{# T36.1 — sandbox editor under the T28 admin layout. #} {# Was a standalone cream-themed page with the old green admin banner; #} {# now extends ``admin_layout.html`` so the sidebar nav, header brand, #} {# role chip, and Tailwind CDN come from the layout — and "Experiments" #} {# lights up while you're on any ``/admin/sandbox/*`` path. #} {% extends "admin_layout.html" %} {% block page_title %}Sandbox · {{ prompt_name }}{% endblock %} {% block breadcrumb %} Admin / Experiments / {{ prompt_name }} {% endblock %} {% block content %}
{{ source_call_id }}
{% if scan_id %}
· scan
{{ scan_id }}
{% endif %}
· prompt
{{ prompt_name }}
Edit either prompt, pick a model, then click Run experiment.
This NEVER overwrites the original LLM call — a new
llm_experiments row is created instead.
Cost is billed to your admin account
(cap: see /admin/sandbox/list).
dump). #}
{# We feed the FULL source dump dict into the partial via inline_dump #}
{# so the system + user + response render together. The partial gates #}
{# on each field being present, so the missing-dump case (where #}
{# ``original_dump_json`` is ``{}``) collapses to the partial's empty #}
{# "No content." state. #}
Original response
{% with chat_view_id='original-response-' ~ source_call_id, inline_dump=original_dump_json %}
{% include 'partials/llm_chat_view.html' %}
{% endwith %}
{# ===================== Footer links ===================== #}