Repository navigation
✨ dashboard: Evidence panel 4 — boundary enforcement (attempted vs blocked writes per tier; sentinel hits/week) - #12388
Conversation
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
|
hive: this PR now conflicts with 🐝 Hive Agent: — hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 |
0b70f62 to
d7a3409
Compare
Adds the Boundary enforcement evPanel card to the Throughput Evidence sub-panel: attempted-vs-blocked agent writes per intent tier (top block reasons in the tooltip) from write_counters, and a sentinel hits per ISO week sparkline from sentinel_hits_per_week. write_counters in GET /api/evidence/summary now passes through blocked_by_reason. Empty state when the counter file is absent. Reporting only; nothing reads or alters merge eligibility. Rebased onto v5 after #12416: adds the EV_EXPLAINERS entry (graded on proxy-blocked ÷ attempted writes via EV_BLOCKED_STRONG/FAIR), sizes the SVG helpers from EV_SVG_W, and discloses that write_counters are all-time proxy totals while sentinel_hits_per_week follows the selected window. Closes #12368 Signed-off-by: Andy Anderson <andy@clubanderson.com>
d7a3409 to
4d6a517
Compare
|
merged by Hive trusted-author auto-merge: author @clubanderson holds role owner and GitHub permission admin on repo; CI green at 4d6a517. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
Closes #12368 — RFC #12062, proposal B (API + dashboard), panel 4 — Boundary enforcement (the N behind "nothing bad happened").
What
evPanelcard Boundary enforcement in the Throughput section's Evidence sub-panel, between panels 3 and 8. Sameev-panelchrome, inline SVG, no new deps, no inlinestyle=.ADVISORY… 3ISSUES_PRS_MERGE, unknown tier keys after), scaled against the busiest tier, blocked % on the row, and the top 3blocked_by_reasoncodes in the row<title>tooltip.sentinel_hits_per_week, first/last week labels and total / peak in the legend.n=is total attempted writes across tiers.write_countersis omitted (counter file absent) and when no sentinel weeks exist.EvidenceWriteCountinGET /api/evidence/summarynow passes throughblocked_by_reason(omitempty) from the counter file so the tooltip has data;api-reference.mdandreview-evidence.mdupdated. Changelog fragment added.evidence_card_ui_test.gopins the helpers, panel key, tier list, reason handling, empty states, note text, CSS and panel order;evWriteCountersSvg/evSentinelSparklineadded to the ReferenceError guard;TestLoadEvidenceWriteCounterschecksblocked_by_reason.Guardrail (RFC #12062 Non-goals)
Record/read-path only. Nothing in this PR reads into or alters merge eligibility or auto-merge; the proxy counters are read from the file, never written, and tests touch only the summary loader and
index.htmlmarkup.cc @mrbobbytables
Update (rebase onto v5 after #12416):
origin/v5; panel order is now 1 · 2 · 3 · 4 · 5 · 8.EV_EXPLAINERS['boundary-enforcement']ⓘ explainer (Why it matters · What is measured · How it is measured · What good looks like · What ours shows). Graded on proxy-blocked ÷ attempted agent writes via new named constantsEV_BLOCKED_STRONG = 0.05/EV_BLOCKED_FAIR = 0.15; row added to the grade table inreview-evidence.md.sentinel_hits_per_weekis built from the bundles selected by the 24h/7d/30d/custom window (hours/from/to) so it follows the selector.write_counterscome from the proxy's durable counter file which only stores all-time totals (no per-time data), so they cannot follow the window — this is disclosed in the panel note and the explainer rather than hidden.EV_SVG_W; fixed a v5 test comparison (EvidenceWriteCountis no longer comparable onceblocked_by_reasonis a map).