{% extends "base.html" %} {% block title %}Samples · FreeWeight{% endblock %} {% block content %} {% if error %}

Samples

{% else %}

{{ test.test_name }} {{ test.test_key }}

Run {{ detail.run.id[:10] }}… · {{ test.status }} · {{ test.completed_cases }} / {{ test.total_cases }} cases × {{ test.repetitions }} repetitions

{% if samples %} {% for sample in samples %} {# The second of the two interactions UI standards §13 allows between a headline metric and its raw record: the case inspector shows this sample's prompt identity, its response, its tool calls, its per-criterion scoring and the telemetry observed while it ran. #} {% endfor %}
{{ samples | length }} sample(s). A sample that could not be scored shows — for its score and keeps its row: it is excluded from the aggregate, not from the count. A skipped sample was never sent — its error column says why.
Case Rep Status Prompt Score Method Output tokens Output chars TTFT (ms) Wall (ms) Finish Error Inspect
{{ sample.case_id }} {{ sample.repetition }} {{ sample.status | replace("_", " ") }} {% if sample.prompt_id %} {{ sample.prompt_id }} {{ sample.prompt_version }} {% else %}{% endif %} {% if sample.score is none %} {% else %}{{ "%.3f" | format(sample.score) }}{% endif %} {{ sample.score_method or "—" }} {{ sample.output_tokens if sample.output_tokens is not none else "—" }} {{ sample.output_chars if sample.output_chars is not none else "—" }} {% if sample.client_ttft_ms is none %} {% else %}{{ "%.1f" | format(sample.client_ttft_ms) }}{% endif %} {{ "%.1f" | format(sample.client_wall_ms) if sample.client_wall_ms is not none else "—" }} {{ sample.finish_reason or "—" }} {{ sample.error_code or "—" }}inspect
{% else %}

No samples yet.

This test has not produced a sample. It is real and empty, not missing.

{% endif %} {% endif %} {% endblock %}