Do I still need this classifier if my model's structured output mode is already on?

Usually, yes. A model’s structured output or function-calling mode makes a schema violation less likely, it doesn’t make one impossible. That guarantee is enforced upstream of your code, by the model provider, and it isn’t available or reliable on every call site in a larger system; if it’s ever misconfigured, absent on a fallback path, or simply not supported for one call site’s setup, nothing else in the pipeline notices until something downstream breaks on a field that isn’t there.

malformed_output doesn’t care what generated the output or what mode was on when it did. It captures the schema the call site declared and validates every response against it, on every trace, regardless of upstream guarantees. That makes it a backstop, not a replacement: a well-configured structured-output mode should mean the classifier rarely fires, and if it starts firing on a call site that’s supposed to be covered, that’s itself worth looking at.

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