writes a finding, and triage decides whether it becomes a case.

keep reading

More on this.

Two ways to run Tessary.

Tessary is an open-source agent reliability platform. Cloud and self-hosted run the same workflow on the OpenTelemetry traces your agent already emits.

Tessary Cloud

We host it for you. Send your first trace with nothing to deploy and no model key.

what's includedper organization
traces
10,000 per calendar month
stored trace data
1 GB
retention
30 days
model credit
$10, one-time, for triage and root-cause analysis
credit card
not required

Self-hosted Tessary

Run the open-source code on your own infrastructure with one command. Add your own model key for triage and root-cause analysis.

Self-host Tessary for me by following https://github.com/tessaryai/tessary/blob/main/setup.md

docker compose -f oci://docker.io/tessaryai/tessary:compose up -d -y