Back to all updatesRead evaluation results from your terminal
List evaluations, runs and dataset versions from the CLI, or ask the assistant how a run scored.
The TraceRoot CLI can now read your evaluation results, so you or your coding agent can check a run without opening the dashboard.
Start with traceroot evals list, narrow to one evaluation with traceroot evals runs list --evaluation-id <id>, then run traceroot evals runs get <run-id> to see the run's status, scorer averages, cost and duration.
support-agent · run #5 · completed
candidate: prompt-v2+haiku-5-5 environment: evaluation
dataset: ds_d0237b5a47f0f2c37a319ddcfb @ 371028929235910658
METRIC VALUE UNIT CASES
correct_category 1 — 12
cites_account_facts 1 — 12
helpfulness 0.583 — 12
duration 8,031 ms 12 (mean per case)
cost 0.00158 $ 12 (mean per case)
Use traceroot datasets versions get <version-id> to read the exact cases a run was scored on. Every command takes --json for scripts and agents.
The in-app assistant reads the same data: link a run and ask how it did.
Upgrade to traceroot-cli 0.5.0 and see evaluations in the CLI.