{% extends "base.html" %} {% import "_ui.html" as ui %} {% block title %}{{ r.report.title }}{% endblock %} {% block crumbs %}Runs{{ ui.icon("chevron-right", "sm") }}{{ r.report.title }}{% endblock %} {% block content %} {%- set ns = namespace(q_ok=0, q_all=0, st_pass=0, approved=0, ev_ok=0, ev_all=0) -%} {%- for p in r.packages -%} {%- set ns.q_ok = ns.q_ok + (p.queries|selectattr("query")|list|length) -%} {%- set ns.q_all = ns.q_all + (p.queries|length) -%} {%- if p.selftest == "pass" %}{% set ns.st_pass = ns.st_pass + 1 %}{% endif -%} {%- if p.review_status == "approved" %}{% set ns.approved = ns.approved + 1 %}{% endif -%} {%- endfor -%} {%- for b in r.behaviours -%} {%- set ns.ev_ok = ns.ev_ok + (b.evidence_verified|select|list|length) -%} {%- set ns.ev_all = ns.ev_all + (b.evidence|length) -%} {%- endfor -%} {%- set max_rank = (r.behaviours|map(attribute="rank_score")|max) if r.behaviours else 0 -%} {%- set n_rules = r.packages|length -%}
{{ r.analysis.summary }}
{% endif %} {% if r.warnings %}{{ r.report.sha256 }}{{ ui.copy_btn(r.report.sha256, "", "btn ghost icon sm") }}{{ r.run_id }} · prompts {{ r.prompt_version }}{% if r.tool_version %} · ti2dac {{ r.tool_version }}{% endif %}ai.prices).| Stage | Model | Calls | Cached | Input | Output | Cache read | Cache write | Est. cost |
|---|---|---|---|---|---|---|---|---|
| {{ stage }} | {{ u.model_price_key }} | {{ u.calls }} | {{ u.cached_calls }} | {{ "{:,}".format(u.input_tokens) }} | {{ "{:,}".format(u.output_tokens) }} | {{ "{:,}".format(u.cache_read_tokens) }} | {{ "{:,}".format(u.cache_write_tokens or 0) }} | {{ "not priced" if u.model_price_key in r.usage.unpriced_models and not u.cost_usd else ("$%.4f"|format(u.cost_usd)) if u.cost_usd is defined else "–" }} |
| Total | {{ r.usage.calls }} | {{ r.usage.cached_calls }} | {{ "{:,}".format(r.usage.input_tokens) }} | {{ "{:,}".format(r.usage.output_tokens) }} | {{ "{:,}".format(r.usage.cache_read_tokens) }} | {{ "{:,}".format(r.usage.cache_write_tokens) }} | {{ "not priced" if r.usage.unpriced_models and not r.usage.estimated_cost_usd else "$%.4f"|format(r.usage.estimated_cost_usd or 0) }} |
Output tokens include the model's thinking. Analyse = 1 call per report; generate = 1 per rule; repair/translate only when needed. Cached calls cost nothing. Costs are priced at the model that actually answered.
A second opinion on this run, checked against the report. Nothing changes until you accept a finding.
{{ review.model }}{% if review.unpriced %} · not priced{% else %} · est. ${{ "%.4f"|format(review.cost_usd) }}{% endif %}
{% else %}ti2dac checks only (no AI call){% endif %}
{{ review.findings|length }} finding(s){% if review.discarded %} · {{ review.discarded|length }} discarded{% endif %}
{{ review.summary }}
{% endif %} {% if not review.findings %}{{ ui.empty("No mistakes found", "Neither the checks nor the reviewer found a problem with this run.", "check-circle") }}{% endif %}{{ f.problem }}
{% if f.report_quote %}{% if f.quote_verified %}{{ ui.icon("check", "sm") }} {% endif %}{% endif %} {% if f.suggested_fix %}{{ f.report_quote }}
A finding that rests on the report must quote it word for word. These quotes were not found, so the reviewer may have misread or invented them. They are kept here so nothing disappears unseen.
{{ f.report_quote }}
review stage on the AI page.TTPs extracted from the report, ranked by how detectable, reproducible and low-noise they are. {% if ai_enabled %}Tick behaviours without a rule to generate more.{% else %}AI is off, so no new rules can be generated here.{% endif %}
{{ p.rule_yaml }}
{{ i }}{{ ui.icon("history", "sm") }} Previous versions: {% for h in p.history %}v{{ h.version }} ({{ h.replaced_at }}: {{ h.reason }}){% if not loop.last %}, {% endif %}{% endfor %}
{% endif %}{{ q.query }}
{% elif st == "not-available" %}{{ q.error }}{{ q.error }}{{ q.notes }}
{% endif %}How each Sigma field is named in each SIEM. Red cells are not supported by that backend.
{% for cl in p.field_mappings|map(attribute="client")|unique %} {% set tids = p.queries|selectattr("client", "equalto", cl)|rejectattr("method", "equalto", "llm")|map(attribute="target")|list %} {% if cl %}Client {{ cl }} (its profile's field names)
{% endif %}| Sigma field | {% for t in tids %}{{ t }} | {% endfor %}|
|---|---|---|
{{ fm.sigma_field }} |
{% for t in tids %}{% set v = fm.targets.get(t, "") %}
{% if v.startswith('UNSUPPORTED') %}unsupported | {% else %}{{ v }} | {% endif %} {% endfor %}
Rules above joined into one detection by ti2dac code (Sigma correlation plus fixed query templates). The AI never writes the join.
Add one with ti2dac chain {{ r.run_id }} rule-001 rule-002 --within 10m --on host.
{{ c.evidence }}{% endif %} {% if c.warnings %}
{{ q.query }}{% else %}Not available: {{ q.error }}
{% endif %}{{ q.notes }}
{% endif %} {% endfor %}