Is malformed output invalid JSON, or just the wrong shape?

Both count. malformed_output fires in two distinct cases: output that doesn’t parse at all, broken JSON, a trailing comma, prose wrapped around what should be a raw object, and output that parses perfectly fine but violates the shape the call site declared, a field missing, the wrong type, an extra field nobody asked for.

The second case is the more common one, and the easier one to miss without something watching for it. Output that parses is easy to mistake for output that’s correct: a response returning a quantity as the string “3” instead of the number 3 loads without an error and then fails somewhere downstream, in code that never expected to receive a string there.

Either way, the classifier treats them the same, as a violation of the declared contract, and records the actual mismatch as evidence rather than just a pass or fail flag.

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