{# T55 sandbox diff — terminal-brutalist re-skin. #} {# Two-column compare: original LLM call vs experiment, with a #} {# token/cost/latency strip on top. #} {% extends "admin_layout.html" %} {% block page_title %}admin :: sandbox :: diff{% endblock %} {% block breadcrumb %} admin/ experiments/ {{ experiment.id[:8] }} {% endblock %} {% block head_extra %} {% endblock %} {% block content %}
Original LLM call on the left, your experiment on the right. The strip below carries delta-friendly stats for cost, tokens, and latency. Hidden empty diff cells render as the chat-view partial's empty state.
{# ===================== Header ===================== #}{{ experiment.prompt_name }}{{ experiment.created_at }}{{ experiment.model }}{{ original.call_id }}{% endif %}
No source LLM call (from-scratch experiment, or source row was deleted).
{% else %}
{% if original.scan_id %}scan {{ original.scan_id }} · {% endif %}
model {{ original.model }}
{{ experiment.id }}
model {{ experiment.model }}