Glossary · Last updated 2026-09-21

Glossary: Decision trace

Decision trace is jev.pro’s name for the audit record of one evaluation: filtered state (or a hash), question ids, full answers (including probabilities and confidence), resolved model (for example jev-1.13.0), usage, and the policy branch your code took. Jev (System One) does not emit a reasoning essay. The trace is the explanation you can show a reviewer. TypeSafe does not ship a trace object.

This glossary entry is for builders who already heard of Jev / System One and need a citable definition before they generate a fake chain of thought. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Confirm the API body on docs.typesafe.ai. We do not sell or issue API keys.

How it shows up

Persist the trace next to handoff tickets. Fields come from the official API body plus your own action enum. Always log response.model: if you sent jev-latest, the versioned id is what actually answered (currently jev-1.13.0 on the vendor snapshot we read). Log the full probabilities map — the top signal hides close races.

Choice and Score include confidence derived from the distribution. Noul has no confidence — store noul. Question IDs are not used in inference; they are your join keys. State is text/JSON only; hash if sensitive, keep enough fields for a replay. Replay is your POST, not a hidden endpoint.

Record the verdict and the threshold that produced it. Calibration work later depends on these rows. RLCD is a vendor training name, not a field on the trace. Schema-safe ≠ factually correct.

Contrast

Decision trace vs audit trail. The trace is one evaluation’s payload. The audit trail is the durable, identity-bearing store of many traces plus actor and policy metadata.

Decision trace vs generated rationale. If you need a paragraph, call an LLM after and label it generated assistance, never “Jev thought.”

Decision trace vs signal. The signal is the answer. The trace is signal + inputs + model pin + branch.

Decision trace vs decision contract. The contract is what you ship. The trace is what happened this time.

Common mistakes

• Storing only choice and dropping probabilities.
• Omitting response.model so alias moves are invisible.
• Writing a “because” and attributing it to Jev.
• Assuming TypeSafe retains your trail — read legal pages; keep a copy you control.
• Treating a 200 schema-safe row as a correct fact.

See also

• Glossary hub
• Audit trail
• Decision traces explainer
• Decision contract
• Verdict
• Confidence

FAQ

Is this an official TypeSafe type?

No. Unofficial working term. Canonical fields: docs.typesafe.ai/api.

Should I generate a rationale for the log?

If you need a paragraph, call an LLM after and label it generated assistance, never “Jev thought.”

Hash or raw state?

Hash if the payload is sensitive; keep enough fields for a replay.

What model id do I store?

The resolved response.model, not only the alias you sent.

Limits

This is a jev.pro working definition unless TypeSafe uses the same word as an API field. We do not invent search volume. Schema-safe answers are not automatically factually correct.

Disclaimer

This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.