* fix(store): promote-on-approve looks up version_no by submission_id
Live bug observed on agnes-development: an entity had 5+
version_history rows sharing the same `hash` (user re-uploaded
byte-identical bundles as v2/v4/v6 of the same skill — the LLM and
inline checks happily approved each one). The runner's
promote-on-approve path looked up the submission's version_no by
hash:
for entry in entity.version_history:
if entry["hash"] == sub_hash:
target = int(entry["n"]); break
The loop matched the FIRST hash collision — always v1, n=1. With
current=1, the forward-only `target > current` guard then skipped
the promote, leaving the entity stuck at v1 even though the new
submission's status flipped to `approved`. UI kept showing v1 as
"current".
Fix: look up by submission_id via the existing
`_version_no_for_submission` helper (already used by retry / rescan
/ download paths). Same lookup applied in
`admin_override_store_submission` which had the identical hash-match
loop.
Test: TestPromoteLookupByByteIdenticalBundles uploads v1 + a
byte-identical v2, drives the LLM with mock-approve, asserts
entity.version_no advances to 2.
* fix: bundle #329 reviewer-Important follow-ups + post-merge polish
Bundled with Vojtech's commit ahead of this (the promote-on-approve
`version_no` lookup-by-submission_id fix) since #330 is the next
release-cut PR and the four #329 follow-ups would otherwise need a
standalone release-cut PR — prohibited by docs/RELEASING.md §
"Release-cut belongs to the PR".
Fixed:
- src/usage_ask.py — SCHEMA_DIGEST + SYSTEM_PROMPT referenced the
dropped `usage_plugin_daily` table. The admin
`POST /api/admin/telemetry/ask` endpoint ships SYSTEM_PROMPT to
the LLM, so any model-emitted SQL against `usage_plugin_daily`
would fail with a DuckDB binder error post-#329 merge. Updated to
describe the new v48 rollups (`usage_marketplace_item_daily` /
`_window`) and rule 5 of the prompt to point at them.
Internal:
- CHANGELOG.md [0.54.20] section restored to its canonical content
from the v0.54.20 git tag. The #329 self-merge carried 226 lines
of author's pre-rebase bullets that ended up mis-attributed; the
published v0.54.20 GitHub Release (FTS BM25 + batch bar) now
matches the CHANGELOG section verbatim. Also fills in [Unreleased]
with this PR's bullets (Fixed + Internal).
- tests/conftest.py — dropped the unused
`conn_with_usage_schema_and_attribution` fixture that INSERTed
into the now-removed `usage_attribution_*` tables. Zero callers
today, but a tripwire — the first future test to request it would
have failed with a binder error.
- app/web/templates/marketplace.html — replaced a customer-specific
token (`groupon-marketplace`) in the Most Popular sort-tiebreaker
comment with a generic `<customer>-marketplace` placeholder per
CLAUDE.md § Vendor-agnostic OSS. Also scrubbed an `agnes-development`
reference in app/api/admin.py and src/store_guardrails/runner.py
(cherry-picked from Vojtech's commit) on the same hygiene rule.
* release: 0.54.22 — flea-market promote-by-submission_id fix + #329 reviewer follow-ups
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* feat(telemetry): marketplace item rollup refactor (schema v46)
Replace the v42 attribution layer with prefix-split + live lookup against
marketplace_plugins / store_entities. The v42 design had a latent bug —
AttributionLookup keyed on bare skill names while Claude Code writes
`<plugin>:<local>` in JSONL, so lookups never matched and
usage_plugin_daily stayed empty in every deployment.
Schema (v46 migration):
- Drop usage_attribution_skills / _agents / _commands (mapping tables,
derivable from marketplace_plugins + plugin tree).
- Drop usage_plugin_daily (always empty in production due to the bug above).
- Create usage_marketplace_item_daily — per-day fact (count, distinct_users,
error_count), composite PK on (day, source, type, parent_plugin, name).
- Create usage_marketplace_item_window — sliding-window snapshot with
true cross-window distinct user counts; period_label='last_7d' refreshes
every tick, 'last_30d' refreshes hourly (tracked via session_processor_state).
- Mark usage_tool_daily as candidate for removal (no product-UI consumer).
Attribution flow:
- MarketplaceItemLookup replaces AttributionLookup. Preloads
marketplace_plugins.name + store_entities.name into memory once per
UsageProcessor tick, then per-event splits identifier on ':',
matches prefix, writes resolved source / parent_plugin into
usage_events. agnes-store-bundle prefix routes to flea entities.
Slash commands with `plugin:` prefix count as type='skill' in rollup.
API:
- BREAKING: MarketplaceItem.unique_users_30d renamed to distinct_users_30d
(now a true distinct count from the window snapshot, not sum-of-daily).
- InnerDetailResponse gains a telemetry field — invocations_30d +
distinct_users_30d surfaced on curated inner skill / agent detail pages.
- Card chip hidden pending UX finalisation; data stays in the response.
Backfill: scripts/backfill_marketplace_rollup.py — one-shot rebuild over
historic usage_events after deploy, idempotent.
USAGE_PROCESSOR_VERSION bumped 4 → 5 so the reprocess loop re-attributes
existing events to the new source/ref_id semantics on the next tick.
Tests rewritten: test_session_processor_usage, test_usage_rollups,
test_marketplace_telemetry, test_api_admin_usage_reprocess,
test_db_schema_version, test_home_stats, test_schema_v42_migration.
New: test_backfill_marketplace_rollup.
* fix(marketplace): refresh Most Popular on search + category changes
`loadMostPopular()` early-exits when `state.q` or `state.category` is
set, but the search + category handlers only called `loadItems()` —
so once the section was visible, typing a query or filtering by
category didn't re-run the hide check and the cards stayed on screen
out of scope. Tab + sort handlers already chained the call.
Add the call to runSearch + category pill click handlers (All +
per-category) so the visibility contract holds for every state
mutation that can flip the early-exit condition.
* feat(marketplace): All-plugins section + 7-day Most Popular
Listing layout:
- Always-visible "All plugins" / "All items" / "Your stack" section
header (label swaps per tab) wrapped in `#mp-all-section` so its
margin-collapse mirrors the sibling `#mp-popular-section` and the
spacing from the filter row stays consistent in both layouts.
- Sort dropdown moved from the filter row into the All-* header,
pinned right via `margin-left: auto`. Anchored to its section so
the relationship between sort + grid is obvious.
- `.mp-section-header` gets `min-height: 32px` + `align-items: center`
so the bare-text Most Popular row matches the dropdown-bearing
All-* row.
- `.mp-section-header` margin tightened 24px → 20px on top.
Most Popular:
- Capacity reduced 8 → 4 cards.
- Now reflects a 7-day window (was 30-day). Backend surfaces
`invocations_7d` + `distinct_users_7d` on `MarketplaceItem`
alongside the existing 30d fields; the loader pulls a wider page
(server still sorts by 30d) and re-sorts + filters client-side
on `invocations_7d > 0` so the strip stays "hot right now".
- Section label updated to "Last 7 days".
- Section now renders on both `curated` and `flea` tabs (was
curated-only). Hidden on `my` and whenever search / category
filter is active. Refresh hooks wired into search + category
click handlers so visibility flips immediately on state change.
Backend (`_load_invocation_stats`):
- Single SELECT pulls both `last_30d` and `last_7d` rows from
`usage_marketplace_item_window`; the result dict carries
invocations + distinct_users for both windows.
- Trend (recent_7 vs prior_7) kept on the daily fact table so it
stays independent of the window snapshot's freshness.
* feat(marketplace): Most adopted sort + hide Trending when no trend data
Add a fourth sort option to the All-items dropdown — "Most adopted
(30d)", keyed on `MarketplaceItem.distinct_users_30d` (true 30d
distinct user count from `usage_marketplace_item_window`). Protects
the listing from power-user skew that `most_used` is susceptible to:
one user × 100 invokes can't beat 10 different users × 1 invoke
under adoption sort.
Hide Trending option when the response has no trend data. User
reported `sort=trending` returning an empty grid because every
plugin's `trend_pct` was None (prior-week threshold of >= 3
invocations didn't clear anywhere). Empty grids on a user-selected
sort are worse UX than just not offering the sort — surface what
works, hide what doesn't.
Backend (`app/api/marketplace.py`):
- `_apply_sort` gains a `most_adopted` branch (DESC distinct_users_30d,
ties by name ASC).
- `sort` Literal extended.
- `ItemListResponse.available_sorts` lists the sort keys the UI
should expose for this response. recent/most_used/most_adopted
always; trending only when at least one item in the tab's stats
carries a non-null trend_pct.
- `_available_sorts(stats_dicts)` helper centralises the rule —
curated and flea branches pass one stats dict, my-tab passes both
(option is available when either source has trend data).
Frontend (`app/web/templates/marketplace.html`):
- New `<option value="most_adopted">Most adopted (30d)</option>`
between Most used and Trending.
- URL state allowlist extended so `?sort=most_adopted` round-trips.
- `applyAvailableSorts(available)` runs after each list fetch:
hides options not in the response's available_sorts; if the user
is on a now-unavailable sort, resets to 'recent' and re-fetches.
Search-mode fan-out unions availability across the curated + flea
responses so a hit on either side keeps the option visible.
* feat(marketplace): funnel chip on cards + deterministic Most Popular sort
Card chip — funnel telemetry between description and footer:
[stack-icon] N installed · [user-icon] N active · [bolt-icon] N calls · ↑/↓ N%
- stack_count (new MarketplaceItem field): for curated it's COUNT(*)
on user_plugin_optouts (post-v28 row PRESENCE = subscribed; system
plugins are fanned out to every user via fanout_system_for_user so
the count includes them naturally). For flea it reuses the existing
store_entities.install_count (bumped on install/uninstall).
- distinct_users_30d (existing) — active users in the 30d window.
- invocations_30d (existing) — call volume.
- trend_pct (existing) — week-over-week, both directions: green ↑ /
red ↓, magnitude only (sign in the arrow). Hidden when null.
Backend additions in app/api/marketplace.py:
- MarketplaceItem.stack_count field.
- _load_curated_stack_counts() — one SELECT per render, GROUP BY
(marketplace_id, plugin_name). Wired into the curated + my-tab
branches; flea reads install_count off the entity row directly.
Frontend (app/web/templates/marketplace.html):
- Heroicons solid 24×24 inlined (one helper per icon, all
fill="currentColor" so per-segment colour tokens apply): rectangle-
stack (mirrors the My Stack tab icon), user, bolt, arrow-trending-
up/down.
- Per-segment colour: installed=amber #F59F0A (My Stack accent),
active=green #0e9b6a, calls=orange #f97316. Text stays neutral so
the chip still reads as metadata, the leading glyph carries the
visual cue. Trend pill keeps the full-segment green/red colour.
- Zero state: chip hidden when stack_count == 0 AND invocations_30d
== 0 — brand-new cards aren't visually penalised by a "0·0·0" row.
- Tooltips on every segment via title="…" so hover explains the
number's meaning to anyone uncertain about the icon.
Most Popular section — deterministic ordering:
Previously sorted by invocations_7d DESC with no tie-breakers, so
several cards with identical 7d call counts would swap places on
refresh (JS stable sort fell back on backend order, and the backend's
own tie-breaker for `most_used` was just name ASC — six `grpn`
plugins from six test marketplaces collapse to the same name and
became indeterminate via list_with_filters' created_at order).
New cascading hierarchy (chosen primary now matches what "most
popular" really means — wide adoption, not power-user volume):
1. distinct_users_7d DESC ← adoption / social proof
2. invocations_7d DESC ← volume at equal adoption
3. distinct_users_30d DESC ← broader adoption fallback
4. invocations_30d DESC ← broader volume fallback
5. name ASC ← deterministic textual order
6. marketplace_slug ASC ← splits duplicate plugin names across
marketplaces
Six levels guarantee any two items end at a different sort key, so
the strip is stable across refreshes.
* fix(marketplace): unify Most Popular on 30d + right-align installed chip
Most Popular section was sorting on the 7d window while its cards
rendered 30d numbers — header label promised one thing, cards showed
another. Unified everything on 30d so a card means the same data
everywhere on the page.
- Dropped the "Last 7 days" meta from the Most Popular header.
- Sort cascade now starts on distinct_users_30d, then invocations_30d,
with 7d adoption/volume as recency-aware fallbacks before the name +
marketplace_slug deterministic tail. Six levels guarantee identical
sort keys never produce indeterminate order across refreshes.
- Filter switched from invocations_7d > 0 to invocations_30d > 0 to
match the new horizon.
- Most Popular now only renders on page 1 of the listing. Past initial
discovery, a top-of-list popularity strip on page 2+ would shadow the
results the user paged into. Pager click handler refreshes the
section so navigating back to page 1 re-mounts it.
Chip layout — split engagement vs adoption visually:
[user] N active · [bolt] N calls · [↑/↓] N% [stack] N installed
└────────── LEFT (time-bounded engagement) ────┘ └── RIGHT (all-time) ──┘
- Installed (stack_count) is all-time, decremented on uninstall. Alone
it says little ("12 people installed it") without the engagement
context next to it ("…but did anyone actually use it?"). Visually
separating the two groups makes that distinction obvious — left
group answers "is it used", right answers "does anyone have it".
- Implemented via flex with margin-left:auto on .seg-installed so
installed drifts to the trailing edge.
- Installed tooltip now reads "Currently installed by N users" — the
count is a real-time net (uninstall drops it), and saying "currently"
makes that explicit. Helps when a card shows 0: signals "nobody has
this in their stack right now", not "data missing".
* feat(plugin-detail): telemetry chip in hero, derived rows in sidebar
Surface the same telemetry funnel the listing card carries on the
curated plugin detail page, so clicking through from /marketplace
keeps a single mental model — figures match, semantics match. The
detail sidebar drops the two raw numbers that used to live there
(Invocations 30d / Users 30d — duplicated by the chip now) and
replaces them with two *derived* signals only the daily series can
provide: Active days + Last used.
Backend (app/api/marketplace.py):
- PluginDetailResponse.stack_count — curated reads via
_load_curated_stack_counts(), flea reuses install_count. Frontend
treats both sources uniformly.
- _build_telemetry() always returns a dict (never None). Frontend
decides chip visibility from stack_count + invocations_30d the
same way the listing card does. daily_series is always 30 entries
(zero-padded) so "Active days" and "Last used" derivations on the
sidebar are trivial array filters.
Frontend (app/web/templates/marketplace_plugin_detail.html):
- New .hero-telemetry slot at the bottom of the hero meta column,
between the pills row and the action buttons. Renders the four
funnel segments — active · calls · trend · installed — joined by
` · `. No left/right split: the hero has space, so a single
coherent metadata strip reads cleaner than the card's split layout.
- Heroicons solid inlined (user / bolt / arrow-trending-up,-down /
rectangle-stack) recoloured against the dark hero — icons in
lighter tokens (mint #6ee7b7, peach #fdba74, cream #fde68a), trend
pill keeps the saturated green/red because direction-coding earns
its own colour.
- Tooltip on installed reads "Currently installed by N users" — the
count is a real-time net (drops on uninstall), and "currently"
makes that explicit when a card shows 0.
- fmtNum helper added so 1.2k / 14M renderings match the card's
format exactly.
- Sidebar swap: Invocations + Users rows removed, replaced by
Active days → "N of 30"
Last used → fmtRelative of the latest non-zero day
Both derived from telemetry.daily_series — engagement consistency
+ recency, neither of which the hero chip exposes on its own.
* feat(item-detail): telemetry chip in hero for curated skill/agent
Bring the funnel chip the plugin detail page got in 4cf38d40 to the
curated inner skill/agent detail page — clicking through from the
listing card now keeps the same metadata strip from grid to plugin
page to inner item page.
Backend (app/api/marketplace.py):
- _load_inner_item_stats() rewritten:
* always returns a dict (never None) so the frontend can decide
chip visibility client-side, same contract as _build_telemetry
* adds trend_pct, computed the same way as plugin level
(recent_7 vs prior_7 from usage_marketplace_item_daily, ≥3
prior-week threshold)
* adds daily_series (30 entries, zero-padded) so the sidebar can
derive Active days + Last used
- InnerDetailResponse.parent_stack_count — new field. Skills/agents
don't have a per-item subscription model, so the hero shows the
*parent plugin's* stack count under a "Plugin:" prefix. The
funnel: "12 installed plugin → 2 actually use this skill".
- curated_skill_detail + curated_agent_detail handlers load
_load_curated_stack_counts() once and pass the parent's value.
Frontend (app/web/templates/marketplace_item_detail.html):
- New .item-detail .hero .hero-telemetry slot beneath the badges
row. CSS mirrors plugin-detail's colour tokens (mint/peach/cream
Heroicons solid + saturated trend pill) so the two surfaces read
as one visual family.
- Installed segment uses a "Plugin:" label rendered with reduced
opacity to signal the metric describes the parent, not the item
itself. Tooltip: "Parent plugin (<plugin_name>) currently
installed by N users".
- Sidebar Invocations + Users rows removed (chip carries them).
Active days + Last used derived from telemetry.daily_series replace
them; only rendered when activeDays > 0 so a brand-new skill
doesn't show "0 of 30" / "Last used —".
- "Type" row dropped from the sidebar — duplicates the hero badge.
- fmtNum helper added (matches listing card + plugin detail).
Plugin detail (app/web/templates/marketplace_plugin_detail.html):
- Hero "Curator: …" line removed. The Details sidebar already
carries that info; duplicating it under the h1 was visual noise.
- Sidebar "Owner" row renamed to "Curator" — for curated plugins
it's a person who curates inclusion in this Agnes instance, not
the upstream code owner. "Owner" was a hold-over label.
* feat(item-detail): unify hero with plugin detail — pills + breadcrumb + cleaner sidebar
- Inner skill/agent hero now uses the same `.pills` / `.pill.cat / .curated /
.flea / .muted` class names + CSS as the plugin detail page; the only
item-only addition is `.pill.type` (Skill / Agent uppercase, plugin detail
has no kind axis).
- Hero `Updated` moved out of the meta-row into a muted pill (mirrors the
plugin detail hero), removed from the Details sidebar to avoid duplication.
- Details sidebar slimmed: dropped Marketplace, Path, Updated rows; Parent
plugin now shows the curator-friendly display name
(`parent_display_name || manifest_name || slug`) instead of the slug.
- Breadcrumb extended to full path: Marketplace > <marketplace_name> >
<plugin display name> > <self>, mirroring the plugin detail breadcrumb.
- Backend: new `InnerDetailResponse.parent_display_name` field, populated via
`_curated_plugin_enrichment` from marketplace-metadata.json — same source
plugin detail hero already uses.
* feat(marketplace): flea inner skill/agent detail + breadcrumb polish
- Flea inner skill/agent detail page parity with curated:
* GET /api/marketplace/flea/{id}/skill/{name} + /agent/{name}
returning InnerDetailResponse (mirror of curated_skill_detail).
* /marketplace/flea/{id}/skill|agent/{name} web routes that render
marketplace_item_detail.html with source='flea' + innerName context.
* Frontend apiURL grows a third branch for flea-inner; breadcrumb
grows to 4 segments (Marketplace > Flea Market > <plugin display
name> > <self>) when innerName is set.
* Telemetry attribution: MarketplaceItemLookup resolves
<flea_plugin>:<inner> prefixes to (source='flea',
parent_plugin=<plugin name>) so nested invocations land in the
same rollups curated nested skills use. USAGE_PROCESSOR_VERSION
bumped 5 -> 6 so the reprocess loop re-attributes historic events.
- Breadcrumb 2nd segment is now a generic clickable "Curated
Marketplace" / "Flea Market" link to /marketplace?tab=... instead
of the opaque per-instance marketplace_name. Applied on both plugin
detail and inner item detail.
- Inner item hero telemetry chip works for both sources: installedCount
branches on parent_stack_count (curated) vs install_count (flea),
installed segment drops the "Plugin:" prefix for flea standalone /
inner items.
- Updated row dropped from Details sidebar on item detail — the hero
pill already carries the value, sidebar row was duplicate.
* feat(item-detail): block stack-install on flea inner items (mirror curated)
Inner skills/agents nested inside a flea plugin can no longer be added
to a user's stack on their own — adoption only happens at the plugin
level, same rule curated nested items have followed since launch.
- Hero action: when innerName is set (curated nested OR flea nested),
render "Open parent plugin →" link + helper text instead of the
install/remove buttons. Flea standalone entities (no innerName) keep
the normal install UX.
- Meta-row: same branch now serves curated + flea inner — "part of
<parent plugin display name> · by <author>" with the parent link
pointing at the right detail page per source.
No API gate change needed: POST /api/store/entities/{id}/install only
accepts existing entity ids (plugin-level), inner items have no entity
id of their own so the endpoint cannot target them directly.
* feat(marketplace): telemetry chip on inner cards + fix flea hero chip visibility
Inner skill/agent cards on the plugin detail page now carry the same
four-segment funnel chip the marketplace listing cards show (N active
. N calls . trend . N installed), for both curated nested skills and
flea nested skills. Plus two fixes that were keeping the hero chip
hidden on flea plugin / flea inner detail pages.
- Backend `_load_inner_items_stats_by_parent(conn, source, parent_plugin)`
bulk loader: one query per plugin against usage_marketplace_item_window
+ one against _daily, returning {(name, type): stats}. Avoids N+1
per-card lookups.
- `InnerItemSummary` gains invocations_30d / distinct_users_30d /
trend_pct / parent_stack_count fields. `curated_detail` and
`flea_detail` (in the entity.type=='plugin' branch) enrich the
skills / agents lists after the existing cover-photo enrichment loop.
- `marketplace_plugin_detail.html`: new `.plugin-detail .inner-card
.inv-chip*` CSS lifted from marketplace.html with the listing-card
rules, new buildInnerCardChip() helper, buildCardSection appends
the chip to each card body. Same gate as the listing card (hidden
on parent_stack==0 && calls==0).
- fix(flea): flea_detail forgot to populate PluginDetailResponse.stack_count
from entity.install_count (listing card does this on line 851; detail
endpoint didn't). Hero chip gate `stackCount===0 && calls===0` then
always hid the chip even when the entity had installs. Now mirrors
listing card semantics: stack_count == install_count for flea.
- fix(flea inner): renderInnerHeroTelemetry was reading `d.install_count`
for any non-curated source. InnerDetailResponse has no install_count
field — it has parent_stack_count (populated server-side from the
parent flea plugin's install_count). Gate + label now read
parent_stack_count for both curated nested AND flea nested scenarios;
install_count remains the flea standalone path.
* fix(marketplace): Owner label on flea + parent-centric sidebar for flea inner
- Plugin detail Details sidebar — authorship row label now tracks the
source: curated bundles get `Curator` (existing behaviour), flea
bundles get `Owner`. The `owner_todo` reminder placeholder stays on
the curated branch only; flea falls through silently.
- Inner item detail Details sidebar — flea-inner (skill/agent nested
inside a flea plugin) now shares the curated nested layout: Parent
plugin / Bundle size / Active days / Last used / Owner. Drops the
flea-standalone shape's `Category`, `Version`, `Installs`, `Released`
rows that didn't apply to a nested item. Active days + Last used were
already wired (telemetryRows) — they just weren't on the flea-inner
branch.
* fix(tests): bump SCHEMA_VERSION assertions 47 -> 48 post-rebase
The marketplace telemetry migration was renamed _v46_to_v47 -> _v47_to_v48
during the rebase onto main (collision with #326 FTS BM25 migration that
took the v47 slot). Two test files still asserted the pre-rebase value:
- tests/test_home_stats.py::test_schema_version_constant_is_46 (CI red)
- tests/test_schema_v46_migration.py::test_schema_version_is_46
Renames the helper fn name + bumps the assertion. The other two test
files (test_db_schema_version.py, test_schema_v42_migration.py) were
already updated in the rebase resolution.
* fix(telemetry): _build_telemetry returns None when invocations_30d == 0
The follow-up commit that introduced the always-return-dict shape broke
the test contract from the original v46 PR (commit b603e998):
tests/test_marketplace_telemetry.py::TestDetailTelemetry::
test_detail_endpoint_telemetry_absent_when_no_data
AssertionError: assert {'daily_series': [...], ...} is None
Both `PluginDetailResponse.telemetry` and `InnerDetailResponse.telemetry`
are declared `Optional[Dict] = None`, the frontend renders are None-safe
(`d.telemetry || {}` guard + `if (!d.telemetry || ...)` on daily_series),
so dropping the dict on zero activity is the cleaner default.
* release: 0.54.21 — marketplace telemetry refactor (schema v48) + flea inner detail parity + listing UX polish
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
Follow-up to #62 / PR #126 — that PR shipped the bulk-edit batch bar
in the Review tab only and deferred the symmetric bar on All Items.
This adds it.
Scope:
- New .batch-bar block in #tab-all with selectAllAll, selectedCountAll,
and the five bulk-edit buttons (Move to category / Move to domain /
Add tag / Remove tag / Set audience).
- renderItemCard signature: third param widened from boolean isReview
to a 'review' | 'all' | 'browse' mode enum so the Browse tab's
call site can explicitly suppress the row checkbox. The earlier
iteration of this PR widened the checkbox condition without
auditing other callers, which left the Browse tab with orphan
checkboxes that fired updateSelectionCount('all') against an
invisible tab. Adversarial-review fix.
- updateSelectionCount('all') toggles the *BtnAll set; renderAllItems
resets the header checkbox + recomputes counts on every re-render so
stale selection state can't survive a list refresh.
Approve / Reject stay scoped to Review per the issue's scope decision
— status-change actions belong with the per-row action buttons or the
keyboard workflow in Review.
Existing JS plumbing already assumed tab-aware selection
(getSelectedIds(tab), toggleSelectAll(tab), openBulkEditModal reads
currentTab); the All-items DOM and the *BtnAll ID set are the only
additions.
Tests in tests/test_admin_memory_page_all_items_batch_bar.py:
- test_admin_page_renders_all_items_batch_bar — all five button IDs
+ the select-all checkbox + the toggleSelectAll('all') callback
are present on the rendered admin page.
- test_all_items_bar_omits_approve_reject — Review-only Approve /
Reject IDs do not appear with the All suffix (scope guard).
- test_browse_tab_omits_row_checkbox — regression guard for the
Browse-tab orphan checkbox: confirms the call site uses 'browse'
mode and renderItemCard omits the checkbox markup on that branch.
Closes#129.
* fix(store): surface review failures + harden publish gate
Four independent fixes to the flea-market submission pipeline, all surfaced
by an admin upload that landed at status='approved' without an LLM review.
1. LLM truncation no longer pins submissions in review_error.
- Raised MAX_RESPONSE_TOKENS 2500 → 6000 in llm_review.py
- Added one-shot retry-with-doubled-budget in anthropic_provider.py
(capped at 4× initial)
2. Flea detail page surfaces the latest submission's failure verdict even
when a previously-approved version is still serving (deferred-promotion
path). The _quarantine_banner gate widened from `visibility != approved`
to also fire on `blocked_inline / blocked_llm / review_error`, with copy
that distinguishes the v2+ edit case ("Latest edit failed review —
previously approved version (vN) keeps serving") from the initial-upload
quarantine wording.
3. Restore button + endpoint no longer allow restoring a version that was
never approved. Added StoreEntitiesRepository.get_with_version_approvals
joining store_submissions, gated the UI button on submission_status in
('approved', None), rendered status pills for non-restorable rows, and
added a 400 version_not_approved guard in POST /restore.
4. **BREAKING (operator-facing)**: publish gate is now fail-CLOSED on
misconfig. The previous get_guardrails_enabled() silently fell back to
"disabled, auto-approve everything" when guardrails.enabled=true in YAML
but no ANTHROPIC_API_KEY was in env. Split into:
- get_guardrails_enabled() (intent — YAML)
- get_guardrails_llm_provider_ready() (readiness — env)
Three-state matrix:
enabled=false → auto-approve (unchanged)
enabled=true + ready=true → normal pipeline (unchanged)
enabled=true + ready=false (NEW) → submissions hold at pending_llm
awaiting admin retry or override
(was: silent auto-approve)
Admin "Retry review" eligibility broadened to include pending_llm.
Boot-time WARNING banner surfaces the misconfig in app/main.py.
docs/STORE_GUARDRAILS.md updated with the three-state matrix.
Operators relying on the auto-fallback for local-dev no-LLM setups must
now explicitly set `guardrails.enabled: false` in instance.yaml.
Tests: 4623 passed. Added TestPublishGateFailClosed (4 tests) and
TestRestoreVersion::test_restore_rejects_* (3 tests). conftest.py adds an
autouse fixture defaulting guardrails OFF so legacy tests don't need to
know about the new toggle.
* fix(store): admin override promotes v2+ edits to current
The override handler at app/api/admin.py:3708 only flipped submission
status → 'overridden' and entity visibility → 'approved'. Under the v37+
deferred-promotion model that's insufficient for v2+ edits / restores:
the new bundle sits in versions/v<N>/plugin/ and the entity row stays at
the prior approved version_no + hash + on-disk live bundle. Installers
kept getting the OLD bytes the admin had just intended to replace.
Mirror the runner.run_llm_review auto-approval branch: look up the
submission's version_hash in entity.version_history, and if its `n`
differs from entity.version_no, promote_version + _swap_live_to_version.
Initial v1 overrides are unaffected — the loop finds n=1 == version_no
and skips promotion.
Tests:
- test_override_v2_edit_promotes_to_current: stage v1 approved + v2
blocked_llm; override the v2 sub; assert entity.version_no=2,
entity.version flips off the v1 hash, and the live plugin/ dir
mirrors versions/v2/plugin/.
- test_override_v1_initial_upload_no_promote: regression guard so the
promote loop doesn't accidentally bump a v1 override.
Audit log gains a promoted_to_version_no field on the override action.
* fix(store): retry/rescan review staged bundle; override forward-only
Two adversarial-review findings from a Codex pass on the publish-gate
work.
C1. Admin retry + rescan were passing live `plugin/` to the LLM. For a
v2+ submission held at `pending_llm` / `blocked_llm` / `review_error`,
live still holds the prior approved version's bytes — so the LLM
reviewed the WRONG bytes, and the runner's hash-match promotion in
`run_llm_review` would then advance the entity to staged bytes that
were never actually reviewed. Resolve the staged
`<entity>/versions/v<N>/plugin/` from the submission's
`version_history` entry, with a fall-back to live for legacy pre-v37
rows that never seeded a versions/ dir. Helpers
`_submission_plugin_dir` and `_version_no_for_submission` added to
`app/api/store.py` so override / retry / rescan share one path.
H1. Override's promote loop used `target != current`, which would
silently demote the live bundle when admin overrode a stale v2
submission while v3 was already approved + live. Changed to
`target > current` so override flips status + visibility on the row
regardless, but on-disk promotion only fires forward. Same `>`
defensive guard applied in `runner.run_llm_review` so a late LLM
verdict racing with a newer approval can't demote either.
Tests:
- TestAdminRetryReviewsStagedBundle::test_retry_v2_blocked_passes_staged_dir_not_live
- TestAdminRetryReviewsStagedBundle::test_rescan_v2_blocked_passes_staged_dir_not_live
- TestOverrideForwardOnly::test_override_stale_v2_does_not_demote_when_v3_current
* review polish: CHANGELOG drift, override eligibility, defensive copy
Three small additions on top of the retry/rescan staged-bundle fix:
1. CHANGELOG: the PR's bullets had drifted into the released
[0.54.17] section during rebase (context-match landed them next
to already-released content). Moved them up to [Unreleased] where
they belong; [0.54.17] now holds only what was actually released
(refresh-marketplace ls-remote, /me/activity hero, CI sharding +
workflow polish).
2. app/api/admin.py: admin override eligibility now accepts
pending_llm alongside blocked_inline + blocked_llm + review_error.
Closes a UX gap from the new fail-CLOSED behavior: under
enabled-but-not-ready, a known-good submission would otherwise
sit indefinitely until the admin set credentials AND clicked
Retry. Override already routes through version_history (and is
now forward-only on promote), so it stays safe for v2+ deferred-
promotion submissions.
3. src/repositories/store_entities.py: get_with_version_approvals
defensively copies each version_history entry before annotating
with submission_status. self.get() re-parses JSON each call today
so this is belt-and-suspenders against any future caching layer
leaking the annotated key into a subsequent plain get() call.
Tests: 112 passed (focused on test_store_entity_versions +
test_admin_store_submissions, covering the retry/rescan staged-
bundle fix the author shipped + this polish).
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
The subtitle was built by ~-concatenating a Markup operand
(user.email | e) with HTML string literals. Under autoescaping,
Jinja2's markup_join escapes every non-Markup part once it hits a
Markup operand — so the literal <strong> tags became <strong>
and the page showed literal "<strong>...</strong>" text around the
email. The | safe in _page_hero.html was too late to undo it.
Switch to {% set %}...{% endset %} block capture: the literal
<strong> stays HTML while {{ user.email }} is still autoescaped.
Regression test asserts the tags render and a hostile email stays
escaped.
The wizard nav buttons used class="btn-primary" / "btn-secondary"
without the .btn base class, so the padding (10px 20px),
border-radius (8px), font-size, and inline-flex centering rules from
.btn never applied. Buttons rendered as ~18px-tall colored boxes with
no padding (visible mismatch against the sibling Cancel <a> which
correctly used class="btn btn-secondary").
Added .btn to all three buttons (#next-btn, #back-btn, #finish-btn).
No CSS change — purely a markup fix.
Playwright before: next.padding="0px" borderRadius="0px" height=18
Playwright after: next.padding="10px 20px" borderRadius="8px" height=38
Consolidates the scattered per-analyst pages into /me/activity (usage
analytics) and /me/profile (account hub). /me/stats and /profile/sessions
301-redirect; /profile, /me/debug, /tokens are removed with every internal
link repointed. Includes an XSS fix in the /me/activity page hero, the
user_id-keyed session-lookup alignment, and the v0.54.15 release cut.
Co-developed by @ZdenekSrotyr and @cvrysanek.
Cuts 0.54.14. Repairs the [Unreleased] changelog state left by three PRs that merged since v0.54.13 without proper changelog hygiene, sweeps dead code orphaned by #305 and earlier PRs, and bumps the version.
- CHANGELOG: #307 bullets moved out of the released [0.54.10] section into [Unreleased]; backfilled missing entries for #305 (Removed) and #308 (Changed); [Unreleased] -> [0.54.14] release cut.
- pyproject.toml: 0.54.13 -> 0.54.14.
- app/web/router.py: removed orphaned gws_oauth / instance_admin_email / connector_prompts keys from the shared _build_context ctx dict.
- app/web/templates/home_not_onboarded.html: swept dead connector-tile, automode, and setup-collapsible/minimize CSS + the orphaned JS click-wiring.
Three coordinated tweaks to the publication discovery surface:
1. Action-row CTA on /marketplace?tab=curated reads 'Submit a skill
or plugin' instead of 'Submit a plugin'. Skills are first-class
citizens of the curated shelf; the old wording made them feel
like an afterthought. Same rename in the empty-state JS innerHTML
so the two paths can't drift.
2. Curated guide page (/marketplace/guide/curated) expanded from a
4-line stub into a 3-step ordered list documenting the Named
Curator handoff (find curator → handoff → publish + lifecycle).
New '.guide-fastpath' callout block points users at the Flea
Market when they want lighter review-bar / faster path. Primary
CTA at the bottom of the curated guide now links to the flea
guide too, so users who skim past the fast-path callout still
see the escape hatch.
3. Flea guide page (/marketplace/guide/flea) expanded from a
3-line stub into a 4-step ordered list (package → upload via
form → automated review → published). Documents the actual
/store/new flow + the automated guardrails (manifest, content
quality, prompt-injection scan) so users know what 'self-service'
actually means before they upload.
Route titles updated to match: 'Submit a skill or plugin to
Curated Marketplace'.
New file: tests/test_web_marketplace_guide.py — three tests covering
the CTA rename, the curated guide's structural elements (Named
Curators lede, 3 steps, fastpath callout, primary-CTA href), and
the flea guide's structural elements (4 steps, no fastpath
asymmetry, /store/new primary CTA).
The dedicated `<details data-section="connectors">` section on /home
duplicated content that the install hero's Step 4 clipboard payload
already inlines. Both surfaces sourced the same prompt strings from
`app/web/connector_prompts.py` (home tiles via `<code id="*-prompt">`,
setup script via `app/web/setup_instructions.py::_connectors_block`),
so users walking the install script visited each connector inline and
then had no reason to scroll back up.
Removed the full block (3 tiles + summary + section-label). Lead
paragraph in the install hero now mentions the connector families
briefly so the benefit is visible before kick-off:
"... your team's curated data, plugins, third-party tools (Asana,
Google Workspace, Atlassian), and shared knowledge ... the install
script also connects your tools for you, so there's no extra page
to visit."
The "Email admin" mailto CTA, previously gated inside the GWS tile
when admin_email was set + GWS unconfigured, moves implicitly to the
install script's GWS step (Claude prompts the user when the OAuth
gating wall lands). Tests updated:
- test_connectors_section_removed_from_home (renamed from
test_connectors_render_flat_when_onboarded_by_default) — asserts
`class="connector-tiles"` and `data-section="connectors"` are absent
in BOTH onboarded states, and that the lead paragraph still mentions
the three connector families so the benefit isn't lost.
- test_home_renders_connector_prompts_from_shared_module — DROPPED.
Was a parity check between the home tiles and the setup script's
connector_prompts.py source. One surface now → no drift risk → test
redundant. Replaced with an inline comment pointing future readers
at where the strings flow (setup_instructions.py::_connectors_block).
- test_home_no_longer_shows_email_admin_button (renamed from
test_home_shows_email_admin_button_when_admin_email_set_and_gws_unconfigured)
— asserts the mailto CTA is gone from /home regardless of
admin_email / GWS-configured state; documents the path-move.
CSS for `.connector-tile*` left in place as dead bytes — small
footprint, no behavior, easy follow-up if/when someone audits.
Each catalog_data bucket renders as its own top-level Data Package card instead of nested accordions under a single Core Business Data wrapper. Tables flat-listed per package, mirroring the Agnes Internal card. Pluralization fix (1 table / N tables).
Includes the 0.54.11 release-cut.
Co-authored-by: pcernik-grpn <pcernik-grpn@users.noreply.github.com>
* docs(plan): design-system unification plan (post-review revisions)
Plan covers consolidating two CSS files into one, introducing
canonical primitives (.btn family, .search-input, .filter-bar,
.page-header, .data-table, .empty-state, .toast, .stat-card,
.tab-strip), unifying the top-nav Admin trigger with sibling
links, and migrating 41 templates that today carry inline
<style> blocks.
Post-review revisions: nav fix moved to first commit (user
complaint lands first); sticky-header and dark-mode skeleton
tasks dropped (defer to follow-up PRs); contract test class
detection tokenizes class="..." attributes properly; baseline
screenshot loop added to Task 0; vendor-token grep widened.
* fix(nav): unify Admin trigger with sibling nav links
The top-nav Admin entry is a <button class="app-nav-link
app-nav-menu-trigger">, siblings are <a class="app-nav-link">.
.app-nav-menu-trigger used to override .app-nav-link with
"color: inherit; font: inherit", resetting font-size from 13px
back to body default and color from --text-secondary to body
color. Active state diverged too: .is-active on links used
--primary blue, [aria-expanded=true] on the button used
--border-light grey.
Fix: expand .app-nav-link so it covers <button>-element resets
(font-family: inherit, border: 0, background: transparent,
cursor: pointer, display: inline-flex for chevron alignment).
Add [aria-expanded="true"] as another active-state selector
so the dropdown's open state highlights identically to .is-active
on links. Delete the now-redundant .app-nav-menu-trigger rules
that stripped button chrome.
Extract the inline <script> from _app_header.html into a new
app/web/static/app.js (loaded by base.html only — base_login.html
has no nav). Sets up window.appUI.wireDropdown for both the user
menu and the Admin dropdown via DOMContentLoaded.
* style(css): consolidate style.css into style-custom.css + add cache-bust
One stylesheet for the whole web UI:
- style.css (1086 lines, legacy Google-inspired tokens + components)
absorbed into style-custom.css under a labeled block, placed after
the modern :root + body so style-custom's component rules continue
to override the legacy ones (preserves the original cascade order
that came from loading style.css first).
- style.css deleted; <link> dropped from base.html + base_login.html.
- static_url() now appends ?v=<mtime> to /static/<path>. Cheap
per-request os.stat — auto-invalidates browser + proxy caches on
redeploy without operator intervention. Mtime survives across
uvicorn restarts as long as the file content is unchanged.
Legacy classes (.btn, .card, .login-*, .badge, .code-block, .flash,
.form-group, .username-box, .btn-copy, .auth-tabs, .divider, etc.)
still render — they live in style-custom.css now. Login pages,
error page, password setup, and the dashboard's Claude Code Setup
card all kept working in browser smoke.
* test(design): contract test for design-system invariants
7 structural invariants enforced from this commit onwards:
- style.css must stay deleted
- no template links style.css via static_url
- exactly one bare :root block in style-custom.css
- canonical primitives declared (.btn, .btn-primary, .search-input,
.filter-bar, .page-header, .data-table, .empty-state, .toast, …)
- no deprecated class names in templates (.users-table, .gp-table,
.marketplaces-table, .audit-table, .users-search, .marketplaces-search,
.modal-btn, .btn-primary-v2, …)
- app.js loaded by base.html, NOT by base_login.html
- 3 helper-level unit tests for the class-attribute tokenizer
(multi-line attrs, Jinja-conditional fragments, false-positive prose)
Two of the assertions intentionally start FAILING after this commit
(missing primitives + legacy class refs in 7 admin templates) and
will turn green as Tasks 4–7 add primitives and Tasks 8–15 migrate
the templates.
* feat(css): canonical button family + legacy token aliases
Adds at top of :root: legacy token aliases (--bg, --card-bg, --text,
--text-light, --secondary, --radius) pointing at modern equivalents.
Absorbed style.css rules referenced these names; without aliases
they fell back to 'unset'. Aliases live until Task 16 alongside
their absorbed rules.
Appends canonical .btn variants at end of file (last cascade):
.btn-primary + .btn-primary-v2 + .modal-btn.primary (alias group)
.btn-secondary + .btn-secondary-v2 + .modal-btn:not(.primary):not(.danger)
.btn-ghost + .btn-ghost-v2
.btn-danger + .modal-btn.danger
.btn-lg
.btn:disabled + .btn:focus-visible (focus ring via --focus-ring)
Existing absorbed .btn, .btn-primary, .btn-secondary, .btn-sm rules
remain — the canonical block adds the missing variants + selector-list
aliases so .modal-btn and v2 markup keep rendering until migration
tasks swap them out.
Contract test: .btn-danger now declared (one less missing primitive).
Browser smoke: /admin/tokens hero + filter pills + empty state render
correctly with the absorbed style.css rules now backed by real tokens.
* feat(css): form-control primitives — .search-input + .filter-bar + .filter-pill + .form-input
Canonical filter bar shape: 36px-height inputs (matches button height
for vertical rhythm), 28px pills with .is-active state, consistent
focus ring via --focus-ring token.
Selector-list aliases for legacy per-page classes:
- .users-search / .marketplaces-search / .kb-search → .search-input
- .filters-card → .filter-bar
- .pill[aria-pressed="true"] also matches the .filter-pill active state
.form-input added as a sibling of .search-input for forms — same
baseline height + radius + focus treatment, with textarea.form-input
auto-sizing to min 96px and using the mono font (matches CSV/SQL
pasted-snippet patterns on /admin/agent-prompt + /admin/workspace-prompt).
Contract test: .search-input + .filter-bar + .filter-pill now declared.
* feat(css): .page-header primitive + variants + .tab-strip
Canonical page-header pattern with title (22px) + optional subtitle +
optional eyebrow + right-aligned actions slot. Two modifiers:
- .page-header--hero: gradient background (primary→primary-dark),
28px white title, semi-transparent subtitle/eyebrow. For
/marketplace, /store, /profile-style pages that already use this
layout via per-page inline <style>. Migration tasks delete the
duplicated rules.
- .page-header--compact: 18px title for dense admin index pages.
.tab-strip + .tab-strip__item — the secondary tab row pattern used by
/marketplace?tab=flea and similar. .is-active / [aria-selected=true]
both flip the active treatment (primary color + bottom border).
Contract test: .page-header / __title / __subtitle / __actions all
now declared (4 fewer missing primitives).
* feat(css+js): .data-table + .empty-state + .toast + .stat-card primitives
Last primitive batch. All 8 canonical-primitives invariants in
test_design_system_contract.py now green; only the template-migration
test fails (expected — Tasks 8–15).
.data-table (+ --compact modifier): selector-list aliases for legacy
per-page table classes (.users-table, .gp-table, .marketplaces-table,
.audit-table) so existing markup keeps rendering until migration.
Compact modifier shrinks padding + font for dense lists (audit log).
.empty-state with __icon / __title / __description / __actions —
replaces the ad-hoc 'no results' rendering scattered across pages
(corporate_memory, admin_users, admin_marketplaces, etc.).
.toast / .toast-container — paired with window.appToast({kind, msg,
timeout}) appended to app.js. Bottom-right stacked, click-to-dismiss,
auto-dismiss after 4s by default. Kind 'success' / 'warning' / 'error'
/ 'info' shows a 3px colored left border.
.stat-card (+ --accent variant) + .stat-row grid — for the dashboard
metric tile row.
* style(templates): migrate 8 templates off deprecated class names
Mechanical class-attribute rewrite via tokenizer (preserves Jinja
conditionals + multi-line attrs):
modal-btn primary -> btn btn-primary
modal-btn danger -> btn btn-danger
modal-btn -> btn btn-secondary
users-table -> data-table
gp-table -> data-table
marketplaces-table -> data-table
audit-table -> data-table
users-search -> search-input
marketplaces-search -> search-input
8 templates touched: admin_groups, admin_marketplaces, admin_tokens,
admin_users, admin_welcome, admin_workspace_prompt, my_tokens,
corporate_memory_admin. 43 lines updated total.
Inline <style> blocks in these templates still define rules for the
old class names — those rules no longer match anything and become
dead code, removed in Task 16's alias cleanup along with the
selector-list aliases in style-custom.css.
Contract test (tests/test_design_system_contract.py) now fully green:
9/9 invariants enforced from this commit onward.
* feat(css): extend .data-table selector list to 13 more bespoke -table classes
Visual unification of remaining tables across the codebase without
per-template edits. The .data-table baseline rules (uppercase header
tracking, 12px padding, hover state, border-radius) now apply to:
.ad-table / .ea-table / .md-table / .members-table /
.obs-table / .overview-stats-table / .registry-table /
.sample-table / .sched-table / .sess-table / .sub-table /
.subs-table / .ud-table
These class names live in 12 templates (activity_center, admin_access,
admin_group_detail, admin_scheduler_runs, admin_sessions,
admin_store_submissions, admin_tables, admin_usage, admin_user_detail,
catalog, me_debug, profile_sessions) that have their own per-page
<style> blocks. Per-page rules with higher specificity still win for
their custom needs (column widths, etc.) — this commit only sets a
shared baseline so every table renders with the same chrome.
Contract test stays green: 9/9 invariants enforced.
* style(css): remove now-unused legacy class aliases
Phase A renamed 8 templates off these names; no markup references
them any more, so the selector-list memberships are dead weight.
Removed from style-custom.css:
.btn-primary-v2 / .btn-secondary-v2 / .btn-ghost-v2
.modal-btn / .modal-btn.primary / .modal-btn.danger /
.modal-btn:not(.primary):not(.danger)
.users-search / .marketplaces-search / .kb-search
.users-table / .gp-table / .marketplaces-table / .audit-table
.filters-card
37 lines smaller. Contract test catches any reintroduction.
KEPT aliases (still in untouched template markup):
- .pill (marketplace_plugin_detail.html, marketplace.html — these
pages weren't part of Phase A's deprecated-class sweep; their
own .pill CSS rules still apply)
- All .data-table family extensions (.ad-table, .ea-table, .md-table,
.members-table, .obs-table, .overview-stats-table, .registry-table,
.sample-table, .sched-table, .sess-table, .sub-table, .subs-table,
.ud-table) — these still render data tables in 12 templates;
selector-list aliasing keeps them visually unified with .data-table
baseline.
- Legacy token aliases (--bg / --text / --text-light / --secondary /
--card-bg / --radius) — still resolve absorbed style.css rules.
Templates' inline <style> blocks still contain dead rules for the
renamed classes (.users-search, .modal-btn, etc.); harmless but
bloat. Optional follow-up: a separate sweep can drop those.
* docs(changelog): design-system unification under [Unreleased]
* feat(css): unify page-shell width — .container baseline 1280px + modifiers
Inventory found 30+ unique max-width values across templates (280px
login → 1600px admin/tables). The legacy .container default was 800px,
which made every admin page set its own wider inline override —
30+ ad-hoc widths drifted as a result.
Canonical: .container max-width = var(--width-app) (1280px). Pages
that need a different shape opt in via modifiers:
.container--narrow → var(--width-narrow) (800px) — long-form text,
setup wizards
.container--wide → var(--width-wide) (1400px) — admin lists,
marketplace grids
.container--full → max-width: none — hero / landing
Pages that already set a NARROWER inline max-width (setup, login flows
inside .login-card, etc.) still render at their narrower size — the
inline override beats the new canonical 1280px. The visible change
hits the ~20 admin pages currently rendering at 800px via the legacy
default, which jump to 1280px and pick up consistent breathing room.
Spacing also normalized: padding 24px 20px → var(--space-6) var(--space-5).
* fix(home+catalog): gut dashboard sections + remove confusing toggle + fix table count
Dashboard /home cleanup:
- Remove 'Your Data' card — Data Packages is already a top-nav entry,
so duplicating data sources on the landing page just adds noise.
- Remove 'Account' card — group memberships + scripts + last sync
belong on /profile, not on the welcome screen.
- Remove entire right-column (Corporate Memory + Activity Center
widgets) — both surfaces have dedicated admin pages reachable from
the Admin dropdown.
- Keep stats row (Tables/Columns/Rows/Data Size/Unstructured),
env-setup-CTA, and Notifications card.
/catalog cleanup:
- Strip the 'Always included' badge + the locked toggle-switch from
Core Business Data and Business Metrics cards. The toggle was
always 'checked disabled' — it visually looked like a switch but
could not be toggled, which was confusing. The 'Always included'
copy itself was redundant once the toggle was gone. Agnes Internal
already rendered without these, so the three cards are now visually
consistent.
Catalog data_stats fix:
- 'total_tables' was len(sync_state) — counted only tables that had
ever synced, so a 30-row table_registry with 0 ever synced rendered
as '0 tables'. Switched to len(tables) — the registered
business-data table list — so the count reflects what's actually
available, not what's been touched.
* fix(home): real stat numbers + drop unstructured tile + cleanup dead CSS
Dashboard stats were hardcoded zeros (columns: 0, size_display:
'0 MB', unstructured_display: '0 MB') and the table counter pulled
from sync_state (synced) instead of table_registry (registered).
On a fresh deployment with 30 registered tables and 0 ever synced,
the page rendered '0 / 0 / 0 / 0 MB / 0 MB' — useless.
Now:
- Tables: COUNT(*) FROM table_registry WHERE source_type != 'internal'.
Matches the /catalog Core Business Data counter.
- Columns: SUM(sync_state.columns). Zero only when nothing's synced yet.
- Rows: unchanged (SUM(sync_state.rows), already correct).
- Data Size: SUM(sync_state.file_size_bytes), human-formatted via
inline _fmt_bytes helper (KB/MB/GB).
- Unstructured: tile dropped — was always '0 MB' and had no source.
- last_updated: now derived from sync_state max(last_sync), wasn't set
before so the 'Synced …' tag never rendered.
Dashboard.html cleanup: ~725 lines of orphan inline <style> removed —
.section-title, .data-source*, .toggle-switch*, .catalog-cta*,
.memory-card / .memory-stat / .memory-description / .memory-footer
/ .btn-memory, .activity-card / .activity-stat / .activity-text
/ .btn-activity, .account-grid / .account-row / .account-scripts
/ .badge-role / .badge-group / .cron-line, .badge-included /
.badge-beta / .badge-demo. All matched markup deleted in the
previous commit; the CSS was dead code until now.
* ui(catalog): rename page heading 'Data Catalog' → 'Data Packages'
The top-nav entry says 'Data Packages' but the page itself said
'Data Catalog' — confusing two-name product. Aligns the heading and
<title> with the nav label. Subtitle trimmed too: 'manage your
subscriptions' was a vestige of the toggle UI that just got removed,
replaced with a one-liner describing what the page is for.
Two other 'Data Catalog' strings stay: they live inside the table-
profiler overlay JS and refer to an EXTERNAL catalog system (e.g.
OpenMetadata / Atlan) that an operator may link to per table — that
is a generic term for any external data-catalog product, not our
page name.
* fix(nav): dropdown clicks always work + mutual-exclusion close
Two bugs in the wireDropdown helper:
1. Clicking trigger B while trigger A's menu was open left both open.
e.stopPropagation() in trigger.click prevented the document-click
handler from firing, so trigger A's open menu had no way to learn
that something else was clicked. Net effect: state diverged across
the two dropdowns the more you clicked.
2. The target-vs-trigger equality check (e.target !== trigger) was
strict. Clicking the chevron <svg> inside the button reports the
svg or its <path> child as e.target — not the button — so removing
stopPropagation alone would trip the close branch in the same
click that just opened the panel.
Fix both at once: drop e.stopPropagation() AND switch the doc-handler
guard to trigger.contains(e.target). Now any click outside both the
trigger subtree and the panel subtree closes; any click on another
trigger closes via the OTHER dropdown's doc handler; clicks inside
the trigger (button OR svg child) are fully ignored by the doc
handler and only the trigger's own toggle handler fires.
* feat(ui): canonical blue-gradient hero on every admin page
The UI had a per-page hero pattern on ~10 onboarding/marketing pages
(admin_tokens / profile / install / setup_advanced / marketplace /
my_tokens / store_upload / home_*), each with its own ad-hoc CSS
(.tokens-hero, .profile-hero, .install-hero, .upload-hero, …). The
admin section's index + detail pages had plain H1/H2 with their own
.users-title / .gp-title / .obs-title / .cfg-title / … inline styling.
Net effect: half the app felt like a product, half felt like a
spreadsheet.
Now:
- .page-header--hero CSS upgraded to match the look analysts already
liked from admin_tokens: 28px/32px/24px padding, 14px radius, soft
primary-tinted box-shadow (0 4px 16px rgba(0,115,209,0.2)), 28px
semibold title, optional uppercase eyebrow + 13.5px subtitle.
Narrow-viewport breakpoint included.
- New _page_hero.html partial wraps the boilerplate. Usage:
{% set page_hero_eyebrow = "Users & Access" %}
{% set page_hero_title = "Users" %}
{% set page_hero_subtitle = "…" %}
{% include "_page_hero.html" %}
- 15 admin templates migrated to it: admin_users / admin_groups /
admin_marketplaces / admin_access / admin_sessions /
admin_session_detail / admin_store_submissions /
admin_scheduler_runs / admin_usage / admin_user_detail /
admin_welcome / admin_workspace_prompt / admin_server_config /
activity_center / admin/news_editor. Each gets a grouped eyebrow
(Users & Access / Data / Agent Experience / Activity Center /
Server) matching the Admin dropdown sections so the page identity
is unambiguous at a glance.
Legacy *-title H2/H1 + adjacent subtitle paragraphs deleted; their
per-page CSS rules are dead now (harmless, retire in a follow-up
sweep alongside other inline-style cleanup the reviewers flagged).
admin_tables.html intentionally NOT migrated — it's a standalone
HTML page that doesn't extend base.html; a separate refactor.
Test: test_admin_users_page_renders_for_admin assertion updated
from .users-title to .page-header__title + .page-header--hero (the
canonical pair). All other web/template tests stay green.
* refactor(ui): dedup _humanbytes, drop 267 lines of dead inline CSS
(1) _humanbytes consolidation:
- Add TB branch + optional precision param (default 2 preserves existing
Store detail callers; dashboard uses precision=1 for headline tiles).
- Delete inline _fmt_bytes from dashboard handler — was a copy of
_humanbytes with different rounding. One canonical helper now.
(2) Dead inline-CSS sweep across 17 migrated templates:
- Conservative regex: a CSS rule is deleted only when its primary class
matches one of the known-dead names AND that name is NOT referenced
from any class= attribute in the same file's markup.
- Per-file 'in-use' guard saved several false positives that the deny
list would have nuked (e.g. .users-toolbar, .gp-search, .obs-subtitle,
.marketplaces-toolbar are still in use; only .users-table, .users-search,
.users-title, .modal-btn, etc. that have NO markup left went away).
- Removed: -267 lines across admin_users (-42), admin_marketplaces (-45),
admin_groups (-31), my_tokens (-38), admin_tokens (-29), admin_access
(-9), admin_user_detail (-6), admin_welcome (-8), admin_workspace_prompt
(-8), admin_server_config (-2), admin_sessions (-1), admin_session_detail
(-1), admin_usage (-1), admin_store_submissions (-3), admin_scheduler_runs
(-3), activity_center (-4), corporate_memory_admin (-36).
Contract test stays green (9/9); all web/template/render/user_management
tests pass.
* feat(ui): canonical hero on /catalog (Data Packages)
Same .page-header--hero treatment as the admin pages — Data eyebrow,
Data Packages title, Browse-the-data-sources subtitle. Removes the
ad-hoc .page-title block (h1 / p / wrapper-div) and its CSS rules
(now dead, 3 rule blocks deleted).
* fix(nav): load app.js from _app_header.html — works on standalone pages
The previous nav-fix commit moved the inline dropdown script from
_app_header.html into app/web/static/app.js + added <script src=…>
to base.html. That broke EVERY page that includes _app_header.html
WITHOUT extending base.html (catalog, corporate_memory*,
admin_tables, install). They got the nav markup but no JS → both
Admin and AD dropdowns dead on those pages.
Fix: emit the <script src=app.js defer> directly inside the
_app_header.html partial. Any page that includes the header now
gets the script automatically — base.html-extenders AND standalone
HTML pages alike. base.html's duplicate <script> line removed.
Also fixes the wide-hero on /catalog: .page-header--hero now sets
its own max-width: var(--width-app) (1280px) so standalone pages
without a .container parent don't render the gradient edge-to-edge.
catalog's .source-cards bumped from 900px → 1280px to match the
hero, otherwise the page reads two-tier (wide blue band, narrow
content) which the user flagged.
Verified locally via agent-browser: Admin + AD dropdowns now click
through on /catalog, /admin/tables, /corporate-memory.
* docs(plan): standalone pages → base.html framework migration plan
Plan + Plan-agent review (8 must-fix items applied) for converting
the 5 templates that ship their own <html><head><body> scaffold
(catalog, install, corporate_memory, corporate_memory_admin,
admin_tables) to extend base.html. Root cause of yesterday's
'dropdown dead on /catalog' regression: shared infrastructure in
base.html doesn't propagate to standalones.
* feat(base): body_attrs block + migrate install.html to extend base
base.html: new {% block body_attrs %}{% endblock %} slot so pages
that need <body> attributes (admin_tables has data-source-type)
can carry them through extends.
install.html: convert from standalone <html><head><body> scaffold
to {% extends "base.html" %} with title / body_attrs / head_extra
/ layout / scripts blocks. Drops:
- <!DOCTYPE>, <html>, </html>, <head>, </head>
- <meta charset>, <meta viewport>
- Duplicate <link rel="stylesheet" href="...style-custom.css">
(base.html already provides one)
- <body> opening + closing tags
- Leading _app_header.html include + _version_badge.html include
(base.html handles both)
Preserves per-page CSS (in head_extra), per-page JS (in scripts),
the Inter font preconnect (kept inline; not hoisted to base in
this PR — separate decision).
Pilots the migration recipe before the 4 larger pages.
* refactor(memory): extend base.html
Same recipe as install.html. corporate_memory.html now inherits
<html>/<head>/<body> + nav + app.js script tag from base.html.
Page-specific CSS and JS preserved in head_extra + scripts blocks.
* refactor(memory-admin): extend base.html
Same recipe as install/corporate_memory. Curation page now in the
shared rendering pipeline.
* refactor(catalog): extend base.html
catalog.html had the most complexity: 7 head-level assets (chart.js,
Prism, prism-sql, metric_modal.css link + 2 preconnects + Inter
stylesheet), 5 body-level <script> blocks including a <script type=
"module"> for the metric modal, 2 duplicate style-custom.css links
in <head>. The migration script preserved all of them — head-level
externals hoisted to {% block head_extra %} in source order, body
scripts relocated to {% block scripts %} in source order (so chart.js
loads before the IIFE that builds Chart instances), duplicate
style-custom.css links dropped (base.html provides one).
* refactor(admin-tables): extend base.html + carry data-source-type
The biggest of the 5 standalones at 3563 lines. <body data-source-
type="{{ data_source_type }}"> attribute carried through via the
new {% block body_attrs %} slot (admin_tables JS reads
document.body.dataset.sourceType to switch between keboola and
bigquery rendering paths).
* release: 0.54.10 — UI design system unification + homepage status frame + initial workspace override + store guardrails
Co-Authored-By: zdenek.srotyr <zdenek.srotyr@keboola.com>
* refactor(web): migrate remaining templates to canonical design primitives
- admin_group_detail: .data-table, .btn family, appToast(), remove duplicate table/button/toast CSS
- admin_store_submission_detail: .data-table, .btn family, appToast(), remove duplicate btn/toast CSS
- profile_sessions: .data-table, _page_hero.html, remove duplicate table/title CSS
- me_debug: .data-table, .btn family, remove duplicate table/button CSS
- marketplace: .btn-primary/.btn-secondary, remove duplicate button CSS
- store_edit: remove duplicate .btn-primary/.btn-link CSS, canonical button classes
- store_upload: remove duplicate .btn-primary/.btn-secondary/.btn-link CSS
Co-Authored-By: zdenek.srotyr <zdenek.srotyr@keboola.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(me/stats): per-analyst Stats dashboard with 4 tabs
New /me/stats page shows the calling user's own analytics across
four tabs, lazy-loaded per activation:
- **Sessions** — paginated usage_session_summary join with a
filesystem scan of un-processed JSONL (mirrors admin
list_user_sessions shape). v44 token columns aggregated per row.
- **Tokens** — daily series (default 30 days), by-model breakdown
(lifetime), top-10 biggest sessions, lifetime totals. Single
CTE per sub-query against per-user partition (idx_usage_session_user).
- **Data access** — audit_log rows where action LIKE 'query.%' for
the caller. Covers query.local / query.hybrid / query.remote /
query.internal. Cursor-paginated on (timestamp, id).
- **Sync activity** — audit_log rows where action is sync.* or
manifest.* for the caller, plus users.last_pull_at for the
header. Per-pull history now persists thanks to the new
manifest.fetch audit row.
Backend: app/api/me_stats.py — single APIRouter at /api/me/stats/*,
four GET endpoints, all gated by get_current_user (server-side
caller scope; the page route itself only renders the shell).
Frontend: app/web/templates/me_stats.html — tab bar + 4 panels,
plain JS lazy-loads each panel's endpoint on first activation,
caches per-tab so switching back doesn't refetch. Small SVG bar
chart on Tokens tab (no external charting dep). 'Stats' link
added to _app_header.html primary nav between 'Data Packages'
and the Admin dropdown.
Side change in app/api/sync.py: /api/sync/manifest now emits a
manifest.fetch audit_log row alongside the existing
users.last_pull_at bump. The column UPDATE only retains the
most recent timestamp; per-pull history needs an audit row.
client_kind='api' for the manifest endpoint (vs. 'web' which
the audit-read deduper uses for AC reads), so the Sync tab can
distinguish CLI pulls from browser-driven manifest peeks.
7 new tests in tests/test_me_stats.py:
- sessions endpoint caller-scope isolation (user A doesn't see B)
- sessions pagination
- tokens empty-user zero shape
- tokens aggregation across daily window + by_model + top + totals
- queries endpoint filters to action LIKE 'query.%' + caller scope
- sync endpoint surfaces both manifest.fetch and sync.trigger
- manifest endpoint writes the manifest.fetch audit row
* ui(me/stats): widen page to 1400px via main.main escape
Default base.html .container wraps content at max-width 800px. Stats
tables (by-model + top-sessions: 6 columns each) felt cramped at that
width — same constraint dashboard.html escapes via the {% block layout %}
override pattern. Mirror that here: render <main class="main"> and
bump .stats-page max-width to 1400px so the 6-column tables breathe
without going edge-to-edge on wide monitors.
* ui(me/stats): narrow from 1400px to 1280px to match /home
/home isn't actually .container's default 800px — style-custom.css
has a body:has(.home-mock) .container { max-width: 1280px } override
that widens it. 1280px is the shared 'wide content' width across the
codebase (top-nav header + /home + dashboard all use it).
Bumping me_stats from 1400px to 1280px so the Stats page reads as
'same chrome' instead of distinctly wider than its sibling pages.
* feat(home): status frame on /home — last sync, sessions, prompts, tokens, projects
Adds the homepage status frame: a 5-card row above the install-hero /
offboard-strip on /home showing the calling user's Last sync (their
last `agnes pull`), Sessions, Prompts, Tokens used, and Projects worked
on, with a 24h/7d pill toggle.
Backed by `GET /api/me/home-stats?window=` (one DuckDB CTE joining
`users` + `usage_session_summary` + `usage_events`) and SSR'd from the
same `compute_home_stats` helper on initial paint so there's no
spinner. The window toggle is the only JS-driven path.
Side surfaces:
- `GET /api/sync/manifest` now stamps `users.last_pull_at` so
`agnes pull` (and the Claude Code SessionStart hook that wraps it)
imprints the analyst's last sync time for the new card.
- `usage_session_summary` gains four BIGINT token counters
(input_tokens, output_tokens, cache_read_tokens, cache_creation_tokens)
summed from JSONL `message.usage.*` per assistant turn.
- `USAGE_PROCESSOR_VERSION` bumps 1 → 2 so the session-pipeline
reprocess loop invalidates stale summaries and backfills tokens
on the next tick.
Schema migration v43 → v44 is idempotent ALTERs (last_pull_at +
4 token columns) — fresh installs receive them from `_SYSTEM_SCHEMA`,
upgrade path runs `_v43_to_v44`. Defaults (NULL / 0) backfill
existing rows cleanly.
9 new tests in tests/test_home_stats.py cover the migration,
endpoint shapes (24h/7d/unknown/empty/missing-user), and the
manifest-side last_pull_at bump.
* docs(CHANGELOG): homepage status frame entries under [Unreleased]
The post-rebase release-cut now belongs to whichever PR lands next
after main rolled to 0.54.9. This PR logs its bullets under
[Unreleased] (Added: homepage status frame, per-user pull tracking,
token counters; Changed: schema v43 → v44 migration) so they ride
out with the next release-cut.
* fix(tests): bump test_schema_v42_migration asserts to v44
CI failed because tests/test_schema_v42_migration.py hardcoded
`assert SCHEMA_VERSION == 43` and `assert v == 43` after init.
v44 (homepage stats frame backing columns) was introduced in the
preceding feat commit; this aligns the existing v42-era migration
tests with the new schema version.
* feat(home): gate status frame on operator flag + user.onboarded
Two gates on the homepage status frame:
1. **Operator master switch** — `get_home_status_frame_visibility()` in
app/instance_config.py mirrors the existing `get_home_automode_visibility()`
shape: env var `AGNES_HOME_SHOW_STATUS_FRAME` > yaml
`instance.home.show_status_frame` > default `True`. Cautious-rollout
instances can disable the frame without forking; the yaml example
documents both knobs.
2. **Onboarded gate** — the template only renders the frame when the
caller's `users.onboarded` is true. First-day users see a clean
install-hero before all-zero stat cards; the frame appears
automatically on the next render after `agnes init` POSTs
`/api/me/onboarded`.
Router skips the `compute_home_stats` DB read entirely when either
gate is closed; `home_stats` arrives at the template as None in that
branch and the `{% if %}` shortcuts the include.
Why both gates: PostHog feature flags evaluated and rejected — this
codebase uses PostHog for analytics capture only, not feature gating;
adding a per-user feature_enabled() call on the /home critical path
would couple the homepage render to a remote eval and still require
an admin master switch. The onboarded gate is a UX coherence rule
layered on top of the operator switch, not an A/B test signal.
3 new tests in test_home_stats.py cover the env-var resolution
(falsey values + default-true). The yaml example gets a `home:`
block documenting both `show_automode` (pre-existing flag, was
undocumented in the example) and `show_status_frame`.
* perf(marketplace): browser-cache cover photos + restore Curated tab filter spacing
Cover photos on /marketplace grid now serve with `Cache-Control: public,
max-age=2592000, immutable` plus URL fingerprinting (`?v=<commit-sha8>`
for curated, `?v=<version_no>` for flea) so browser refresh stops hitting
the server entirely for unchanged assets. Per-plugin RBAC dropped from
the three image endpoints (curated_asset, curated_mirrored, get_entity_photo)
in favor of login-only auth — eliminates _system_db_lock contention on
parallel image requests. Per-request magic-bytes revalidation also dropped
from curated_asset (it was re-reading the file just to discard the bytes,
then FileResponse read it again).
Spacing bug: sort-dropdown commit (6be1cee) wrapped .mp-filter-row in a
new flex container with inline margin-bottom:4px, masking the original
12px CSS rule. Curated tab (where .mp-type-row is hidden) ended up with
4px between filters and the card grid. Wrapper margin restored to 12px.
See CHANGELOG entry under [Unreleased] — the RBAC relaxation is called
out under ### Security with explicit threat-model rationale for AI/human
reviewers.
* test(marketplace): update renamed-html-as-png test for dropped magic-bytes check
Magic-bytes body validation was dropped from `curated_asset` in the previous
commit — the request path now relies on extension allowlist + pinned
Content-Type + nosniff + strict CSP to neuter mismatched payloads at the
browser layer. Update the test to assert the new defense-in-depth posture
(200 served, but Content-Type=image/png + nosniff + CSP=default-src 'none')
rather than the gone 415.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
Three tweaks to the post-PR-#291 Getting Started card:
1. Chronologically first. Moved from below the install-hero (where
it sat as a static white card) to ABOVE it, inside the same
`{% if not onboarded %}` guard. The blue hero is now the actual
install flow that the card points at, not a peer that competes
for attention.
2. Collapsed by default. Switched from <section> to <details> with
no `open` attribute, so the page lands with just a quiet pill
(`Getting Started — Two quick next steps — click to expand ›`).
Expand to reveal the two rows. Chevron rotates 90deg when open
via the `[open]` selector. Per-device dismiss X stays — generic
`.home-card-close[data-dismiss-key]` handler now uses
`closest('section, details')` so it works on both container types.
3. First row → #install-hero in-page anchor. Was `/setup` (which
would round-trip to the same hero via a redirect through /setup).
Anchored directly to the blue hero on the same page; copy reads
"One-time install — walkthrough in the section below" so the
user knows it's a scroll-to, not a navigation. Install-hero <div>
gained `id="install-hero"`. `.install-hero { scroll-margin-top:
88px }` keeps the hero's eyebrow clear of the 72px sticky header
on the jump.
Second row link to /setup-advanced and the dismiss key unchanged.
GS disappears alongside the install-hero when the user is onboarded,
so the in-page anchor never dangles. Tests updated to assert the new
markup + onboarded-state hiding.
Addresses post-merge review findings on #290:
- Admin Rescan is the only post-v30 producer of status='blocked_inline'.
Re-add it to admin queue 'Needs review' filter chip and to
TERMINAL_BLOCKED_STATUSES in the bundle-purge job so rescan-produced
rows surface in the default operator view and bundles get TTL-swept
instead of lingering indefinitely.
- Update three doc-drift sites still referring to the pre-#290 spam
counter scope (counted blocked_inline). The counter now narrows to
blocked_llm + review_error; fix the comment in app/api/store.py,
the docstring in get_guardrails_blocked_quota_per_day(), and the
operator-facing hint rendered on /admin/server-config.
- Add positive test for _reject_inline_or_continue validation branch
(code='validation_failed', checks payload shape, no-DB-write
contract). Locks the frontend wizard's detail.checks contract.
- Tighten test_quota_disabled_with_zero — assert (200, 201) explicitly
instead of !=429 so a 500 regression no longer passes.
- _reject_inline_or_continue takes plugin_dir and lazy-computes
bundle_meta only on the security branch. Validation rejects no
longer pay for a SHA256 walk on the bundle.
- Surface store.upload.security_blocked audit-log write failures via
logger.exception instead of swallowing — that audit row is the only
forensic trace by design.
* feat(initial-workspace): per-instance agnes init override
Adds Initial Workspace Template — an admin-configurable per-instance
override for the agnes init analyst workspace. When configured, agnes
init downloads a server-rendered zip from a Git repo the admin registered
and extracts it into the analyst's workspace, fully bypassing Agnes-default
CLAUDE.md / settings.json / hooks / slash commands / AGNES_WORKSPACE.md.
Repo layout convention: only the contents of a top-level `workspace/`
subdirectory ship to analysts; admin docs (README, CI configs) at the
repo root stay in the repo and never reach an analyst. Sync rejects
repos without `workspace/` at root.
Server side:
- src/initial_workspace.py — clone (or fetch+reset), validate, build zip
with strict path checks and reserved-path rejection
(workspace/.claude/init-complete reserved by Agnes)
- app/api/initial_workspace.py — admin CRUD + sync endpoint + analyst-
facing status/zip/applied endpoints; config persists to instance.yaml
overlay, PAT to .env_overlay
- app/secrets.py — refactor: persist_overlay_token shared helper with
threading.Lock for .env_overlay writes (closes pre-existing race
between concurrent marketplaces saves)
- app/web/templates/admin_server_config.html — new "Initial Workspace
Template" section + modal + Sync/Edit/Delete/Download buttons (matches
existing cfg-section visual language)
CLI side:
- cli/lib/override.py — single source of truth for is_override_workspace
sentinel detection
- cli/lib/initial_workspace.py — probe status, safe zip extraction with
../absolute/symlink rejection, typed-YES force confirmation
- cli/commands/init.py — override branch (skips Agnes-default workspace
writes); extended sentinel with override:true, template_source,
template_sha so future agnes self-upgrade does not auto-refresh hooks
- cli/lib/hooks.py + cli/lib/commands.py — short-circuit on override
workspaces (install_claude_hooks, install_claude_commands,
maybe_refresh_claude_hooks)
Audit-event strategy: server writes initial_workspace.fetch_started
inside GET /api/initial-workspace.zip (cannot be spoofed by PAT-holder);
CLI POST /applied writes initial_workspace.applied as best-effort
confirmation. Admin mutations log via the existing _audit pattern.
Tests: 27 server (clone/validate/zip + workspace-subdir convention +
concurrent persist_overlay_token + endpoint shapes + audit rows) + 29
CLI (override sentinel parse + probe fall-through + safe extraction +
YES strictness + hook guards + e2e mocked init).
Risk acceptance — documented in docs/initial-workspace-override.md +
CHANGELOG Internal section so AI reviewers understand the deviations
from defaults are intentional:
- maybe_refresh_claude_hooks deliberately no-ops on override workspaces
- --force on override does NOT back up CLAUDE.md (admin's repo is the
source of truth)
- .claude/CLAUDE.local.md IS overwritten by override extraction when
admin's repo ships one
* test+vendor-agnostic: drop Groupon tokens from #292 fixtures + extend admin-gate coverage
Two fixes from the takeover review on #292:
1. **Vendor-agnostic OSS rule**: Replace `Groupon` / `groupon/template`
tokens in test fixtures with `Acme` / `acme/template` (8 sites in
test_cli_init_override.py + 1 in test_initial_workspace_api.py).
Per CLAUDE.md "Vendor-agnostic OSS — no customer-specific content"
rule: customer-specific tokens don't belong in shipped artifacts,
even in test fixtures. The pre-existing FoundryAI mentions in
test_instance_config.py + test_setup_instructions.py are out of
scope for this PR (didn't introduce them).
2. **Admin-gate coverage gap**: `test_admin_endpoints_require_admin`
only covered GET /api/admin/initial-workspace + POST .../sync. The
register-write (POST .../initial-workspace) and delete (DELETE
.../initial-workspace) endpoints used the same `Depends(require_admin)`
wiring but had no regression test. Loop now covers all 4 verbs so
a future refactor that drops the dependency from one endpoint
fails here instead of silently exposing the write/delete paths to
any analyst with a PAT.
* release: 0.54.9 — Initial Workspace Template (per-instance agnes init override)
Last commit on the PR per CLAUDE.md hard rule. Patch bump (0.54.8 →
0.54.9) for Mina's Initial Workspace Template feature.
No DB migration (config lives in instance.yaml overlay). No
mandatory operator action — empty default keeps OSS-default
agnes init behavior. Operators wanting full template control link a
Git repo on /admin/server-config → "Initial Workspace Template".
See docs/initial-workspace-override.md for the full
responsibility-transfer contract.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* feat(store): hard-reject inline guardrail failures, trace security only
Inline failures (manifest + content validation, static-security
deny-list hits) now hard-reject upstream of any DB write or bundle
persistence. The v30 contract that landed every inline failure as a
hidden+blocked_inline entity + admin-rescannable bundle is replaced
with two response shapes:
- 422 code=validation_failed — manifest/content issues. Banner-only,
no submission row, no audit_log entry. Submitter fixes and retries.
- 422 code=security_blocked — static_scan finding. Banner-only on
the wire, plus one audit_log row (store.upload.security_blocked)
carrying findings + sha256 + size for admin forensics.
Quarantine + admin rescan/override apply only to the async LLM path
(blocked_llm / review_error) — the cases that genuinely benefit from
admin judgment.
Spam-quota counter narrows to blocked_llm + review_error. Admin queue
filter chip drops blocked_inline. Bundle TTL purge stops sweeping
blocked_inline. Legacy blocked_inline rows from instances that ran
the v30 contract remain reachable via the "All" tab.
New _reject_inline_or_continue helper in app/api/store.py centralises
the two-tier rejection across create_entity, update_entity, and
restore_version. Frontend templates render the new payloads as inline
banners (no redirect on failure) and keep submission_blocked as a
one-release back-compat branch.
Tests: new _seed_quarantined_entity helper replaces the older
_make_eval_skill_zip-driven setup wherever a test needs a
hidden+blocked_llm entity. 199 store tests pass under -n auto.
* release: 0.54.8 — store inline hard-reject (BREAKING)
Last commit on the PR per CLAUDE.md hard rule. Patch bump (0.54.7 →
0.54.8) wrapping Vojta's hard-reject refactor.
**BREAKING for store-upload clients**: validation failures now return
422 with `code='validation_failed'` (no entity row, no submission row,
no audit_log entry) instead of the v30 `submission_blocked` 200
response that landed a hidden `blocked_inline` row. Frontend wizard +
edit + restore still understand the legacy code for one release as a
fallback for stale clients hitting an older deploy. Operators with
custom integrations against `POST /api/store/entities` should update
to handle the new `code='validation_failed'` / `code='security_blocked'`
422 responses.
No DB migration required (legacy `blocked_inline` rows from instances
that ran the v30 contract remain reachable via the admin queue's
"All" tab; bundle-purge job no longer covers them but they linger
harmlessly).
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* feat(home): Getting Started + Overview + Usage modes sections
Three new content cards rendered between the install-hero and the
existing connector tiles on /home. Order: Getting Started → Overview
→ Usage modes → connectors.
- Getting Started — dismissible card with two clickable rows linking
to /setup (install flow) and /setup-advanced (deeper reference).
Subsumes the legacy `.advanced-pointer` row that sat above the news
section. Per-device dismiss via a generic localStorage handler:
`.home-card-close[data-dismiss-key="..."]` inside a <section> wires
itself up at page load — drop in any future dismissible card without
per-card JS.
- Overview — operator-owned HTML body sourced from the new
`instance.overview` yaml field (env override
`AGNES_INSTANCE_OVERVIEW`). HTML in, HTML out via the same `| safe`
filter as news_intro. Empty default hides the section entirely,
keeping the OSS vendor-neutral; operators paste their product
framing / privacy posture into instance.yaml. New helper
`get_instance_overview()` in app/instance_config.py mirrors
`get_instance_logo_svg()`.
- Usage modes — three OSS-shipped tiles (Terminal / VS Code / Claude
Desktop · claude.ai) explaining each surface and linking to the
matching /setup-advanced anchors. Closes the gap for users
wondering "where do I actually run this".
Supporting changes:
- setup_advanced.html gains a new `#claude-app` section between
#vscode and #workspace, anchored by the Usage modes Claude Desktop
tile. Covers the marketplace registration paths and when to prefer
the terminal. Added to the table of contents.
- Three new tests in test_web_home_page.py pin the Getting Started
card markup, the Overview-on-when-yaml-set path, and the
Overview-off-by-default path. All 13 tests in the file pass.
Operator follow-up (separate infra PR — NOT this PR): paste the
Foundry-specific Overview body into instance.yaml's
`instance.overview` field. OSS ships with an empty default.
* fix(home): Overview is operator-owned content — drop dismiss button
Earlier iteration added a close X to the Overview section to match
the Getting Started card's dismiss UX. Wrong call: Overview is
operator-authored reference content (privacy posture, telemetry
policy, project framing) and a per-device localStorage hide means
returning users who want to re-read the policy can't recover it
without clearing storage.
Reverts the close button + the data-dismiss-key on the Overview
section. Test inverted to assert the dismiss key is absent (defends
against a future drive-by adding it back). Getting Started still
dismisses — that's procedural getting-started content users
legitimately stop needing once they've finished setup. Overview is
always reachable; whole section is still opt-in at the operator
level via the empty-yaml default.
* fix(home): Terminal usage-mode tile is informational (no click-through)
The setup hero above /home's Usage modes already walks the user
through the Claude Code CLI install — the Terminal tile click-through
to /setup just round-trips back to content the user already scrolled
past. Switch Terminal to a non-anchor <div> and scope the hover
affordance to a.home-usage-item so VS Code + Claude Desktop tiles
keep their click-through (those legitimately deep-link into
/setup-advanced anchors).
* fix(home): point Usage modes guidance at ~/{workspace}/Projects/ subfolder
The bundled plugin scopes the session-analysis loop and the
central-catalog sync to ~/<workspace>/Projects/, not the workspace
root itself — that convention already appears in the install hero's
Step 4 manual-fallback note ('Don't create ~/<workspace>/Projects/
manually — the bundled plugin offers to set it up after install').
Usage modes' footer guidance now matches: 'create every project
under ~/<workspace>/Projects/'. Also calls out that the
session-analysis loop is scoped to that root so users understand
why working outside the workspace dir is invisible to the platform.
* feat(brand): inline operator SVG logo + drop header subtitle (release 0.54.6)
Three header tweaks, one PR:
1. _app_header.html drops the small uppercase subtitle line below the
brand. instance.subtitle still flows into the CLAUDE.md preamble +
init welcome template ("Operated by …"); only the web header chrome
loses it.
2. get_instance_logo_svg() in app/instance_config.py reads
instance.logo_svg (yaml) / AGNES_INSTANCE_LOGO_SVG (env). The
yaml field was already documented in instance.yaml.example and the
template already supported inline <svg> via {{ config.LOGO_SVG |
safe }}, but router.py:344 hard-coded LOGO_SVG = "" — the middle
wire was missing. Now operators can paste a lockup directly into
their instance.yaml under instance.logo_svg: | and have it render
in the header. Resolution mirrors get_instance_brand (env > yaml >
""). instance.name remains independent: drives browser <title>
tags + page h1s + CLAUDE.md heading; the SVG is the web-header
visual only.
3. .app-header-logo svg gains max-height: 40px; width: auto; so any
operator's lockup scales via its viewBox to fit the 72px header
without per-asset width/height edits. Pairs with #2 — without the
clamp, raw artwork (e.g. a 1600x430 lockup) overflows the chrome.
Release-cut included per the same-PR rule (Unreleased contained only
these bullets after rebase onto 0.54.5).
* revert: keep app-header-subtitle span — out of scope for this PR
Initial commit dropped the subtitle line on the assumption that
the user wanted both the secondary header line AND the future-SVG
brand cleaned up. The actual ask was narrower: drop the hostname
suffix that renders inside instance.name ("Foundry AI (hostname)"),
which is a startup.sh concern, not a template one. Restore the
subtitle span and the CHANGELOG bullet that announced its removal.
PR scope narrows to LOGO_SVG wiring + CSS clamp only.
* fix(header): hide subtitle span when instance.subtitle is empty
Pre-fix the template fell back to the literal string 'Data Analyst
Portal' when INSTANCE_SUBTITLE was unset, so operators who left the
field empty saw a stray hardcoded label below their brand. Switched
to a Jinja {% if %} guard around the whole <span class="app-header-
subtitle"> so an empty subtitle produces no element at all — clean
header chrome instead of placeholder leak.
* feat(home): hide install-hero once onboarded + X close button
- Wrap the entire install-hero in `{% if not onboarded %}` so once
`users.onboarded=true` (auto-flipped by `agnes init` POSTing
/api/me/onboarded, or by the new X / existing fallback button) the
blue hero disappears entirely. Pre-PR the onboarded branch reused
the same shell with a "Welcome back" header + "Steps 1–4 done" badge
+ minimize toggle, which visually outweighed the actual nav hub.
- Add a circular × close button (top-right of the hero, rendered only
when not-onboarded). Click → window.confirm() asking the user to
acknowledge onboarding → POST /api/me/onboarded → reload. The
confirm string intentionally avoids the literal phrase
"Mark me as offboarded" because cli/commands/onboarded.py::status
scans /home's rendered HTML for that exact marker as a fallback for
the api/me/profile check.
- Lift the offboard escape hatch out of the hero into a discrete
`.offboard-strip` rendered below, gated `{% if onboarded %}`. Lets
the analyst flip back to the install view after wiping their
workspace folder.
- Centralize the /api/me/onboarded POST into a `postOnboarded()` JS
helper reused by the hero X, the existing "Mark me as onboarded"
fallback button, and the new offboard button.
Tests updated to match the new behavior:
- `test_home_onboarded_user_sees_nav_hub` — asserts the hero is gone
and the offboard strip is the only setup-flow remnant.
- `test_minimize_toggle_no_longer_rendered` (renamed) — asserts the
minimize toggle is absent in both states (was previously rendered
inside the now-hidden onboarded branch of the hero).
- `test_home_no_auto_transition_after_post_until_reload` — checks
offboard-strip presence post-flip instead of the removed
"Welcome back" hero copy.
* fix(home): X-close button used invalid source enum, hit 422
The X button's data-target-source was 'self_acknowledged_x' to give
audit_log a separate marker for X-vs-button-driven flips. But
app/api/me.py:38's OnboardedRequest pins source to a Literal of
['agnes_init', 'self_acknowledged', 'self_unmark'] — pydantic
returned 422 on every X click.
Confusing side effect: both buttons share self-mark-status as the
status element, so the failed X click rendered 'Failed (422)' next
to the still-functional 'Mark me as onboarded' button. Looked like
the button itself broke.
Fix: drop the _x suffix. Both surfaces now POST source='self_acknowledged'.
Distinction in audit_log is not load-bearing — the source field
captures user intent ('I'm onboarded'), not the specific UI affordance.
* feat(store-guardrails): admin-configurable content thresholds
Adds the flea-market content guardrail floors to the /admin/server-config
editor so operators can tune the bar without code changes. Defaults are
unchanged (60 chars description, 25 chars command, 5 distinct words, 200
chars body) — patching guardrails.* in instance.yaml or via the admin UI
overrides any of them and the next inline check picks up the new value.
src/store_guardrails/content_check.py now resolves the four floors via
helper functions (_min_desc_chars / _min_command_desc_chars /
_min_distinct_words / _min_body_chars) that read app.instance_config at
call time. Module-level _DEFAULT_* constants remain as fallbacks if
the import fails (defensive — keeps the guardrail module loadable
without the app package on its path).
app/instance_config.py grows four matching getters returning the live
value with sane defaults + integer coercion.
app/api/admin.py registers 'guardrails' as an editable section + ships
nine known-fields entries (min_description_chars,
min_command_description_chars, min_distinct_words, min_body_chars,
enabled, review_model, blocked_quota_per_day, blocked_bundle_ttl_days,
stuck_review_grace_seconds) with operator-facing hint copy explaining
what each knob does.
app/web/templates/admin_server_config.html gets a SECTION_META entry
so the section renders as 'Flea-market guardrails' with a help string
instead of a bare section ID.
app/web/router.py threads the live thresholds into /store/new and
/store/examples via a small _guardrail_thresholds() helper so the
disclosure copy, char counter, and "Why these limits" table render
the configured value (not a hardcoded 60). End-to-end smoke verified:
PATCH guardrails.min_description_chars=90 → /store/new immediately
renders "90 characters" + JS DESC_MIN=90 on the next request, no
restart required (helpers read live config per call).
* chore(store-guardrails): address PR review safe-fix findings
Code-review safe_auto findings on PR #281 (review run
20260513-100126-64052520):
- CHANGELOG: add Unreleased entry covering the new
/admin/server-config Flea-market guardrails section, the four live
threshold getters, and the route-helper rendering knobs. Required by
the project's non-negotiable "Changelog discipline" rule.
- content_check.py: narrow `except Exception` to `except ImportError`
on the four `_min_*()` resolver helpers. Surface-level TypeError /
ValueError on a malformed YAML value belongs to the
instance_config getters' own try/except — the resolvers should only
defend against the in-tree import itself failing, not silently
swallow real bugs in the getters.
- store_upload.html: refresh the stale "30-char threshold" comment to
reflect the configurable floor (default 60), and add `|default(60)`
/ `|default(25)` / `|default(5)` filters to the disclosure-copy
bindings so the upload form matches store_examples.html's
belt-and-suspenders rendering if a future route ever renders the
template without populating the `guardrail` context.
- router.py: tighten `_guardrail_thresholds()` return annotation
from bare `dict` to `dict[str, int]`.
Residual work (left for separate change after operator direction):
- Add round-trip test (PATCH guardrails -> next inline check uses
new value) — primary testing gap.
- Decide policy on `min_*=0` (currently coerced to 1 via
`max(1, int(val))`) vs treating 0 as a disable sentinel like
neighbour getters (`blocked_quota_per_day`,
`blocked_bundle_ttl_days`).
- Add POST-time integer validation for `guardrails.*` so a typo'd
YAML value (bool / string / float) errors loudly instead of
silently falling back to the default.
* test(store-guardrails): cover admin-configurable thresholds + PATCH round-trip
Closes the "primary testing gap" Vojta noted in the safe-fix commit
on PR #281 — the four new `get_guardrails_min_*` getters and the
PATCH-takes-effect-on-next-check live-config flow had no direct
coverage.
10 new tests in `tests/test_store_guardrails_admin_config.py`:
- TestGuardrailGetterDefaults (4 tests) — each new getter returns the
documented default (60 / 25 / 5 / 200) when nothing is configured.
- TestGuardrailGetterOverlay (5 tests) — overlay-driven overrides win,
string values that look numeric coerce via int(), garbage strings
fall back to default via the (TypeError, ValueError) branch, and the
`max(1, int(val))` floor pins zero/negative inputs to 1.
- TestPatchRoundTrip (1 test) — PATCH `/api/admin/server-config`
`guardrails.min_description_chars=90`, then call content_check
against a 75-char description that previously passed: must now fail
with `too_short`. Then PATCH back to 60 and verify the next check
passes again. Closes the cache-invalidation contract Vojta relies on
for the "no app restart" claim — broken without the
reset_cache() bracket in /api/admin/server-config.
The TestGuardrailGetterOverlay.test_zero_or_negative_floored_to_one
test pins the current `max(1, int(val))` policy. Vojta's safe-fix
commit explicitly left "policy on min_*=0 vs disable-sentinel" as
residual work — pinning the current behavior here ensures any future
change to use 0 as a disable sentinel must update this test (and the
reviewer sees the policy decision).
Verified: 4509 tests pass locally (4499 existing + 10 new).
* release: 0.54.2 — admin-configurable flea-market guardrail thresholds + tests
Last commit on the PR per CLAUDE.md hard rule. Patch bump (0.54.1 →
0.54.2) bundling Vojta's admin-configurable thresholds for the
flea-market content guardrail (9 knobs in /admin/server-config) plus
the test coverage closing the "primary testing gap" he punted in the
safe-fix commit.
No DB migration; defaults unchanged from PR #276 — instances that
don't set `guardrails.*` keep the original bar transparently.
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
Co-authored-by: ZdenekSrotyr <139972147+ZdenekSrotyr@users.noreply.github.com>
* docs(spec): admin observability spec + Activity Center MVP plan
Parent spec (480 lines) + executable plan (2295 lines, 14 TDD tasks).
Covers Activity Center rebuild (/admin/activity), with /admin/sessions
and /admin/feedback deferred to follow-up plans.
Already incorporates reviewer-pass revisions across three angles
(security, production resilience, code architecture):
- _get_db import path corrected to app.auth.dependencies
- Test fixtures aligned with seeded_app / admin_user / get_system_db
- All new audit writes wrapped in try/except + logger.exception
- Filename sanitization on session uploads
- DuckDB DESC index behavior documented; upgrade window flagged
- Migration idempotency + evolved-DB test cases
- reveal_raw + shared-cache multi-worker explicitly deferred
Targets schema v40 (audit_log gains params_before, client_ip,
client_kind, correlation_id + 3 indices).
* feat(db): schema v40 — audit_log gains params_before, client_ip, client_kind, correlation_id + 3 indices
* chore(test): clean up Task 1 — drop unused import, rename stale test
* feat(audit): AuditRepository.log() accepts params_before/client_ip/client_kind/correlation_id
* test(audit): strengthen params_before assertion to round-trip JSON content
* feat(audit): AuditRepository.query() rich filters + keyset cursor pagination
* feat(sync): SyncStateRepository.list_recent() cross-table feed
* feat(audit): POST /api/sync/trigger writes audit_log row
* feat(audit): POST /api/scripts/run-due writes audit_log row
* feat(audit): POST /api/upload/sessions writes audit_log row + sanitizes filename
* feat(audit): GET /api/data/{table_id}/download writes audit_log row
* feat(activity): /api/admin/activity timeline + /health + /sync endpoints
* feat(ui): /admin/activity rebuilt — health pulse, timeline, sync grid; /activity-center → 308 redirect
BREAKING: removed demo executive-pulse / maturity-roadmap content from activity_center.html.
The page now reflects real audit_log + sync_history data.
* feat(ui): admin nav + dashboard widget point at /admin/activity
* feat(activity): recursive-audit suppression for AC read endpoints (60s window per actor+filter)
* feat(activity): emit PostHog events when integration enabled (no-op default)
* fix(audit): move v40 indices out of _SYSTEM_SCHEMA + update test_repositories to unpack query() tuple
_SYSTEM_SCHEMA CREATE INDEX on audit_log(timestamp) failed when migration
tests hand-roll a bare audit_log (id, action) without the timestamp column.
Fix: remove indices from _SYSTEM_SCHEMA; add ADD COLUMN IF NOT EXISTS guards
for timestamp and other pre-v40 columns in _v39_to_v40() so the upgrade path
is safe on any hand-rolled schema; call _v39_to_v40 explicitly in the
fresh-install (current==0) path to restore index creation there.
Also unpack the (rows, next_cursor) tuple from AuditRepository.query() in
the three TestAuditRepository tests that still treated it as a list.
* docs: CHANGELOG entry for Activity Center MVP
* chore: refresh stale module docstring in app/api/activity.py
* feat(cli): agnes admin activity — terminal access to Activity Center (timeline + health + sync)
* fix(db): _v39_to_v40 — add IF NOT EXISTS guard for 'action' column
The v39→v40 ladder step adds defensive ADD COLUMN IF NOT EXISTS for
every audit_log column so a hand-rolled bare audit_log (id only) is
safe through the ladder. 'action' was missing from the guard list,
causing CREATE INDEX idx_audit_action_time to fail on tests that
stub audit_log with only an id column (tests/test_e2e_extract.py::
TestSchemaMigration::test_migration_preserves_and_extends).
Local 6/6 schema tests + the previously-failing CI test pass.
* docs(spec): platform telemetry epic — Boss directive + Activity Monitoring plan rebased onto v40 (stacked on zs/spec-activity-center)
* feat(db): schema v41 — 7 usage_* tables for telemetry (events, summary, rollups, attribution)
* chore(db): tighten v41 — usage_session_summary.session_id NOT NULL + upgrade test asserts all 7 tables
* feat(usage): UsageAttributionRepository — replace/delete/lookup over usage_attribution_* tables
* refactor(marketplace): extract list_inner_skills/agents/commands to src/marketplace_listing.py for reuse
* feat(usage): explode plugin attribution on marketplace sync + store entity write; backfill script
* refactor(marketplace): finish src/marketplace_listing.py extraction — drop duplicate _list_inner_* + _parse_frontmatter from app/api/marketplace.py
* feat(usage): promote attribution helpers to src/usage_attribution_helpers.py; hook update_entity rename + bundle-swap; clarify best-effort semantics
* feat(usage): UsageProcessor real extraction + rollup rebuild + 10 fixture-driven tests
* fix(usage): include tool_id in event hash + executemany + rollup transaction (critical multi-tool-turn drop fix)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(marketplace): popularity stats — invocations_30d + trend + sort=most_used|trending + Most Popular section
* feat(admin): /admin/users/<id> Sessions section — list + single-file + bulk-zip downloads (audit-logged)
* feat(usage): admin export endpoint + CLI — csv/json/parquet streaming, filters, audit-logged
* feat(usage): agnes admin ask — LLM Text-to-SQL over usage_events with SELECT-only validator (audit-logged)
* feat(usage): reprocess + prune endpoints + scheduler daily prune job + CLI
* docs: PLATFORM_SETUP.md operator playbook + HOWTO/ cookbook (5 guides + index)
Adds docs/PLATFORM_SETUP.md as a consolidated operator playbook covering
bootstrap, TLS, marketplaces (curated + flea), scheduler env vars, telemetry
extraction/export/ask/prune, privacy posture, and daily routine.
Adds docs/HOWTO/ with 5 analyst cookbook guides: first query, snapshots for
remote tables, private sessions, feedback + admin ask, and customizing skills.
Existing setup docs (QUICKSTART, DEPLOYMENT, ONBOARDING, HEADLESS_USAGE)
get a one-line cross-reference at the top pointing to PLATFORM_SETUP.md.
* docs(changelog): platform telemetry epic — usage_* foundation + surfaces + admin access + docs
Comprehensive [Unreleased] entry covering: usage_events/session_summary/
tool_daily/plugin_daily tables (v41), attribution lookup tables, backfill
script, marketplace Most Popular + invocation chips + sort, admin Sessions
section, export/ask/reprocess/prune endpoints + CLI mirrors, Activity Center
(v40), PLATFORM_SETUP.md + HOWTO/ docs, and operations notes for v41 upgrade.
* fix(security): block DuckDB read_*/http_*/glob functions in usage_ask validator + symlink escape guard in session zip + clarify mark-private semantics
* fix(admin): parquet export tempfile cleanup on COPY failure + correct processed-first sort on /admin/users/<id>/sessions
* feat(audit): close 8 production audit gaps — query (local/remote/hybrid), catalog/schema/sample, snapshot estimate/create, check-access
* feat(ui): /admin/usage summary dashboard + per-user activity tab on /admin/users/<id>
* fix(audit): cap error messages at 200 chars + audit user_activity reads + recursion guard on usage.summary
* fix(audit): catalog.list audits on error path + clean up deferred json import
* fix(ux): client_kind=cli for PAT auth + timeline empty state + email-instead-of-uuid + nav reorder + help text + loading indicators + ask doc
* feat(observability): unify /admin/activity into single page with saved views
- KPI cards (events, users, error rate, p95) clickable as quick-filters
- Faceted filter dropdowns populated from audit_log in the current window
- Sortable audit table, cursor pagination, per-row JSON side panel
- Saved views (schema v43: user_observability_views) — per-user state
- Top bar: window selector + 30s Live toggle + saved views dropdown
- /admin/scheduler-runs → 308 redirect (source=scheduler filter)
- New endpoints: /api/admin/observability/{facets,kpis,views}
* test: update activity + scheduler-runs tests for unified page
- test_admin_activity_page_renders asserts new structural anchors
- test_admin_scheduler_runs_page_admin_only asserts 308 redirect
* fix(observability): respect [hidden] on modal + side panel
CSS `display: flex` on .obs-modal beat the [hidden] attribute's UA
display:none, so the save-view modal rendered on page load and Cancel
clicks couldn't dismiss it. Gate the modal's flex layout on
:not([hidden]); add the same display:none guard prophylactically to
.obs-panel and .obs-views-panel.
* feat(observability): user enrichment in audit + interactive /admin/usage
Activity:
- /api/admin/activity now joins users for user_email + user_name per row
- User column renders "name (id-prefix)" or "email (id-prefix)" instead
of an opaque truncated UUID; falls back to id when the user record is
missing
Usage:
- /admin/usage rewritten as the same filter/group-by/search pattern as
/admin/activity. Faceted dropdowns (User / Tool / Source / Event type)
populated from usage_events; debounced free-text search across
tool_name / skill_name / subagent_type / command_name
- New endpoints /api/admin/usage/{facets,kpis,query}; the query endpoint
supports group_by in {day, username, tool_name, source, ref_id} with
sort + offset pagination, plus an ungrouped raw-events mode
- 4 KPI cards (events, distinct users, distinct tools, error rate) are
clickable quick-filters; clicking a grouped row applies the bucket as
a filter
- Old static `?window=7d|30d|all` server preload removed; all state is
client-side via since_minutes + group_by + filters in the URL
* fix(observability): clearer labels, all-column sort, drop saved views UI
- Rename page titles: "Activity" → "Server activity", "Usage" → "Tool usage"
with a one-line subtitle on each explaining what the page covers and
linking the other one. The two pages source different data (audit_log
vs usage_events) and the previous labels conflated them.
- Drop the saved-views dropdown + save modal from /admin/activity. The
modal pop-open bug was the trigger; the value wasn't there yet. The
/api/admin/observability/views CRUD + DuckDB table stay in place.
- Rename "Live (30s)" to "Auto-refresh (30s)" with a tooltip clarifying
that it's the re-fetch rate, not the time range. Time range now
labeled "Time range" instead of "Window".
- All audit-table columns are sortable (User, Source, Action, Resource,
Result added); sort is page-local with a Jinja comment explaining the
trade-off. Same for raw usage rows.
- Fix duplicate sort-arrow bug — the literal "▼" in the Time th HTML was
rendering alongside the CSS ::before arrow. Removed the literal; CSS
is the single source of truth.
* feat(observability): global Sessions browser + transcript viewer + CLI
Web:
- /admin/sessions — list every collected session JSONL across all users
with time-range, user, model, errors-only and free-text filters. Default
sort surfaces error-heavy sessions first. KPI cards (sessions, distinct
users, sessions w/ errors, tool error rate) clickable as quick-filters.
- /admin/sessions/<username>/<file> — transcript viewer rendering the
JSONL chronologically: user prompts, assistant text, tool calls (with
JSON input) and tool results (with flattened output). Errors get a red
border + chip and a "Next error" navigation button at the top.
- Admin dropdown gains a "Sessions" link.
API:
- GET /api/admin/sessions/{list,kpis,facets} — filtered cross-user reads
off usage_session_summary
- GET /api/admin/sessions/{username}/{file}/transcript — parses JSONL via
the existing services.session_pipeline.lib, returns chronological events
- GET /api/admin/sessions/{username}/{file}/download — JSONL stream, same
path-safety guards as the per-user endpoint, audit-logged
CLI:
- `agnes admin sessions list [--user X] [--errors] [--since 7d]` — table
output with `!` prefix on rows that hit a tool error
- `agnes admin sessions show <username> <file>` — transcript dump, with
`--errors` to print only the failed tool_result blocks
- `agnes admin sessions download <username> <file> [-o path]`
- `agnes admin sessions kpis` — top-level numbers
* feat(internal): expose telemetry tables to agnes query with row-level RBAC
Three new registered tables backed by system.duckdb, queryable through
the same /api/query plumbing analysts use for Keboola / BigQuery /
local sources:
agnes_sessions → usage_session_summary (filter: username)
agnes_usage → usage_events (filter: username)
agnes_audit → audit_log (filter: user_id)
RBAC is per-row, not per-table: admins see every user's rows; non-admins
see only their own. The filter is built server-side from the auth user
dict; non-admin filter values are regex-validated before SQL interpolation.
Implementation:
- new connector connectors/internal/ with access (filter+exec) + registry
(idempotent table_registry seed at startup)
- /api/query detects internal table refs and short-circuits to a CTE
wrapper that prepends "WITH agnes_x AS (SELECT * FROM <src> WHERE …),
…" then "SELECT * FROM (<user_sql>) AS _q". DuckDB cursor on the
shared system.duckdb handle — opening parallel handles / ATTACH on the
same file is blocked process-wide.
- mixing internal + BQ / registered local tables in one SELECT is
rejected (v1 limitation)
- src.rbac.can_access_table waves internal tables through for all
authenticated users; row scoping is the actual security control
- /api/v2/schema and /api/v2/sample gained internal branches; sample
intentionally skips its cache because rows are RBAC-scoped per caller
- audit row written as action='query.internal' with is_admin flag
Tests: connectors/internal/access — RBAC, filter clause, schema, CTE
wrapper coexistence with user-supplied aggregations, unsafe-username
rejection. 16/16 passing.
Motivating queries this enables:
SELECT tool_name, COUNT(*) FROM agnes_usage
WHERE is_error GROUP BY 1 ORDER BY 2 DESC
-- analyst self-introspection: which tools fail for me?
SELECT user_id, COUNT(*) FROM agnes_audit
WHERE action = 'session.transcript_view' GROUP BY 1
-- admin: who's been looking at whose session transcripts?
* feat(admin): group dropdown into 5 named sections + internal tables in /catalog
Admin dropdown gains section headers so admins can land on the right
page without re-reading the full menu:
Activity Center Server activity / Tool usage / Sessions
Users & Access Users / Groups / Resource access / Tokens
Data Tables
Agent Experience Curated Marketplaces / Flea Submissions /
Agent Setup Prompt / Agent Workspace Prompt
Server Server config
"Agent Experience" frames the curated content + prompts as one cluster
— it's all admin-controlled material that shapes what an analyst's AI
agent encounters. "Configuration" → "Server" since only one item lives
there now.
Renamed the section's first two items:
"Activity" → "Server activity" (matches page H1)
"Usage" → "Tool usage"
Also fixes /catalog visibility of the internal tables (agnes_sessions /
_usage / _audit) for non-admin users: ``app.auth.access.can_access``
short-circuits to True for resource_type='table' + an internal-table id.
Without this, non-admins saw the tables in /api/v2/catalog (which uses
the same RBAC bypass) but not on the /catalog HTML page (which calls
can_access directly, requiring a resource_grants row internal tables
don't have).
CSS for `.app-nav-menu-section`: small caps, muted, non-clickable; first
section trims top padding so the panel doesn't open with an awkward gap.
* refactor(admin): move corporate memory into Admin > Agent Experience
Memory link was the only admin-only entry in the primary nav (gated by
session.user.is_admin). Moves it into the Admin dropdown under Agent
Experience, alongside Curated Marketplaces / Flea Submissions / Prompts
— all admin-curated content that shapes what an analyst's AI agent
encounters.
Renamed the nav label to "Shared Knowledge" to match what the page
actually is (admin-curated organisational knowledge from session
verification, surfaced to agents). URL stays at /corporate-memory; the
route still gates on require_admin per the existing comment.
Side effect: primary nav (Home / Marketplace / Data Packages) is now
uniform for every authenticated user — no conditional admin-only entry.
* ui: rename admin entries to Curated Knowledge / Init Prompt / Workspace Prompt
- "Shared Knowledge" → "Curated Knowledge" (parallel with "Curated
Marketplaces" in the same Agent Experience section; "curated" tells
the admin what they do there — review + approve)
- "Agent Setup Prompt" → "Init Prompt" (matches the `agnes init` flow
it actually drives)
- "Agent Workspace Prompt" → "Workspace Prompt" (the "Agent" prefix
was redundant — every item in the section is agent-facing)
Renames page titles + H1s on /admin/agent-prompt and
/admin/workspace-prompt to match.
* refactor: rename Usage → Telemetry across user-facing surfaces
External surfaces all switch; internal Python module / file names and the
physical DB tables (usage_events, usage_session_summary, usage_tool_daily,
usage_plugin_daily) stay — renaming them would force a schema migration
+ a redo of the LLM Text-to-SQL prompt for no analyst-visible win.
Changes:
- Admin dropdown: "Tool usage" → "Telemetry"
- Page H1 / <title>: same
- URL: /admin/usage → /admin/telemetry; old URL 308-redirects
- API prefix: /api/admin/usage/* → /api/admin/telemetry/*
- CLI: primary command `agnes admin telemetry …`; `agnes admin usage` kept
as a deprecated alias so existing operator scripts keep working
- Internal data-source table id: agnes_usage → agnes_telemetry. The
registry seed now evicts any stale internal-source row whose id no
longer matches INTERNAL_TABLES, so the old `agnes_usage` row is
removed from table_registry on next app boot
- All tests + JS endpoint paths updated
* test(rbac): include auto-appended internal tables in expectations
get_accessible_tables now appends agnes_sessions / agnes_telemetry /
agnes_audit to every authenticated user's accessible-tables list so the
internal data source shows up in /catalog. The two existing rbac tests
asserted hardcoded list shapes that pre-dated the change.
Rewritten to assert "granted tables + the canonical internal-table set"
instead of literal lists, so the test stays correct if the internal
table roster changes again later.
* ui: visual dividers between admin-dropdown sections
Adds a 1px top border + 6px top margin to every section header except
the first, so the five named groups (Activity Center, Users & Access,
Data, Agent Experience, Server) read as visually separated clusters.
The header itself stays small-caps + muted as before — the border is
additive.
* ui(memory): match obs-topbar visual on /corporate-memory
The Curated Knowledge page (linked from the admin dropdown's Agent
Experience section) opened straight into the stats bar — no title,
no subtitle, no shared chrome with the other admin pages. Adds an
obs-topbar-style header at the top of .container-memory:
- H1 "Curated Knowledge"
- subtitle explaining what the page is + how AI agents pull from it
The `.ck-*` class set duplicates the inline obs-* styles from
/admin/activity etc. for this one page; promoting the obs-* class set
to style-custom.css for shared reuse is the obvious next step (4 pages
already inline the same CSS), tracked as a follow-up.
Page <title> also renamed from "Corporate Memory" → "Curated Knowledge".
* ui(tables): list Agnes internal tables in /admin/tables + group in /catalog
/admin/tables previously rendered three per-source-type listings
(BQ / Keboola / Jira) and dropped any row whose source_type didn't
match — so the agnes_sessions / agnes_telemetry / agnes_audit rows
seeded into table_registry were invisible. Adds a fourth read-only
section "Agnes internal tables" that filters source_type === 'internal'
and renders the same registry-table layout the other sections use,
with two changes:
- no Register button (these rows are seeded on every app boot from
connectors/internal/registry.py)
- Edit + Delete actions hidden (any change would be reverted on the
next start). Manage access stays so admins can still inspect.
Mode badge picks up a new mode-internal CSS class (teal accent) so the
display doesn't lie and call it "local".
In /catalog, internal tables now group under an "agnes" accordion
section (bucket="agnes" on seed) instead of falling into the catch-all
"default". Single source of truth for which tables exist; admins find
them where they expect.
* ui(tables): Agnes internal as a 4th tab next to BQ/Keboola/Jira
Previous iteration mounted the internal-table listing as a separate
standalone card under the tab strip. Reshapes it to a proper
tab-content section so admins switch between data sources via one
consistent nav (BigQuery / Keboola / Jira / Agnes internal).
- New tab button "Agnes internal" in the tab-nav.
- The listing card becomes <section id="tab-content-internal"
class="tab-content">; switchTab() already routes by id so no JS
change beyond extending the hash allowlist for direct #internal
links.
- Tab content keeps the read-only treatment from the previous commit
(no Register button, no Edit / Delete in renderRegistryListing).
* ui: rename Curated Knowledge → Curated Memory
Settles the naming back on "Curated Memory" — parallel structure with
"Curated Marketplaces" in the same Agent Experience section, and zero
rename ripple: URL (/corporate-memory), API (/api/memory/*), CLI
(agnes admin memory), and Python modules all stay on "memory" so the
admin label finally lines up with the underlying surfaces.
The "Curated" prefix still tells admins what they do on the page
(review pending → approve / mandate / reject) and reads as a sibling
of "Curated Marketplaces" right next to it in the dropdown.
Touches: admin dropdown label, page <title>, page H1. DB tables stay
on knowledge_* (already the canonical naming for the data shape).
* ui: rename "Server activity" → "Audit log"
"Audit log" is what the page actually is — server-side audit_log table
rendered with KPI cards + filter bar + sortable table. The "Server
activity" label confused the term with Claude Code session telemetry
(Telemetry page) and didn't make the source/concept clear.
Touches:
- Admin dropdown nav label
- /admin/activity page H1 + subtitle
- /admin/telemetry subtitle cross-link
- test_activity_api page-renders assertion
URL (/admin/activity) and API (/api/admin/activity/*) stay — the
"activity" name has stuck at the route layer for a year; rerouting
those would churn dashboards/bookmarks for zero analyst-visible win.
* ui(admin-nav): gray band on each section header for clearer separation
Previous iteration used a 1px top border between section labels — the
labels still blended into the items above/below at a glance. Switches
to a light gray background band per section header, extended edge-to-
edge inside the panel via negative horizontal margins. Bolder
font-weight (700) reinforces the separation; bumping the font color
isn't needed because the band itself does the work.
First section's header tucks into the panel's top border-radius so the
band reaches the corners without a gap.
* ui(catalog): rename internal-table category to "Agnes Internal"
`bucket` is what /catalog renders as the accordion category header
verbatim — "agnes" lowercase didn't read as a real category name and
got confused with a system identifier. Bumps to "Agnes Internal".
Seed re-applies on every app boot so existing rows pick up the new
bucket value via `ON CONFLICT (id) DO UPDATE`.
* ui(catalog): split Agnes Internal into its own card on /catalog
Previously the three internal tables landed inside the "Core Business
Data" card under an "Agnes Internal" accordion alongside Keboola / BQ
buckets — readers conflated system telemetry with business datasets,
and the data_stats header counter ("3 tables · ~X rows total") only
ever counted synced rows so internal tables looked invisible.
Split the catalog page into two cards:
- Core Business Data: only non-internal source_types (Keboola, BQ,
Jira). Accordions group by bucket as before. Stats counter reflects
this card's tables.
- Agnes Internal: a dedicated card with its own visual treatment
(teal accent matching the mode-internal badge in /admin/tables).
Flat list (no accordion — only 3 rows, never grows here), each
row carries the canonical `agnes query` snippet. Read-only — no
profiler click, no In-stack toggle, no sync metadata.
Route adds `internal_card` context object; template renders the new
card only when it's non-None.
* fix(rbac): hide internal tables from /admin/access + drop "my" framing
Two related cleanups for the Agnes-internal tables:
1. /admin/access (resource grants) no longer lists them. The
`can_access` check has a hardcoded internal-table bypass — security
is row-level (per-request view filter), so a table-grain
`resource_grants` row would do nothing. Surfacing them in the UI
let admins set up grants that silently no-op. Filter at the
`_table_blocks` projection so the UI tree never sees them.
2. Display names drop the analyst-perspective "my" framing:
"Agnes — my sessions" → "Agnes sessions"
"Agnes — my telemetry events" → "Agnes telemetry events"
"Agnes — my audit log" → "Agnes audit log"
The "my" only makes sense from the querying analyst's seat
(`SELECT … FROM agnes_sessions` returns *their* rows); on /admin/*
pages where admin sees / configures them across users, the
pronoun was misleading. Description text now spells out the
row-level RBAC contract explicitly.
Display names update via TableRegistryRepository.register's ON CONFLICT
UPDATE on next app boot; no manual cleanup needed.
* ui: subtitle notes about agnes_* tables on each Activity Center page
The recursive observability story — Agnes serves its own audit /
telemetry / session data through the same `agnes query` plumbing
analysts use for business data — wasn't surfaced anywhere on the
admin pages that show that data. Three pages get a one-liner with
the canonical `agnes query` snippet + the RBAC contract (analysts
see their own rows, admin sees all):
- /admin/activity (Audit log) → agnes_audit
- /admin/telemetry (Tool usage) → agnes_telemetry
- /admin/sessions → agnes_sessions
Sets up the discovery moment for admins: they're reading the page,
they see "you can query this from Claude Code", they remember it
when an analyst asks "how do I find my own failed tool calls?".
* ui(tables): explain "Show log" empty-state on /admin/tables
Cache warmup log <pre> renders with a dark background and is only
populated by the SSE stream during a Re-warm all run. Opening the
page cold + clicking Show log just revealed a black bar with no
context — admins couldn't tell what they were looking at.
Adds an inline paragraph above the <pre> explaining what the log is,
the row format, when it fills in, and where to find the historical
audit trail (/admin/activity). The actual <pre> stays empty until
SSE events arrive, but the surrounding copy carries the meaning.
* ui(tables): auto-open cache-warmup log on Re-warm all click
A Re-warm all run takes ~24s per remote BQ row. With the <details>
collapsed by default, operators saw the button disable, watched a
quiet ~24s pass, and assumed nothing had happened — the streaming
log was hidden behind a closed disclosure.
Two small JS tweaks:
- cacheWarmupRun() opens the details on click, so streamed lines
appear without an extra interaction
- cacheWarmupOnStart() hides the inline hint paragraph the moment
real log content lands, so the dark log block isn't competing
with redundant context
Hint paragraph also clarifies that only `query_mode='remote'` BQ
rows are warmed — operators with only materialized/internal tables
would see total=0 and the page would "do nothing" by spec.
* ui: trim Agnes internal copy across surfaces
Descriptions had grown to explain the extraction pipeline ("parsed
out of session JSONLs"), the underlying table ("Backed by
usage_session_summary"), the RBAC mechanic ("row-level RBAC at query
time — analysts see their own; admin sees all"), and the SQL snippet.
Every implementation detail meant another rewrite on the next iter.
Strips to one stable line per surface: what the data is, plus
"Also available locally for analysis". Mechanics live in code +
docs; the page copy says what the user needs to know.
Touched:
- connectors/internal/access.py: INTERNAL_TABLES descriptions
- activity_center.html / admin_usage.html / admin_sessions.html
subtitles
- catalog.html Agnes Internal card description + row strip
- admin_tables.html "Agnes internal" tab hint
* fix(internal): is_user_admin arity bugs + + saved-view payload cap
Round-1 code review (PR #278) caught two blocking bugs and three nits.
Blocking — both `is_user_admin(user)` (single dict arg) calls raised
TypeError. is_user_admin signature is `(user_id, conn)`. Affected:
- app/api/query.py:_run_internal_query — every POST /api/query that
references agnes_sessions / agnes_telemetry / agnes_audit blew up
with a 500. The headline analyst-facing feature of this PR was
unusable through the API.
- app/api/v2_sample.py — same shape; `GET /api/v2/sample/agnes_*`
returned 500.
Both fixed to call `is_user_admin(user.get("id"), conn)`. Added two
FastAPI-level tests in test_internal_data_source.py that go through
the TestClient — the existing unit tests on `execute_internal_query`
and `build_filter_clause` skipped the request-handler layer where the
bugs lived, which is why this landed.
Nits also closed:
- connectors/internal/access.py: `+` allowed in _USERNAME_RE /
_USER_ID_RE so RFC 5321 email local-parts (alice+test@x) resolve
correctly without hitting InternalAccessError.
- app/api/observability.py: saved-view payload capped at 64 KiB to
prevent an admin from bloating system.duckdb with a malformed save.
* fix(security): close non-admin data-leak via underlying-table refs
PR #278 R2 review surfaced a non-admin-exploitable bypass: SQL whose
string literal contains 'agnes_sessions' routed into the privileged
internal-query path, then queried the underlying physical table
(usage_session_summary / usage_events / audit_log) directly, escaping
the CTE wrapper's row filter. Two reinforcing defenses:
1. find_internal_refs() now strips single-quoted string literals
before scanning for alias names — a literal alone no longer
routes the request into the privileged code path.
2. execute_internal_query() rejects non-admin SQL that references
the underlying physical tables (usage_*, audit_log). The CTE
wrapper only scopes the agnes_* aliases; a direct FROM on the
base table — or a shadowing inner WITH that still has to read
the base table — bypasses RBAC. Block before execution with an
actionable error pointing to the agnes_* alias. Admins are
unaffected (god-mode short-circuit on the filter clause).
3. tests/test_internal_data_source.py — three new negative tests
covering literal-only matches, direct-table refs, and CTE
shadow attempts.
Also tightens usage_ask.py's SELECT-only validator: pragma_table_info,
pragma_storage_info, pragma_database_*, and duckdb_tables / columns /
views / indexes / schemas are reflection functions that leak metadata
the analyst question shouldn't reach. \bPRAGMA\b in _FORBIDDEN never
matched the function-call form (word-boundary between `A` and `_`).
* fix(security): dynamic denylist for non-admin internal queries
R3 review (PR #278) caught a wider data-leak than R2: the underlying-
physical-table guard listed only the 7 usage_* + audit_log tables,
but system.duckdb has 30+ other sensitive tables — users (emails +
ids), personal_access_tokens, resource_grants, user_groups,
user_observability_views, store_*, marketplace_*, knowledge_*, etc.
A non-admin SQL like
SELECT * FROM agnes_sessions
UNION ALL SELECT email, id, … FROM users LIMIT 1
would leak every user's row.
Replaces the hardcoded denylist with a **dynamic allowlist** —
non-admin SQL may reference ONLY the registered agnes_* aliases.
Every other table in `information_schema.tables` (main schema) is
rejected. Future migrations that add a new sensitive table are
automatically covered without re-editing this module.
Also strips SQL comments (`/* */` and `--`) before the identifier
scan so a comment-wrapped table name (`/**/users/**/`) can't slip
past the regex.
Four new negative tests pin: `users`, `personal_access_tokens`,
block-comment wrap, line-comment wrap.
Plus: per-user view-count cap (100) on /api/admin/observability/views
so an admin can't fill system.duckdb with thousands of saved views.
* release: 0.54.0 — Activity Center + Telemetry + Sessions + internal datasource
Cuts the work shipped across this PR (Activity Center build, recursive
internal data source) into a versioned release. Bumps pyproject.toml
to 0.54.0; renames the top of CHANGELOG.md from [Unreleased] to
[0.54.0] — 2026-05-12 with a header summary; opens a fresh
[Unreleased] section for the next round.
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
* feat(store-guardrails): enforce per-component description quality
Two-tier hard guardrail on flea-market submissions. Empty / placeholder /
single-word descriptions now block before any LLM call; vague-but-passes-
floor descriptions block on the substantive LLM review layer.
Tier 1 — inline mechanical check (src/store_guardrails/content_check.py).
Walks the baked plugin tree, evaluates each component (plugin manifest,
agents, skills, commands) plus the submission-level form description
against a 60-char / 25-char (commands) / 5-distinct-word / 200-char-body
floor with a placeholder denylist (TODO, TBD, {{var}}, etc.). Floors
calibrated against real ecosystem norms: Claude / superpowers /
compound-engineering skill packs cluster 150–220 chars, npm / Docker /
VS Code at 100–120. InlineResult.passed now ANDs in content.status.
Tier 2 — LLM review extension (prompts.py + llm_review.py). System
prompt gains a content-quality criterion; REVIEW_JSON_SCHEMA carries a
content_quality {verdict, issues[]} object alongside the existing
security findings. is_safe() requires content_quality.verdict == 'pass'.
Single LLM call covers both dimensions. MAX_RESPONSE_TOKENS bumped
2000 → 2500 for the extra payload. Verdicts missing content_quality
treated as pass (backwards compat with already-recorded rows).
Submitter UX:
- /store/new wizard now carries a "Before you upload — what passes
review" collapsible disclosure on both step 1 and step 2 with the
bar + patterns that work. Live char counter on the description
field. Per-component preview table (green/red dots from the new
summarize_for_preview helper) renders after the ZIP /preview round
trip, scoping each finding to its file.
- New /store/examples page with rejected/passes pairs for skill /
agent / plugin / command plus a "Why these limits" research table.
Anchored sections (#skill / #agent / #plugin / #command) so the
rejection banner can deep-link by component_type.
- Quarantine banner _content_findings.html groups findings by file
(one "See <type> example ↗" per component, not per field) and
translates field codes (frontmatter.description / body / etc.) to
plain-English labels. _content_howto_fix.html surfaces a static
"Re-upload as new version" + "See examples" action row beneath any
content failure on the entity detail page.
- _parse_frontmatter moved to src/store_guardrails/_frontmatter.py so
the new check module shares the parser without inverting the
app → src dependency direction.
Tests:
- New tests/test_store_guardrails_content.py (29 cases) covering
every failure code per component type plus submission-level checks
and the summarize_components / summarize_for_preview helpers.
- Extended test_store_guardrails_inline.py for the new
InlineResult.content field + aggregate behaviour.
- Extended test_store_guardrails_llm.py for the new
content_quality verdict pathways (fail blocks, missing field passes).
- Backfilled fixture descriptions across test_store_api.py,
test_store_entity_versions.py, test_store_put_atomic.py,
test_admin_store_submissions.py, test_marketplace_api.py,
test_marketplace_v32_endpoints.py so existing happy-path tests
clear the new 60-char floor.
* fix(content-guardrail): align agents walker with preview + drop import-time .format()
Two cleanups from the takeover review on #276 (vr/guardrails-content).
1) `_iter_components` for agents now skips files lacking frontmatter
(no `name` AND no `description`). Pre-fix the walker greedily
evaluated every `*.md` under `agents/` — `agents/README.md` and
helper docs got flagged as "frontmatter.description empty"
rejections. Worse: `summarize_for_preview` for `type=agent` ALREADY
filters the same shape, so the upload preview gave a green dot
while the post-bake check gave a red rejection on submit. Two new
regression tests in TestAgentsWalkerSkipsNonAgentFiles pin both
shapes (README + _NOTES.md) so the preview/check parity stays
aligned.
2) `body_too_short` hints now use the same runtime-kwarg substitution
pattern as every other hint in the table. Pre-fix the skill +
agent body_too_short hints called `.format(min_chars=_MIN_BODY_CHARS)`
at module-load time, but the call site `_hint_for(type_,
"body_too_short")` didn't pass `min_chars=`, so the format() was
just baking the constant at import. Cosmetic inconsistency; pass
`min_chars=_MIN_BODY_CHARS` at the call site instead and let
`_hint_for` do the substitution like it does for `too_short`.
Verified end-to-end:
- New TestAgentsWalkerSkipsNonAgentFiles cases fail on the unfixed
walker (verified by reverting to the pre-fix file and re-running);
pass cleanly after the fix.
- Full content-guardrail suite: 25/25 (23 existing + 2 new).
- Full pytest: 4189 passed, 25 skipped.
* release: 0.53.5 — content guardrail (flea-market submitter UX) + catalog ENTITY column + BQ hint dispatch
Bundles three threads landed in [Unreleased]:
- Vojta's flea-market content guardrail (two-tier mechanical + LLM)
- Zdeněk's `agnes catalog` ENTITY column replacement for FLAVOR
- Zdeněk's `/api/query` remote_estimate_failed hint dispatch fix
Plus the takeover hygiene from #276 review (agents walker preview/check
parity + body_too_short hint runtime kwarg consistency) and the
backslash-escape fix follow-up to v0.53.4 #275.
No DB migration; no API change. Patch upgrade lands transparently.
Upload form's new "Before you upload" disclosure + per-component preview
table appear on the next dev-VM auto-pull. Quarantine banner now groups
findings by file with "See <type> example ↗" deep-links to the new
/store/examples reference page.
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
Three client-side fixes in admin_tables.html plus a regression test
file pinning the server-side PUT contract the new JS relies on.
Bug 1 — saveBqTabEdit (synced/custom) nulled bucket/source_table on
every save; the null was supposed to clear stale state on a true
remote→materialized mode flip but fired on every save, silently
wiping persisted bucket/source_table when admin edited only the
description on an already-materialized row. Now gated by
_editOriginalQueryMode !== 'materialized'.
Bug 2/3 — _buildBigQueryPayload (synced/whole) at register time did
not send bucket/source_table — only source_query — so whole-table
materialized rows persisted with bucket=NULL. Edit modal then loaded
empty Dataset/Table inputs over a SELECT * SQL. Register now sends
both fields; _openEditBqModal additionally parses source_query as
a fallback for rows that registered pre-0.53.1.
Closes#266.
- instance.brand (env AGNES_INSTANCE_BRAND, default "Agnes") +
instance.workspace_dir replace hard-coded "Agnes" / "~/Agnes" across
/home, /setup, /setup-advanced, /login, /install, /me/debug, and the
Claude Code clipboard setup script. Terraform-friendly env override;
defaults preserve existing Agnes branding.
- Explicit "create workspace folder" step on /home (OS-tabbed mkdir+cd)
+ same step baked into the clipboard script as step 2. Drops the
implicit assumption that `agnes init --workspace .` lands in a
sensibly-cd'd shell.
- Final "Restart Claude Code" step in the setup script (unconditional,
between connectors and Confirm) so freshly-installed plugins, MCP
servers, and SessionStart hooks load on the next Claude Code session.
- Asana reverted from hosted Remote MCP back to PAT + raw REST against
app.asana.com/api/1.0. MCP envelope shape consumed ~5x tokens per
call; the PAT path lets the agent read flat REST fields. Existing
MCP registration is detected and the user is asked whether to remove
it (default Y, with benefits listed: token cost, no third-party hop,
no OAuth refresh dance, deterministic envelope shape).
- Atlassian connector instructs picking the longest API-token expiry
(today "1 year") to cut re-mint friction. No public query-parameter
hook exists on id.atlassian.com to pre-select expiry, so the prompt
documents the manual click and acknowledges that limitation.
- Uniform ✅ / ❌ per-connector marker contract (Asana, GWS, Atlassian)
for the Confirm summary to grep. Each connector now ends with a
Claude-driven end-to-end test that uses Claude Code's own bash to
exercise the stored credential and prints
"✅ <Connector> integration verified — ..." (or the failure variant).
* release: 0.53.0 — Tier B trackers + admin UI bugfix
Closes#259 (init resume sentinel), #260 (startup parquet-lock sweep),
#261 (materialized schema uses local parquet, not BQ), #265 (admin
tables apostrophe → HTML-entity escape).
Tracker notes: #262 closed as obsolete (pre-empted by 0.51.0 changes),
#266 left open pending UX clarification.
* fix(init): move resume sentinel from .agnes/ to .claude/
The clean-install integration test (test_clean_install_integration.py)
forbids creating .agnes/ in the workspace root via its
forbidden_unconditional list — that path is reserved for ~/.agnes/ in
the user's HOME (marketplace clone, CA bundle).
.claude/ is already created by agnes init for settings.json + hooks,
so dropping init-complete next to those keeps the resume sentinel
consistent with the rest of Claude Code's workspace surface and lets
the clean-install assertions pass.
Issue #259.
* docs(changelog): point #259 entry at new .claude/init-complete path
Follows the sentinel move from .agnes/ → .claude/ to keep the changelog
in sync with what 0.53.0 actually ships.
* fix(memory-admin): pass RBAC user_groups (not YAML config) to /corporate-memory/admin
`GET /corporate-memory/admin` was passing `corporate_memory.groups` from
instance.yaml (a dict, default `{}`) as the template's `groups=`. The
template's `renderItemCard` does `GROUPS.map(g => ...)` to build the
mandate-form audience picker, which throws `{}.map is not a function`
the moment any pending item renders. The thrown TypeError bubbles up
through `renderReviewItems` and gets swallowed by the `loadReviewQueue`
catch block, which paints "Error loading pending items." over a
perfectly valid `/api/memory/admin/pending` response.
Bug was dormant since the initial system commit because `renderItemCard`
only runs when at least one pending item exists, so test fixtures and
empty queues never tripped it.
Mandate form actually targets RBAC user_groups (audience strings like
`group:Admin`), not the unrelated `corporate_memory.groups` YAML
section. Route now passes the user_groups list shaped as
`[{name, members_count}]`. Template additionally guards the `.map`
call with `Array.isArray(GROUPS) ? GROUPS : []` so a future shape
regression degrades to "no group options" instead of crashing the list.
No DB migration; no API change.
* test(memory-admin): assert /corporate-memory/admin renders GROUPS as array
Server-side regression for the bug Vojta's previous commit fixed: the route
used to pass `corporate_memory.groups` (a dict, default `{}`) into the
template as `groups=`, so `const GROUPS = {{ groups | tojson }}` rendered
as `{}` and `GROUPS.map(...)` inside renderItemCard threw at runtime.
The bug was dormant because renderItemCard only runs when at least one
pending item exists — empty queues never tripped it. Test seeds one
pending knowledge_item and asserts the rendered HTML contains
`const GROUPS = [` (array prefix), so any future regression that flips
the variable back to a dict shape fails immediately rather than waiting
for an operator to seed the queue and notice the broken admin page.
* release: 0.51.1 — corporate-memory admin pending-banner hotfix
Patch bump: ships the /corporate-memory/admin GROUPS-shape fix
(dict → array of {name, members_count}) and its server-side regression
test. No DB migration, no API change, no operator action required —
upgrades land transparently.
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* Rename agnes-metadata.json to marketplace-metadata.json
Curated marketplace enrichment file (.claude-plugin/agnes-metadata.json)
becomes marketplace-metadata.json. Clean cut, no fallback — curators of
upstream marketplace repos must rename the file on their side.
Python API renames mirror the file rename: read_agnes_metadata →
read_marketplace_metadata, AGNES_METADATA_REL → MARKETPLACE_METADATA_REL,
AGNES_METADATA_MAX_BYTES → MARKETPLACE_METADATA_MAX_BYTES. Synth Claude
Code marketplace strip rule (.agnes/** + the metadata file) follows the
new filename.
* Marketplace detail polish: window cover + 715:310 aspect + helper alignment
- Plugin & item (skill/agent) detail hero: 160x160 square cover replaced
with a macOS-style window frame (3 traffic-light dots + titlebar label
showing the entity name). Body is constrained to 715:310 so curator-
uploaded covers no longer crop to a square. Window is 380px wide; meta
column and absolutely-positioned top-right install/remove actions stay
put. Fallback when no cover_photo_url (translucent gradient + PL/SK/AG
initials) is unchanged, just inside the window body.
- Inner skill/agent cards in the plugin detail's Internal structure
section adopt the same 715:310 aspect (was fixed 78px tall). No window
chrome on inner cards — just the matching proportions so covers read
consistently across hero, grid tiles, and listing cards.
- Curated nested item helper text ("This skill is part of ... — add the
bundle to your stack to use it") now stacks UNDER the "Open parent
plugin" button instead of being a side-by-side flex sibling in the
actions-row. Added align-self: flex-end so the 260px helper box
anchors at the right edge of the 300px actions column, matching the
button's right edge.
* Marketplace My tab: surface the same category + type filters as Flea
- Frontend: mp-cat-row and mp-type-row now show on tab=my (previously
hidden — type was flea-only, category was flea/curated-only). Curated
browse stays plugin-only and continues to hide the type pills.
fetchOne() sends the `type` param for tab=my too, so the items
endpoint's existing my-branch filter actually receives it.
- Backend categories endpoint, tab=my branch: when the type filter is
set to skill/agent, skip counting curated subscriptions. Curated
plugins are always type='plugin', so they wouldn't survive the items
endpoint's type filter; including them in the category counts made
the pill numbers overstate what users could actually see in the
grid. type=None or type='plugin' keeps the previous behaviour.
- CHANGELOG entry under [Unreleased].
* Marketplace plugin detail: render rich content from marketplace-metadata.json
Adds five optional plugin-level fields to marketplace-metadata.json and
renders them on the curated plugin detail page + listing card:
* display_name — friendly h1 / listing-card name / mac-window titlebar
label (overrides the technical plugin id)
* tagline — punchy 1-line value prop for the hero subtitle and the
listing card description (replacing the verbose marketplace.json
description on cards)
* description — multi-paragraph markdown body, server-side rendered
through markdown-it-py and sanitized through nh3 with a
description-scoped allowlist (no iframes / no raw HTML / no
javascript: links). Powers the "What it does" panel.
* use_cases[] — {title, description, prompt} entries that render as a
3-column "When to use it" card grid; each card shows the literal
prompt as a code chip so users can copy-paste into Claude Code.
* sample_interaction — {user, assistant} dialog rendered in a Claude
Code-style dark Catppuccin Mocha transcript panel: monospace user
row with a green ">" prompt indicator + sans-serif assistant body
with markdown formatting (peach bold, yellow italic, pink inline
code, mantle-dark fenced code blocks).
All five fields are optional; UI sections only render when populated,
so plugins without enrichment look identical to before. Fields are
read on-demand from the working tree (cached by mtime per marketplace
slug) so curator edits land at the next request without waiting for
a sync cycle — same pattern as the existing inner-skill/agent
enrichment path. No DB schema bump.
Skill / agent rich-content rendering is deferred to a later phase
(needs a source-of-truth decision: extend plugin.yml? LLM-generate
from SKILL.md / agent.md?). The schema accepts the same fields at
skill/agent level today for forward compatibility but the UI ignores
them for now.
Also: stripped a stale `background-color: var(--bg)` from the global
`code` rule in style.css (was making inline code visually disappear
on the page background).
* Skill / agent detail: render rich content from marketplace-metadata.json
Brings the skill/agent detail pages to parity with the plugin detail
page. Same rich-content schema (display_name, tagline, description as
markdown, use_cases[], sample_interaction) plus two per-item additions:
* invocation — curator-provided literal command string. When set,
overrides the computed "<manifest_name>:<inner_name>" chip and
cleanly supports both "/" skill prefix and "@" agent prefix (the
hardcoded "/" in the chip markup is hidden when the curator provides
the invocation, so /grpn-eng:query <q> and @grpn-eng:cto-architect
both render correctly).
* when_to_use — markdown disambiguation block ("Use this for X. For
similar Y, see /other-skill") rendered into a new "When to use this"
panel below the Example section.
Skill / agent category is now per-item overridable in
marketplace-metadata.json. When absent, the API keeps the parent
plugin's category as the badge so existing items don't lose their
category until curators opt in to per-item categorization.
The new "Example" Q&A panel uses the same Claude Code-style dark
Catppuccin Mocha transcript treatment as the plugin detail —
monospace user row with a green ">" prompt indicator + sans-serif
assistant body with markdown formatting.
All new fields are optional and read on-demand from the working tree.
Skills / agents whose marketplace-metadata.json doesn't carry rich
content render exactly the same way they did before (frontmatter
description + computed slash command + cover from existing v32
enrichment). No DB schema bump.
* Fix TypeError in skill / agent detail when curator sets per-item category
`curated_skill_detail` and `curated_agent_detail` were passing both
`**parent` (from `_curated_inner_parent_fields`, which returns the
parent plugin's category as a fallback) and `**enrichment` (from
`_curated_inner_enrichment`, which returns the per-item category
override when the curator set one) into `InnerDetailResponse(...)`.
Python function-call kwargs unpacking with overlapping keys raises
`TypeError: got multiple values for keyword argument 'category'`
— it doesn't merge like a literal dict does. The bug only surfaced
when the marketplace-metadata.json carried a `category` field at
skill / agent level (curator opting into per-item categorization);
items without that override hit the endpoint cleanly because only
parent provided the key.
Fix: build `merged = {**parent, **enrichment}` first (literal-dict
syntax DOES merge, with the right-hand-side winning) and unpack the
merged dict. Curator override still wins via the merge order, and
the same pattern is future-proof for any other field that lands in
both layers later.
Plus a regression test in test_marketplace_metadata.py asserting
that the inner-resolver carries `category` for downstream merging.
* Marketplace detail: tolerate partial curator JSON
Server constructed UseCase / SampleInteraction via raw dict indexing
(uc["title"], sample["assistant"]), so a curator commit missing any
required Pydantic field crashed the whole plugin / skill / agent detail
endpoint with a 500. Route both constructions through _safe_use_case /
_safe_sample_interaction helpers — partial input silently drops the
malformed card / section instead of breaking the page.
Regression test in test_marketplace_api.py covers the three shapes:
use_case missing a key, use_case with an empty string, and
sample_interaction with only user (no assistant). Sibling rich fields
still render.
* Address PR-251 review (must-fixes + S2/S3 polish) + release-cut 0.50.0
Five must-fixes from the review pass (3 from @cvrysanek's two-stage
review, 2 from my independent pass), plus the 0.50.0 release-cut as the
last commit on this PR per CLAUDE.md (CLAUDE.md "Release-cut belongs
to the PR" rule added in v0.49.1).
Must-fixes
----------
1. Cache eviction: bounded LRU instead of per-marketplace predicate.
The previous predicate (`k[0] == marketplace_id and k[1] != mtime_ns`)
only swept stale entries for the CURRENT marketplace; with N>100
distinct marketplaces each holding one mtime key, the cap silently
failed and memory grew linearly. Replaced with OrderedDict-backed
bounded LRU at cap=256, drop oldest insert on overflow.
Cache stress test pinned in test_marketplace_metadata.py.
2. Render CPU cap: per-field byte cap on description / when_to_use /
sample_interaction.assistant via MARKETPLACE_METADATA_FIELD_MAX_BYTES
(= 64 KiB). Without this, a 1 MiB curator markdown body × QPS =
curator-controlled CPU burn through pure-Python markdown-it-py.
Truncation respects UTF-8 boundaries and logs a warning so the
curator sees the cap fire on the next sync. Test for cap +
UTF-8-boundary preservation.
3. Inner-detail bypassed the metadata cache. _curated_inner_enrichment,
_curated_inner_cover, and curated_detail all called
read_marketplace_metadata directly, defeating the mtime cache the
plugin listing already shared. Routed all three through
_read_metadata_cached so skill/agent detail hits are O(1) re-parses
per marketplace per mtime instead of O(QPS).
4. Truthy-vs-presence trap in plugin/inner enrichment merge. API-layer
writers used `if resolved.get(k):` which silently dropped any
future falsy-but-valid resolver field (bool featured=False, int
priority=0, str category=''). Switched to presence check
(`if k in resolved`) so the resolver is the authority on field
presence; `{**parent, **enrichment}` merge respects whatever the
resolver decided to ship.
5. Vendor-agnostic OSS cleanup. Removed operator-specific token
references (/grpn-eng:, @grpn-eng:, .foundryai/) from
src/marketplace_metadata.py docstring, app/web/templates/
marketplace_item_detail.html JS comment, docs/curated-marketplace-
format.md, and tests/test_marketplace_metadata.py fixtures. Replaced
with generic /my-plugin:tool / @my-agent:role / .example/ placeholders.
CHANGELOG
---------
- New "### Fixed (PR #251 follow-ups)" section documenting all 4
code-side must-fixes
- New "### Internal" section noting the vendor cleanup + new tests
- BREAKING bullet for the file rename now covers operator-side
migration: running instances see plugin enrichment disappear from
the UI until upstream curator renames + nightly sync overwrites the
working tree; POST /api/marketplaces/{id}/sync forces refresh sooner
- Stripped /grpn-eng: leaks from the existing skill/agent rich-content
bullet
Tests
-----
128 targeted tests pass (test_marketplace_metadata, test_marketplace_api,
test_marketplace, test_markdown_render, test_marketplace_synth_strip,
test_marketplace_filter). New tests added:
- 6 XSS regression tests on render_safe (javascript:/data:/vbscript:
schemes via autolink, reference link, and mixed-case + positive
http/https/mailto + noopener noreferrer rel)
- 3 byte-cap tests (truncation + UTF-8 boundary + under-cap pass-through)
- 1 cache eviction stress test (>256 marketplaces -> bounded at cap)
- 1 truthy-vs-presence resolver-contract test
Release-cut
-----------
- pyproject.toml 0.49.1 -> 0.50.0 (minor; BREAKING file rename per
pre-1.0 CHANGELOG note: "breaking changes called out under Changed
or Removed with the BREAKING marker")
- CHANGELOG [Unreleased] -> [0.50.0] - 2026-05-12, new empty
[Unreleased] on top.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* fix(cta): fall back to textarea+execCommand when Clipboard API rejects
The "Setup a new Claude Code" CTA fetches /auth/tokens, parses the JSON
response, renders the setup script, THEN calls
`navigator.clipboard.writeText()`. Modern browsers (Safari, Firefox, and
Chrome on stricter configurations) reject `writeText` with
NotAllowedError when transient user activation has been consumed by an
intervening `await` — which is exactly the case here. Users perceived
this as "the browser blocked the copy" and got the manual-paste fallback
modal even though the textarea + `document.execCommand('copy')` path
WOULD have worked synchronously without needing fresh user activation.
`copyToClipboard` now:
- prefers the modern Clipboard API (unchanged for the happy path)
- on writeText rejection, falls back to `copyViaTextarea` instead of
surfacing the rejection to the caller's catch block.
`copyViaTextarea` is the previously-inline textarea fallback factored
out into a named helper, with two small hardening touches:
- `readonly` + `tabindex=-1` so the hidden textarea doesn't steal
focus or pop the virtual keyboard on mobile.
- explicit `setSelectionRange(0, text.length)` to belt-and-braces the
selection on iOS Safari (where `.select()` alone sometimes selects
zero chars on touch-focused textareas).
Only the CTA button needed this — the Step-1 install-command and the
connector-copy buttons all call `writeText` synchronously inside the
click handler (no awaits in between), so they keep their existing
user-gesture context and didn't hit the same rejection. No template
changes there.
* refactor(home): fold Atlassian MCP registration into connectors block
The standalone "Register the Atlassian MCP server" step (was step 6 in
the unified setup script) moves INTO the Atlassian connector's prompt
body so all Atlassian-related setup lives in one logical group. Same
intent that #247 carried for connectors, applied one level deeper:
the hosted Remote MCP registration is part of "set up Atlassian", not
its own ungrouped step.
What changed:
- `app/web/connector_prompts.py` — the Atlassian prompt's step 5
replaces the speculative "Register the on-demand Atlassian MCP under
.claude/mcp/atlassian" line with the actual hosted Remote MCP
registration: `claude mcp add --transport sse atlassian
https://mcp.atlassian.com/v1/sse || true`. The `|| true` keeps re-runs
idempotent and the body explains the OAuth-on-first-use contract.
Both /home's Atlassian tile and the inlined setup-script Atlassian
sub-block emit this line — single source of truth holds.
- `app/web/setup_instructions.py` — `_mcp_servers_block` deleted; the
`mcp_servers` step is removed from `_step_numbers`; resolve_lines no
longer calls it.
- Renumbering: install (1), init (2), catalog (3), preflight (4),
marketplace (5), diagnose (6), connectors (7), confirm (8). Was:
6 = mcp_servers, 7 = diagnose, 8 = connectors, 9 = confirm.
- `tests/test_setup_instructions.py` — Confirm step 9→8, Connect 8→7,
diagnose 7→6, mcp_servers references dropped.
`test_step_numbering_with_connectors_step` now asserts
`"mcp_servers" not in steps`. Stray-Confirm assertion lists shift
by one position.
- `tests/test_setup_page_unified.py` + `tests/test_web_ui.py` — same
step-number shifts in the rendered /setup preview assertions.
The `claude mcp add` line is still the Atlassian Remote-MCP path that
the 2026-05-10 init-report Fix C added — only its position in the
flow changes. /home Atlassian tile copying continues to install the
MCP too (the prompt body the tile pastes contains the same line).
112 tests pass.
* feat(atlassian): operator-overrideable base URL via AGNES_ATLASSIAN_BASE_URL
Adds an env var / YAML key the operator (Terraform module, customer-VM
template, OSS instance.yaml) can set to bake the Atlassian Cloud site
root into the connector prompt — so end users don't have to guess /
paste their org's `https://<myorg>.atlassian.net`.
When set, the Atlassian connector prompt (rendered on both /home tile
and inlined into the setup-script step 7 Atlassian sub-block) replaces
step 1's "Ask me for my Atlassian Cloud site URL and email" with a
one-line note that the URL is already provisioned by the operator and
asks only for the email. Step 4's helper-script body has the
`BASE_URL='<the site URL I gave you>'` placeholder substituted with
the literal value. When unset (empty), the existing "ask the user"
flow remains — no regression for OSS instances.
Resolution + normalization in `get_atlassian_base_url()`:
- env `AGNES_ATLASSIAN_BASE_URL` > yaml `instance.atlassian.base_url` > ""
- strips trailing slash + trailing `/wiki` so the canonical value is
the bare site root. Matches the per-user helper script's
normalization at storage time (atlassian_prompt step 4 guard 2), so
the literal baked in by the operator stays consistent with what the
user's helper script would have computed from their input.
Plumbing:
- `app/instance_config.py`: new `get_atlassian_base_url()` resolver.
- `app/web/connector_prompts.py`:
- `atlassian_prompt(*, base_url: str = "")` — string-replace two
explicit placeholder phrases when base_url is truthy; otherwise
return the prompt unchanged.
- `all_connector_prompts(..., atlassian_base_url: str = "")` —
forwards the kwarg.
- `app/web/router.py` (`_build_context`): reads
`get_atlassian_base_url()` and passes it through to
`all_connector_prompts(...)` so both the /home tile context AND the
inlined-script `resolve_lines(...)` call use the same value.
- `src/welcome_template.py` (`compute_default_agent_prompt`): same
threading via the existing import-on-demand path.
Tests (`tests/test_home_route_resolution.py`):
- `get_atlassian_base_url` resolver: default empty, env override,
trailing-slash strip, trailing-`/wiki` strip.
- `atlassian_prompt(base_url=...)`: literal URL baked in, ask-step
removed, placeholder replaced, operator-baked-in copy appears.
- `atlassian_prompt(base_url="")`: existing ask-the-user flow
unchanged.
- `all_connector_prompts(atlassian_base_url=...)`: kwarg threads
through to the rendered atlassian prompt.
135 tests pass.
* feat(asana): register hosted Asana Remote MCP in connector prompt
The Asana connector prompt only stored a PAT in the OS keychain + ran
a curl verify against /api/1.0/users/me. That set Claude Code up for
direct `curl` calls but didn't actually wire Asana into Claude's tool
list — so the user couldn't ask Claude to "find my open Asana tasks"
and have it work. Symmetric oversight to the Atlassian connector's
original speculative `.claude/mcp/atlassian` line that this branch
already replaced with `claude mcp add --transport sse atlassian
https://mcp.atlassian.com/v1/sse`.
Adds a new step 5 that registers Asana's hosted Remote MCP:
claude mcp add --transport http asana https://mcp.asana.com/mcp || true
This is the V2 endpoint (streamable HTTP transport, launched February
2026). The V1 SSE endpoint at https://mcp.asana.com/sse was deprecated
2026-05-11 (today) and must NOT be used — calling it out explicitly
in the prompt body so a future operator who finds an old reference
doesn't paste the dead URL. OAuth is handled by Claude Code at first
use, same model as the Atlassian MCP step.
The PAT stored in step 3 stays for direct `curl` calls (precheck +
ad-hoc scripts) — the MCP path uses its own OAuth grant, not the PAT.
Old step 5 (revoke instructions) renumbers to step 6 and adds the
`claude mcp remove asana` cleanup hint.
Same single-source-of-truth invariant holds: /home Asana tile + the
inlined Asana sub-block in the setup script (step 7 connectors) both
emit identical text from `asana_prompt()`.
71 tests pass.
* feat(asana): drive MCP OAuth login + end-to-end validation post-register
`claude mcp add --transport http asana ...` only registers the
server in Claude Code's local config — it does NOT trigger OAuth.
The browser tab opens the first time any `mcp__asana__*` tool gets
invoked. So the previous step 5 left a user looking at a "registered"
MCP that, in practice, hadn't authed yet and would fail on first
real use. Same blind spot Atlassian's prompt also has, but Asana was
the one called out in the latest review pass.
Adds a new step 6 between MCP registration (step 5) and the revoke
instructions (now step 7):
a. Tell the user verbatim what's about to happen — a low-impact
read through the MCP will pop the OAuth browser tab; sign in
with the same account whose PAT they stored in step 3 and
approve. Frames the OAuth as one-time so users don't wait
for it on every later call.
b. Drive an actual MCP read. Don't prescribe the exact tool name
because the Asana MCP's exposed surface (`mcp__asana__*`) is
versioned upstream and we don't want to pin to a name that
gets renamed. Instead: tell Claude to pick the lightest read
from its surfaced tool list (users-me / list-workspaces /
equivalent). Document the recovery path when Claude Code
times out waiting for the OAuth tool use: `claude mcp list`
to confirm registration before retrying.
c. Print a single one-line proof that combines wiring + auth:
"Asana MCP connected as <name> — <N> workspace(s) visible."
Explicit anti-echo callout for tokens, task content, comments.
On failure, surface the exact Claude-Code error and stop —
no silent pass.
d. Sanity-check that the MCP OAuth identity and the PAT identity
reference the same Asana account. Easy mistake to make when
the user has multiple Asana accounts — flag only on mismatch,
keep quiet when they match. Recovery: `claude mcp remove asana
&& claude logout asana` then redo step 5.
Step 7 (revoke) absorbs both the keychain delete + the
`claude logout asana` line so users have a single place to undo
everything.
43 tests pass.
* fix(init): clear stale CA env vars on Windows before any TLS handshake
Reported by the 2026-05-11 Windows test pass: after `agnes init` the
gws connector failed with `UnknownIssuer` TLS errors because
`SSL_CERT_FILE` and `REQUESTS_CA_BUNDLE` were still set in Windows
User scope pointing at `C:\Users\localadmin\.config\agnes\ca-bundle.pem`
— a file that did not exist on the test host. Past Agnes installs
(the setup-prompt trust block + older bootstrap helpers) write those
pointers when they materialize a combined Agnes-CA bundle; when the
bundle file later disappears (re-init on a new VM, machine swap, the
~/.agnes dir wiped), the pointers go stale and every native Windows
TLS handshake fails before Agnes itself runs. SSL_CERT_FILE in
particular REPLACES (not appends to) the trust store, so a stale
pointer is silently catastrophic.
`agnes init` now clears stale pointers in two layers before the first
server roundtrip:
1. Current-process env (os.environ) — what the immediately-following
`api_get` to /api/catalog/tables actually reads. Without this, init
itself blows up before it gets to step 2.
2. Windows User-scope env via PowerShell
`[Environment]::SetEnvironmentVariable(name, $null, 'User')` — what
every future shell + every native tool (gws, claude.exe, pip, uv)
inherits. The 2026-05-11 reporter expected this exact cleanup
("init was supposed to clear these but they persisted").
The cleanup is best-effort and conservative:
- Only deletes a var when its value points at a path that does NOT
exist on disk. Intentional operator config (e.g. SSL_CERT_FILE
pointing at a corp certifi bundle) stays put.
- PowerShell missing / restricted execution policy / WSL-without-pwsh:
swallowed silently. The current-process leg still runs, which
unblocks init even on hosts where the User-scope leg cannot fire.
Tests (`tests/test_init_ca_cleanup.py`, 6 cases):
- Stale pointers → removed from process env.
- Real-path pointers → preserved.
- Non-Windows hosts: PowerShell is not invoked.
- Windows hosts: PowerShell IS invoked with a script that checks
all three vars + uses Test-Path + SetEnvironmentVariable.
- PowerShell FileNotFoundError: cleanup swallows it, does not raise.
- `_is_windows_host()` reflects sys.platform.
* refactor(asana): MCP-first flow — drop PAT storage, precheck via `claude mcp list`
The Asana hosted MCP at https://mcp.asana.com/mcp authenticates via
OAuth (Claude Code holds the grant; browser tab pops on first tool
use). The earlier prompt walked the user through creating + keychain-
storing an Asana Personal Access Token AND registering the MCP — two
parallel auth surfaces for one connector. Once the MCP works, the PAT
has no consumer: the precheck/verify steps that used `curl
$BASE/api/1.0/users/me` are just redundant proof that Asana itself is
reachable, which the OAuth handshake already establishes.
Removed:
- Step 0 keychain probe + curl verify against /users/me with PAT.
- Step 1 open developer-console / create PAT.
- Step 2 click "+ New access token", warn shown-ONCE.
- Step 3 helper-script for keychain-storage (per-OS bodies: macOS
`security add-generic-password`, Linux `secret-tool store`, Windows
`cmdkey /generic`).
- Step 4 PAT-side `users/me` verify.
- Step 5's split that kept the PAT around for direct curl scripts.
- Step 6d's "MCP vs PAT identity sanity check" — there is no PAT
anymore, nothing to mismatch against.
New flow (3 steps total):
- Step 0 precheck: `claude mcp list | grep ^asana` — if found, the
server is registered AND Claude Code is holding its OAuth grant
(otherwise prior failure would have removed it); print
"Asana MCP already registered — skipping setup" and stop. Tells the
user the explicit reset command (`claude mcp remove asana && claude
logout asana`) so a re-register stays one paste.
- Step 1: `claude mcp add --transport http asana
https://mcp.asana.com/mcp` — no `|| true` because step 0 should have
caught the "already exists" case. Step explains the V2-vs-V1
endpoint distinction (V1 SSE deprecated 2026-05-11) and the
abort-clean recovery if the precheck somehow missed the existing
server.
- Step 2: same OAuth + low-impact-read validation pattern as before.
- Step 3: revoke instructions (mcp remove + logout + Asana-side app
revoke at app.asana.com/Settings → Apps).
Both surfaces (the /home Asana tile and the inlined Asana sub-block
in the setup script's step 7) emit the new text from the same
asana_prompt() — single-source-of-truth invariant intact.
77 tests pass.
* Make /home install-hero links readable against blue background
The Claude license-options link added in the previous commit inherited
the default `<a>` style (`var(--hp-primary)` blue), which renders as
blue-on-blue and is unreadable inside the blue install-hero. Add a
scoped `.install-hero a` rule that uses white with an underline
(matching the existing lead-paragraph contrast pattern) so any link
nested in the hero stays legible.
* Reorder /home install flow: auto-mode is now Step 2, Agnes install becomes Step 3
Step 3 (was Step 2) pastes a ~20-command bash bootstrap into a fresh
Claude Code session. Without auto-mode enabled first, each Bash/edit
command needs a manual approve click — bad UX for first-time users.
Move auto-mode from the outside-hero `<details>` reference block into
the install-hero as a real Step 2, between "install Claude Code" and
"install Agnes". Content is the persistent `acceptEdits` snippet
(write to ~/.claude/settings.json) plus a one-liner pointing at
Shift+Tab for users who are already inside a running Claude Code
session. YOLO mode for full Bash auto-approve stays on
/setup-advanced behind the existing link.
The outside-hero `setup-collapsible[data-section="step3"]` block is
dropped — auto-mode is no longer reference content, it's a real
install step, and duplicating it would just diverge over time.
Onboarded users no longer see the auto-mode block at all (consistent
with Steps 1 + 3 also hiding post-onboarding).
Completion banner copy updated: "Step 1, 2 & 3 done — Claude Code
installed, auto-mode set, Agnes ready". Dashboard CTA partial and
other templates don't reference step numbers for this flow, so no
adaptation needed there.
* Simplify /home Step 2 to Shift+Tab only — drop the JSON snippet
Operator pointed out two issues with the prior Step 2:
1. The settings.json snippet is redundant. Claude Code's first
Shift+Tab cycle to auto-accept mode already prompts the user
whether to persist it as default — Claude writes the config
itself, no manual file edit needed.
2. The snippet only showed the POSIX path `~/.claude/settings.json`,
which doesn't translate to native Windows.
Replace the snippet + copy button with a plain Shift+Tab instruction,
explicitly call out the first-time "make this the default?" prompt,
and note that Claude handles the config write itself — same flow on
macOS / Linux / WSL / Windows. Adds a fallback line for users who
already closed the post-OAuth session.
* Tighten /home Step 2 install-note to two paragraphs
Operator: drop the 'Claude writes the setting itself, so this works
the same on macOS / Linux / WSL / Windows...' line plus the
'auto-approves file edits going forward; Bash commands stay gated
— that's the safe default' line. Both were filler — the make-default
prompt already implies persistence, and gated Bash is the obvious
default users won't be surprised by.
Result: paragraph 1 carries Shift+Tab + first-time make-default
say-yes + closed-session fallback in one breath; paragraph 2 keeps
the verbatim YOLO link. Same affordances, less vertical space.
* System plugin tier with mark/unmark fanout (schema v39)
Adds a mandatory plugin tier so admins can pin a small set of curated
plugins into every user's stack from day one. Marking a plugin via the
new toggle on /admin/marketplaces materializes resource_grants for every
group and user_plugin_optouts subscriptions for every user, so the
existing resolver pulls the plugin into every served set without a new
filter layer. Hooks on user-create (Google OAuth, magic-link, admin
POST, scheduler) and group-create propagate the same materialization to
new principals. UI locks: /admin/access disables the checkbox with a
SYSTEM pill; /marketplace cards swap the "In stack" green pill for an
amber "Required" badge with shield icon; the plugin detail install
button reads "Required by your org"; /my-ai-stack toggle is disabled.
Bypass paths return 409 (DELETE /api/admin/grants for system grants,
PUT /api/my-stack/curated/.../{enabled:false}, DELETE
/api/marketplace/curated/.../install). Unmark only flips the flag —
materialized rows persist so admins curate cleanup at their leisure
through the now-unlocked /admin/access checkboxes.
* Marketplace UX polish + drop legacy /store and /my-ai-stack pages
Two-part cleanup post-v39:
(1) Page deletion. /store and /my-ai-stack were already replaced by
/marketplace?tab=flea and /marketplace?tab=my respectively, but the
standalone routes lingered. Hard delete in dev mode — no redirects,
stale bookmarks 404. The /store/new upload wizard, the flea
detail/edit pages, the admin queue, and all /api/store/* +
/api/my-stack endpoints (CLI consumers) stay. Internal hardcoded
hrefs in the upload wizard's Cancel button and the advanced-setup
page repointed to the marketplace tabs.
(2) Detail-page install button rework. The single button that morphed
between "+ Add to my stack" and "✓ In your stack" did not
communicate uninstall affordance. The installed state now renders an
inline white status label *before* a separate red-bordered
"✕ Remove from stack" button on the same row, both at identical
height to avoid layout shift. System plugins keep their locked amber
"✓ Required by your org" pill (no Remove button — API refuses 409).
The post-action hint panel now fires on remove too with the title
flipped to "✓ Removed from your stack" — Claude Code needs the same
/update-agnes-plugins refresh either way.
Also: /admin/marketplaces Details modal "Mark as system" toggle
redesigned. The button was near-invisible (matched neutral row
metadata). It's now a balanced amber-toned chip with shield icon
and a structured confirm modal replacing the native confirm() dialog
that summarizes fanout consequences before commit.
* Move stack-hint inside hero with glass-on-gradient styling
The post-action hint card ("✓ Added to your stack" with the
/update-agnes-plugins recipe) used to live below the hero in
panel-what (gray card on white page body). Clicking add/remove
inserted/removed it between the hero and content, shifting the
panels below — a noticeable scroll jump.
The hint is now anchored inside the hero's top-right corner alongside
the install/remove buttons, both as flex children of an absolutely
positioned .actions container. The card uses a translucent
white-on-glass treatment that adopts the hero's kind color (blue for
plugin, green for skill, purple for agent) without per-kind branching.
Hero is always tall enough (160px photo) to contain the action+hint
stack without overflow, so toggling the hint visibility doesn't grow
the hero or shift body content.
The hero-head grid reserves a third 300px column for the absolute
actions overlay so meta gets the proper 1fr free space instead of
being squeezed by a padding-right hack. Responsive breakpoint at
1100px reflows the actions stack below hero-head when the viewport
isn't wide enough to keep meta + actions side-by-side comfortably.
* Add optional -DataPath bind mount to run-local-dev.ps1
When the operator wants to inspect DuckDB files (system.duckdb, extracts,
marketplaces, store/, …) directly from Windows Explorer, the named volume
inside the Docker Desktop WSL VM isn't reachable. The new -DataPath param
generates a transient compose override that rebinds /data on app, scheduler,
extract (and Caddy's /srv:ro mirror) to a Windows host folder.
Fully additive — when -DataPath is omitted everything behaves exactly as
before: no override file is generated, $composeFiles array is unchanged,
finally cleanup is a no-op. Existing positional invocations
(.\run-local-dev.ps1 up | down | logs) keep binding to $Action because
$DataPath is a named-only parameter with no Position attribute.
The override is written via [System.IO.File]::WriteAllText so the YAML is
BOM-less across PS 5.1 / 7+ — Compose rejects BOM-prefixed YAML on Windows.
The override file is unique per PID and removed in the script's finally
block so concurrent invocations and crashes don't leak files.
* factor mark_system fanout into UserCuratedSubscriptionsRepository
The endpoint imported UserCuratedSubscriptionsRepository, ignored it
(noqa: F841), then duplicated the user-side fanout SQL inline. Adds
fanout_system_for_plugin() symmetric to the existing
fanout_system_for_user() and routes mark_plugin_system through it —
removes the dead import + 14 lines of inline SQL, returns the same
`affected_users` delta count, no behavior change.
* drop customer-specific path from .ps1 example
Per CLAUDE.md vendor-agnostic OSS rule: replaced
C:\\Business\\Groupon\\Agnes\\agnes-data with the generic
C:\\Users\\<you>\\agnes-data placeholder so the docstring
example reads cleanly on any reviewer's box.
* release: 0.48.0 + parallelize Release-workflow pytest
Cuts the release shipped via #228#230#231#232#233#234#236#237#238#239#240 plus this PR (#241). Major changes:
- System plugin tier (schema v39) — admins mark a plugin mandatory; fans
out RBAC grants + subscriptions to every existing user/group plus
hooks for new principals
- BREAKING: removed standalone /store + /my-ai-stack page routes
(replaced by /marketplace?tab=flea + /marketplace?tab=my)
- Setup-prompt + bootstrap recovery fixes (#240)
- DuckDB CHECKPOINT-on-shutdown + 60s compose grace (#235)
- Marketplace + flea-market UX polish, agnes-metadata.json enrichment
Bonus: switch release.yml test step to `-n auto` (matches ci.yml).
Single-threaded was 15-20 min and frequently the bottleneck on PR
mergeability — now ~6 min.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* Setup-prompt + bootstrap fixes from David's 2026-05-10 init report
Three issues from clean-machine bootstrap evidence:
1. `agnes refresh-marketplace --bootstrap` failed to recover when the
local clone existed but Claude Code's marketplace registry had lost
the `agnes` entry. Bootstrap path now parses
`claude plugin marketplace list`, re-runs
`claude plugin marketplace add ~/.agnes/marketplace` when missing,
and treats `add` failures as fatal (was warn-and-continue, root cause
of the cascade into "Marketplace 'agnes' not found" plugin install
errors).
2. Setup prompt now always emits the marketplace-registration block,
even when the operator has zero plugin grants. Pre-wires the
SessionStart hook so future admin grants land automatically without
re-running setup. Block copy adapts: empty list shows
"no plugins granted yet", populated list shows "install plugins".
3. Setup prompt registers the Atlassian Remote MCP server unattended
(`claude mcp add --transport sse atlassian
https://mcp.atlassian.com/v1/sse`). Hosted Remote MCP, OAuth handled
automatically by Claude Code on first use. Asana / GWS stay on the
/home connector cards (PAT/keychain flows don't fit unattended
bootstrap).
Confirm step nudges the user toward the /home connector cards for the
PAT-flow services. CLAUDE.md template renames the marketplace section
to "Agnes Marketplace" and documents that all plugins are addressed as
`<plugin>@agnes` regardless of upstream slug.
Layout: Confirm shifts from step 6/8 to step 9 across all variants
(preflight, marketplace, MCP all unconditional). Tests updated.
* Link Claude license options from /home install pane
Step-1 Claude install on /home pointed users to OAuth without
explaining what to do if they don't have a Pro/Max subscription. Add
a one-line follow-up link to the plan-tier section on /setup-advanced
(new `#claude-plan` anchor) so first-time users discover the
subscription tiers rather than bouncing on the OAuth screen.
* Add idempotent + no-TLS-bypass guardrails to /home connector prompts
The Asana / Google Workspace / Atlassian connector prompts on /home
already shipped a precheck step that short-circuits when the service
is already wired, but they didn't carry the same idempotency +
surface-errors-verbatim + don't-disable-TLS-verification guardrails
the bash bootstrap prompt has. Add a one-paragraph 'Ground rules'
block at the top of each prompt so a connector failure doesn't
tempt the model into bypass workarounds, matching the same posture
David's 2026-05-10 init report flagged for the bash flow.
* skip Source: lines in marketplace registry detector
`claude plugin marketplace list` prints a `Source: <local path>` line
under each registered marketplace; the local clone almost always lives
under a path containing the marketplace name itself
(`~/.agnes/marketplace`). A naive \\bagnes\\b match over the full
stdout therefore false-positives whenever ANY unrelated marketplace
sits under `~/.agnes-…/` or similar. Filter Source: lines out before
matching so the recovery path actually re-adds when needed instead of
silently falling through to a broken `marketplace update agnes`.
Adds regression test covering the substring-only case.
* drop customer-specific tokens from CHANGELOG entries
Per CLAUDE.md vendor-agnostic OSS rule ("nothing customer-specific
... in changelogs"):
- "agnes-vrysanek.groupondev.com" -> "a private-CA Agnes deployment"
- "Groupon Marketplace / groupon-marketplace" -> "<Org> Marketplace /
<org>-marketplace" (placeholder example)
- Removed "David flagged" attribution language; init-report context
stays intact, just stripped of the named host + brand
---------
Co-authored-by: ZdenekSrotyr <zdenek.srotyr@keboola.com>
* feat(store): flea-market entity edit feature with version history (schema v38)
Owner + admin can now edit a store entity from a real Edit page at
/marketplace/flea/{id}/edit, replacing the prior "coming soon"
placeholder. Editable: display name, description, category, video
URL, cover photo, and an optional new bundle. Type is locked (400
type_locked). Display-name change renames the on-disk slug for both
live plugin/ and version dirs (reuses rename-on-archive helper).
Schema v38 (originally drafted as v37; renumbered after rebase onto
main where v37 was taken by the curated marketplace enrichment).
Versioning model:
* Each bundle update bakes into ${DATA_DIR}/store/<id>/versions/v<N+1>/plugin/
and runs the standard guardrails pipeline.
* DEFERRED PROMOTION: live plugin/ + entity.version_no stay at the
prior approved version through the LLM review window so existing
installers keep receiving the previously approved bundle. Live swap
+ version_no/version/file_size bump happen only on LLM approval.
Blocked verdicts leave the prior version serving forever.
* store_entities gains version_no INTEGER + version_history JSON.
Each version_history entry carries hash, sha256, size, submission_id,
created_at, created_by.
* Existing entities backfill to v1 with a single-entry history seeded
from the row's current `version` hash. Initial create also seeds
versions/v1/plugin/ so future restore can copy v1 bytes forward.
Concurrency:
* Block-while-pending: an in-flight LLM review blocks any further edit
with 409 prior_version_pending. Owner waits 5-30s; Edit button on
detail page renders disabled in the same window via the new
edit_in_flight flag (decoupled from quarantine_sub since the
deferred-promotion flow keeps visibility='approved').
Rollback:
* New endpoint POST /api/store/entities/{id}/versions/{n}/restore
(owner + admin). Copies vN bundle forward as v<max+1> and re-runs
guardrails (rules tighten over time; pre-approved bundles re-validate).
Forward-only history. Same deferred-promotion semantics — live stays
at prior version until LLM approves the restored copy.
UI:
* New /marketplace/flea/{id}/edit page (owner + admin gated).
* Versions card on plugin + item detail templates (owner/admin only)
via shared _flea_versions.html partial.
* Admin queue gains v# column with current badge + separate Hash
column. Submission detail surfaces Version + Bundle hash rows.
* Activity timeline split into per-submission + entity-wide cards;
entity-wide rows render vN chips when audit row params reference
a specific version.
* Section headers (Manifest / Static / Quality / LLM review) tag
with vN chip via shared macro.
* Reviewed-by-model field surfaces explanatory text per status.
* Banner upload-failure now redirects to detail page on
submission_blocked instead of staying stuck.
Tests: 24 in tests/test_store_entity_versions.py covering metadata-
only edit, bundle-edit version bump, type lock, block-while-pending,
name change disk rename, restore flow + 404/400/403 paths, edit page
404 for non-owner, versions card visibility gating, admin queue v#
column, admin detail Version/Hash rows, deferred-promotion installer
contract (pending review doesn't break installer / blocked verdict
keeps prior / approved promotes), admin can edit/restore non-owned,
restore deferred promotion, audit log per-version params. 214 tests
green across guardrails + edit + admin + repo + schema suites.
* docs(store): refresh update_entity docstring to match deferred-promotion + submission-status gate
Bring the docstring in sync with the actual fixes from the prior
commit. The pre-fix wording said the gate read
visibility_status='pending' AND submission status — under deferred
promotion that would never fire for v2+ edits. Now describes:
- Block-while-pending gates on submission.status DIRECTLY,
independent of visibility (so v2+ deferred-promotion edits don't
slip through).
- Display-name + bundle change defers the live rename to promotion;
metadata-only renames stay immediate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* Move marketplace plugin updates from hook to /update-agnes-plugins skill
The SessionStart hook used to run `agnes refresh-marketplace --quiet`,
which performed a full fetch+reset+install cycle on every Claude Code
session start. That work was invisible to the user, slowed session
startup, and was unrecoverable interactively when something failed.
Split the responsibility:
- `agnes refresh-marketplace --check` is a new lightweight detector:
`git fetch` only, compares local HEAD with remote FETCH_HEAD, emits
a Claude Code hook JSON message pointing the user at
`/update-agnes-plugins` when the marketplace has changes. No reset,
no plugin install/update side effects.
- `/update-agnes-plugins` is a new slash command (installed by
`agnes init` into `<workspace>/.claude/commands/`) that runs
`agnes refresh-marketplace` (default chatty path). Output streams
into the Claude Code transcript so the user sees install/update
progress and can react to errors interactively.
- The SessionStart hook now runs `--check`. Existing workspaces
auto-upgrade on next `agnes init` (substring marker matches both
the old `--quiet` entry and the new `--check` one).
BREAKING: `agnes refresh-marketplace --quiet` is removed. Old hooks
calling it silent-noop after the CLI upgrade (the hook's `|| true`
swallows the unknown-flag error) until re-init rewrites them.
* Point marketplace 'Added to your stack' hint at /update-agnes-plugins
The post-install green panel on plugin and skill/agent detail pages
referenced the SessionStart auto-install path and a shell-prompt
`agnes refresh-marketplace` invocation. With the hook now being
detect-only, that copy was misleading — the actual install path is
the new slash command.
Condensed to a single instruction: "Open a new Claude Code session
and run:" followed by `/update-agnes-plugins` in a copy-chip.
JS clipboard string updated to match.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
* Curated marketplace enrichment via agnes-metadata.json + curator metadata
Adds a second well-defined metadata file `.claude-plugin/agnes-metadata.json`
that upstream marketplace repos can opt into, providing per-plugin (and
per-skill / per-agent) cover photo, demo video URL, doc links, and
category override. The Claude Code marketplace contract is untouched —
agnes-metadata.json + the convention `.agnes/` directory are stripped
from the synthetic Claude Code marketplace served via /marketplace.zip
and /marketplace.git/*, so user instances see a clean Claude Code repo
with no Agnes-only metadata.
Highlights:
- DB schema v32 — adds curator_name + curator_email on marketplace_registry,
cover_photo_url + video_url + doc_links on marketplace_plugins.
- Mandatory curator at marketplace registration, editable later through
the admin UI; surfaces on cards + detail pages in place of owner_todo.
- External-asset mirror cache at ${DATA_DIR}/marketplace-cache/<slug>/
with conditional GET, 60s timeout, 10 MB body cap, SSRF guards, and
Wikipedia-policy-compliant User-Agent.
- Strict drop semantics — anything Agnes can't deliver as a real PDF /
Markdown / plain text doc, or a real PNG / JPEG / WebP cover, is
dropped from the served metadata; UI looks identical to no-entry case
(gradient placeholder for missing covers, no row in the doc list).
- Doc allowlist + image allowlist enforced on both the curated mirror
flow and the Flea upload flow (/store/new); shared module
src/marketplace_assets.py.
- New /api/marketplace/curated/{mp}/{plugin}/{asset,doc,mirrored}/...
endpoints with path-traversal guards + RBAC + Content-Disposition
attachment for docs.
- Curator-focused format guide at /marketplace/format-guide; canonical
source is docs/curated-marketplace-format.md, also linked from the
admin /admin/marketplaces page next to + Add Marketplace.
See CHANGELOG.md under [Unreleased] for the full breakdown.
* Fix format-guide test assertion to match shortened disclaimer
The 'Flea Market' phrase was trimmed out of the disclaimer in
docs/curated-marketplace-format.md after the curator-focused rewrite.
Update the rendered-HTML test to assert the channel-scoping phrase
that's actually present ('Curated Marketplace channel only') rather
than the 'Flea Market' contrast that's no longer in the doc.
* Drop unused 'version' field from agnes-metadata.json schema
The parser never read it; it was a YAGNI placeholder for future
schema evolution. Curators don't need to wonder what to put there
when adding the file for the first time. Will be re-added if and
when we actually introduce a backwards-incompatible schema change.
* Harden asset mirror against SSRF via redirect + DNS rebinding
The pre-flight _is_safe_url check validated only the initial URL;
urllib.request.urlopen then followed redirects and re-resolved DNS for
the actual connection — both bypassable. Attacker-controlled origin
could 302 to http://169.254.169.254/... and exfil cloud metadata;
attacker-controlled DNS could return public IP first / 127.0.0.1 second.
Replace urlopen call with a shared OpenerDirector wired through three
custom handlers: _SafeRedirectHandler re-runs SSRF allowlist on every
redirect Location (max 5 hops, down from urllib's 10), and
_PinnedHTTPHandler / _PinnedHTTPSHandler connect to the IP that passed
validation rather than re-resolving the hostname. TLS SNI + cert verify
stay bound to the original hostname.
_resolve_safe returns the validated IP (the existing _is_safe_url
2-tuple wrapper stays for backwards compatibility) and rejects round-
robin DNS that mixes a public + private record. _UnsafeRedirectError
is a typed exception so _fetch_url can map redirect blocks to terminal
'rejected' status (not transient 'failed'). _http_open is the single
call site so tests can mock at one well-defined seam.
Tests cover redirect blocking (link-local, loopback), redirect-error
unwrapping inside URLError, pinned-IP connection target, and the
end-to-end DNS-rebinding scenario. Existing tests that mocked
urllib.request.urlopen are migrated to mock _http_open.
* Harden /asset/ endpoint against stored XSS
The endpoint served any file in the cloned marketplace repo with
stdlib-detected Content-Type, so a curator who landed evil.html (or a
renamed evil.png carrying HTML bytes) in the working tree got a
same-origin XSS — the response shares cookie scope with /admin and
/api/me/*.
The asset endpoint is image-only by contract (cover photos referenced
from agnes-metadata.json + inner skill / agent cards), so applying the
same allowlist + magic-bytes pattern that /doc/ already uses closes
the gap without breaking any legitimate use case. Three layered
checks: extension in IMAGE_EXTENSIONS (.png/.jpg/.jpeg/.webp; SVG
excluded — <script> inside SVG executes), validate_image_file magic
bytes (defeats rename-extension attack), Content-Type pinned from the
validated extension (never stdlib mimetypes).
Defense-in-depth: X-Content-Type-Options: nosniff stops browser MIME
sniffing; Content-Security-Policy: default-src 'none' blocks script /
iframe execution even if a future regression let HTML through.
Tests cover the .html extension reject, the renamed-HTML-as-PNG magic-
bytes reject, the .svg reject, and the happy-path PNG with security
headers attached. The pre-existing path-traversal test seeds a real
PNG instead of ok.txt now that the endpoint is image-only.
* Enforce mandatory curator on marketplace PATCH
The POST handler enforced curator_name + curator_email at create time,
but PATCH treated empty / missing curator inputs as 'no change'. Legacy
rows that pre-date v32 (curator_name=NULL) could be edited indefinitely
without ever filling the curator gap, and OWNER_TODO_PLACEHOLDER lingered
on every /marketplace card.
Reject the PATCH with 400 when the post-merge row would persist with
empty curator. The check fires after the existing field-merge logic, so
once-filled rows that don't touch curator still pass through (their
existing values fall through from the DB row). DB column stays nullable
so untouched legacy rows continue to coexist — the gate fires only the
moment an admin opens the edit modal.
Existing PATCH semantics preserved: empty-string input still means 'leave
existing value alone', and once-filled curator can't be cleared (those
test cases pass unchanged). New test seeds a legacy row directly via the
repository, then exercises url-only PATCH (rejected), partial-fill PATCH
(rejected), and full-fill PATCH (succeeds); a follow-up no-curator PATCH
on the now-formed row also passes.
* Drop unused curated-marketplace helpers (PR #234 review)
* build_db_payload — imported by src/marketplace.py but never called.
The strict-drop semantics it would have implemented were re-written
inline in _refresh_plugin_cache (see the comment block there). The
standalone helper still carried the old fall-back-to-original-external-
URL-on-mirror-failure behaviour, which contradicts the documented
drop-when-can't-deliver contract — a future contributor who re-wired
it would have introduced a silent regression. Delete with the helper
+ the import + the comment that referenced it.
* _resolve_marketplace_name — one-line shim with no remaining call
sites. Callers use _resolve_marketplace_meta which returns name +
curator together, avoiding the double DB hit the shim exists to
hide.
* '# noqa: F401 Optional kept for forward-compat' was wrong — Optional
IS used in src/marketplace.py (line 70 and line 238). Drop the noqa
comment so a future ruff run doesn't try to remove a real import.
Removing build_db_payload also drops the only remaining use of Optional
in src/marketplace_metadata.py, so the import comes out there too.
* Cap agnes-metadata.json size + catch RecursionError on parse
The reader is invoked once per marketplace per sync and the file is
curator-controlled. Two failure modes were unguarded:
* Multi-GB JSON: path.read_text() pulled the whole file into memory
before json.loads even ran. A curator with commit access to an
upstream repo could OOM the sync worker.
* Deeply-nested JSON under any size cap: cpython's recursive object /
array parser raises RecursionError at ~1000 levels of depth.
RecursionError is a RuntimeError, not ValueError, so the existing
catch let it propagate up and abort the entire sync — every other
marketplace in the same pass got skipped.
Add AGNES_METADATA_MAX_BYTES = 1 MiB (a real metadata file with covers,
docs, categories for ~50 plugins fits in <100 KB so the cap is
generous) and gate the size check on path.stat().st_size before the
body read. Broaden the parse except to (ValueError, RecursionError)
with a unified log line. Both failure modes degrade to the same
empty-dict fall-back the malformed-JSON path already used, so one bad
upstream never aborts the rest of the sync.
Tests cover the size cap firing before json.loads (whitespace-padded
valid JSON exceeding the cap) and the recursion path (5000 nested
arrays — past cpython's default recursion limit but well under the
size cap).
* Persist asset-mirror manifest per body write, before unlink
sync_assets wrote each body atomically (tmp + rename) but persisted
the manifest only at the end of the batch. A kill -9 mid-Phase 2 left
on-disk files the manifest never referenced. Once a curator dropped
that URL from agnes-metadata.json, Phase 3's cleanup had no record of
the file and the orphan stayed forever — there's no GC pass walking
the cache dir today, so disk would slowly bloat.
Phase 2 (body-write iteration): after the in-memory manifest mutation,
persist BEFORE unlinking the previous body. The crash window narrows
from 'all of Phase 2' to 'between persist and unlink' (microseconds).
A persist failure mid-batch keeps the previous body on disk — the on-
disk manifest still references it, and a stale-but-existing file beats
a 404. Cost: one extra tmp+rename per body write; manifest is a few KB
so the overhead is negligible vs. the HTTP fetches.
Phase 3 (curator-removed URLs): same discipline. Collect the to-delete
relpaths, persist the manifest with the entries already gone, THEN
unlink. A crash mid-cleanup leaves at most a microsecond window where
files exist despite the manifest no longer naming them. The next sync
reads the (correct) manifest and the orphan stays orphaned, but the
served state is consistent.
Tests cover per-body persist call count, the post-update on-disk
manifest content, and Phase 3 ordering verified by reading the on-disk
manifest from inside Path.unlink.
* Consolidate marketplace video embeds + format-guide CSS
The YouTube nocookie / Vimeo / <video> / link-fallback detection logic
was duplicated verbatim in marketplace_plugin_detail.html and
marketplace_item_detail.html (~40 JS lines each, with subtly-different
inline styles). Both templates now {% include %} a single
_marketplace_video_embed.html partial inside their IIFE so the regex,
the nocookie attribute set, and the unknown-host link fallback live in
ONE place — future tweaks (new host, new attribute, fixed sandbox flag)
no longer need to be applied twice in lockstep.
The .video-wrap selectors (one inline <style> rule in plugin_detail,
one inline style='...' attribute in item_detail) are replaced by the
existing .video-embed 16:9 wrapper in style-custom.css, with new
.video-embed video / .video-embed a child rules added so the wrapper
handles all four embed shapes uniformly without per-template
positioning.
The 60-line inline <style> block in marketplace_format_guide.html
moves verbatim to style-custom.css under a new 'Marketplace format
guide page' section, scoped to .format-guide so other pages aren't
affected.
No user-visible behaviour change: the rendered HTML for valid
YouTube / Vimeo / mp4 / external links is byte-identical to before,
and the format-guide page renders the same.
* Maintainability cleanup batch (PR #234 review)
#10: drop _path_under from app/api/marketplace.py — it was a byte-
equivalent clone of _safe_join (same Path.resolve(strict=True) +
relative_to() containment check). The three v32 endpoint handlers
(/asset, /doc, /mirrored) now share the existing helper.
#14: rename src/marketplace_assets.py → src/marketplace_asset_validation.py
so the file's purpose is obvious from the name and the previous
overlap with src/marketplace_asset_mirror.py is gone. Six call-site
imports updated in lockstep; CHANGELOG references under [Unreleased]
updated to track the new path.
#11: consolidate the URL builders that resolve
/api/marketplace/curated/<slug>/<plugin>/{asset,doc,mirrored}/...
paths. _internal_asset_url / _internal_doc_url / _mirrored_asset_url
lived in src/marketplace.py, while a copy named _mirrored_url lived
in app/api/marketplace.py with a 'must stay aligned' comment. New
module src/marketplace_urls.py is the single source of truth — both
call sites import from it and a future URL-format tweak only needs
to change one file. The _ROUTE_PREFIX constant collapses the per-
function f-string repetition. The route-handler endpoints themselves
still own the path string literals (keeping the builders identical
to the route declarations remains a checklist item, not a runtime
guarantee).
* Re-key asset-mirror manifest by (plugin, url) + dedup HTTP fetches
The manifest used to be keyed by URL alone, so two plugins in the
same marketplace referencing the same external image (a shared CDN
icon, a common cover) collided on entry.plugin_name — last writer
won. The DB row for the losing plugin then stored a served URL
pointing under the winning plugin's tree, and require_resource_access
denied legitimate access on one side and let the other plugin's user
reach the wrong asset.
In-memory: Dict[Tuple[str, str], MirrorEntry] keyed (plugin_name, url).
On disk: format flips from {url: entry} dict to [entry, ...] list of
self-describing entries (each carries plugin_name + url + the
previous fields). JSON keys can't be tuples; encoding 'plugin::url'
would just shift the parsing burden.
Phase 1 of sync_assets deduplicates fetches by URL — three plugins
sharing one URL share one HTTP request. The conditional-GET prior is
picked from any owning plugin's prior entry; if their etags diverge
(rare) we miss one 304 and pay for a full re-download instead.
Phase 2 still creates a per-(plugin, url) manifest entry pointing
under the plugin's own subdir, and Phase 3 cleanup is keyed the same
way so dropping a URL from one plugin's metadata doesn't disturb
another plugin still referencing it.
Body files stay per plugin (RBAC-clean isolation: deleting plugin A's
cache can't strand plugin B). Bandwidth saved by fetch dedup.
Consumer code re-keyed: src.marketplace._refresh_plugin_cache rebuilt
served_url_for / mirror_status as composite-keyed maps;
app.api.marketplace._resolve_external_via_mirror /
_curated_inner_cover / _curated_inner_enrichment look up by
(plugin_name, url).
Tests cover per-plugin manifest entries with shared URL, the single
HTTP fetch for N plugins, and Phase 3 drop-one-keep-other. All
existing tests migrated to composite key access; v2 list format
assertions verify on-disk shape.
* Migrate asset mirror from urllib.request to httpx
The asset mirror was the only HTTP call site in Agnes still using
urllib.request; every other module (CLI, Jira / OpenMetadata / OpenAI
connectors, scheduler, Telegram bot) already used httpx. The asset
mirror was added in this PR's base commit, so this is the only chance
to bring it into convention before someone copies it as 'the pattern
for HTTP fetches in Agnes'.
Three concrete benefits beyond consistency:
* SSRF defence collapses from five urllib classes
(_PinnedHTTPConnection, _PinnedHTTPSConnection, _PinnedHTTPHandler,
_PinnedHTTPSHandler, _SafeRedirectHandler) into one
_SSRFGuardTransport. httpx invokes handle_request() on every redirect
hop, so re-validation is free — we don't need a custom redirect
handler at all.
* DNS-rebinding defence: the transport rewrites request.url.host to the
SSRF-validated IP before delegating to super().handle_request().
httpcore connects to whatever URL.host says, so this pins the
connection without subclassing HTTPSConnection. The original hostname
goes into the Host header + the sni_hostname extension so TLS / vhost
routing still bind to the curator-supplied hostname.
* Error handling: one httpx.HTTPError catch-all for transport errors,
plus specific httpx.TimeoutException / httpx.TooManyRedirects branches
for clearer diagnostics. Matches the _translate_transport_error shape
in cli/client.py.
The shared httpx.Client is built lazily at module load (same pattern as
cli/client.py:_get_shared_client) with follow_redirects=True,
max_redirects=5, timeout=HTTP_TIMEOUT_SEC, and our custom transport.
Externally observable behaviour is unchanged: same FetchOutcome
statuses, same manifest format, same conditional GET semantics, same
body-size cap.
Tests migrated from urllib-shaped fakes to httpx-shaped (status_code,
iter_bytes, context manager). Five urllib-specific tests replaced with
httpx equivalents — three transport unit tests + one DNS-rebinding
integration test that verifies host rewrite via monkey-patched
super().handle_request. One test deleted without replacement
(unwrap-URLError-wrapping-an-_UnsafeRedirectError — urllib-specific,
not applicable to httpx).
* Surface curated agnes-metadata enrichment on My Stack tab
GET /api/marketplace/items?tab=my built each curated row from the
on-disk marketplace.json by way of resolve_allowed_plugins, which
doesn't carry the agnes-metadata enrichment columns
(cover_photo_url, video_url, category override, doc_links). The
handler then hard-coded cover_photo_url=None on the synthetic row.
Result: once a user clicked '+ Add to my stack' on a curated card,
the same plugin in tab=my rendered with the gradient placeholder
instead of its cover photo — confusing parity break vs. the curated
tab where the same row goes through MarketplacePluginsRepository
and gets the enriched columns.
Pre-load the enriched marketplace_plugins rows for every marketplace
the user is subscribed to, then look each granted+subscribed plugin
up by (marketplace_id, plugin_name). Fall back to the on-disk
synthetic shape only when the DB row is missing — happens during
the rare race where RBAC is granted before the first sync cycle
ingests the plugin. RBAC gating (granted set from
resolve_allowed_plugins) is unchanged so this fix can't widen
visibility; it just upgrades the data shape behind cards the user
was already going to see.
Per-marketplace list_for_marketplace beats N gets — typical user is
subscribed to <5 marketplaces, so this is at most a handful of
queries vs. one per subscribed plugin.
Regression test seeds a plugin with cover_photo_url + category
override, subscribes the user, hits /api/marketplace/items?tab=my,
and asserts photo_url + category come through. The misleading
'fall through to gradient until the user re-visits the curated tab'
comment is gone.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
* feat(store): flea-market upload guardrails + soft delete + JOIN-based admin queue
Adds an end-to-end guardrails pipeline for store uploads (manifest +
static-security + LLM review), persists blocked bundles for forensics,
introduces soft-delete (Archive) semantics, consolidates the legacy
/store/{id} surface into /marketplace/flea/{id}, and reworks the admin
queue so lifecycle filters read live entity visibility via LEFT JOIN
rather than a denormalized submission column.
Schema v29 → v35:
* v29 store_submissions table + store_entities.visibility_status
* v30 file_size, bundle_sha256, bundle_purged_at on submissions
* v31 reshape store_submissions (drop legacy unique on entity_id)
* v32 store_entities.archived_at/by + 'archived' visibility value
* v33 drop store_submissions.retry_count (unused)
* v34 ensure idx_store_submissions_entity exists post column-drop
* v35 broaden visibility_status enum + JOIN architecture cutover
Pipeline (src/store_guardrails/):
* Inline checks: manifest_check, static_scan, quality_check
* LLM review configurable haiku|sonnet|opus (default haiku)
* BackgroundTasks-driven async path with structured-output JSON
* Per-submitter daily quota (default 50)
* 30-day TTL purge job (POST /api/admin/run-blocked-purge)
* Bundle SHA256 + size persisted; sha256 survives purge for forensics
Visibility model:
* pending | approved | hidden | archived
* _enforce_visibility returns 404 (no leak) for non-owner non-admin
* Owner sees own non-approved entries via include_owner_id widening
* Install refused with 409 entity_not_approved when not approved
Soft-delete (DELETE /api/store/entities/{id}):
* Default = soft (visibility_status='archived'); existing installs
keep getting served the bundle so users don't lose the plugin
* ?hard=true admin-only: drops bundle + cascades user_store_installs
* Hard-delete preserves entity_id on submission as tombstone so
audit_log linkage survives for the activity timeline
Admin queue lifecycle (the JOIN refactor):
* Verdict (store_submissions.status) is immutable forensic record
* Lifecycle (store_entities.visibility_status) is live state
* /admin/store/submissions Archived chip translates to
`e.visibility_status='archived'` via LEFT JOIN — any path that
flips visibility surfaces in the queue immediately
* Detail page renders Status (verdict) and Entity lifecycle side by
side so admins see "approved at review, now archived" at a glance
URL consolidation:
* /store/{id} deleted (no redirect, stale bookmarks 404)
* /marketplace/flea/{id} is the canonical detail surface
* Three in-tree callers (upload-success, my-stack card, store
listing card) updated to point at the new URL
* Quarantine banner extracted to _quarantine_banner.html partial,
self-guarded, included from both flea detail templates
* Banner JS auto-refreshes when the verdict lands by polling
/api/marketplace/flea/{id}/detail (visibility_status +
submission_status — the latter is needed because blocked_llm
keeps the entity at visibility_status='pending')
Audit log resource format:
* runner.py emits prefixed `store_submission:{id}` (post-fix)
* Detail-page timeline query handles three patterns: prefixed
submission, helper-emitted `store_entity:{sub_id}`, and bare-id
legacy rows — all surface in the activity timeline
UX fixes:
* Owner sees Under review / Quarantined / Hidden banner with status
* Install button gray-disabled (not blue) when non-approved
* Owner cannot delete quarantined entries (403); admin can
* Admin queue: filter chips, sortable columns, paging, page-size
* Auto-refresh queue every 5s while pending rows are visible
* Store upload page file picker no longer opens twice (label →
input default action collided with explicit JS handler)
Tests: 168 passed across the guardrails suites (admin submissions,
store API, inline / LLM / purge guardrails, store repositories,
marketplace filter, schema version). New regression coverage
includes: archive surfaces via JOIN even when API path is bypassed;
deleted submission renders activity timeline (tombstone); flea
detail surfaces submission_status only for owner/admin; detail page
renders Entity lifecycle row; audit log resource format covers both
helper and runner paths.
* fix(store-guardrails): PR #233 follow-up — prompt injection, atomic PUT, BG race, schema, reaper, sort whitelist
Addresses 9 of the 23 findings from the PR #233 review (spec at
docs/superpowers/specs/2026-05-09-pr233-guardrails-fixes-spec.md).
Merge-gate items #1-#6 plus high-value mediums #7, #9-#12, #23.
Architectural items (#8 enum split, #14 factory) and pure
maintainability (#15-#22) deferred to follow-ups.
Security:
* #1 prompt injection — SYSTEM_PROMPT now passed via the SDK's
dedicated system= parameter; bundle wrapped in <bundle>...</bundle>
sentinels declared data-only by the system prompt; literal
sentinel strings in user content are escaped so an adversarial
README can't forge a close tag.
* #6 static scan honesty — module docstring + admin copy + docs
declare static scan as signal not gate; .md/.txt/.rst/.html/.json/
.yaml/.yml/.toml skipped to avoid false positives on prose.
AST mode for Python deferred (separate flag, FP comparison work).
Correctness:
* #2 PUT atomicity — bundles bake into plugin.staging-<rand>/
alongside live, atomic-rename on success; failed checks leave
live tree byte-for-byte intact.
* #3 BG-task race — set_visibility_if_pending guards verdict flips
to the (pending, hidden) review window; admin archives during
review survive; skipped flips audit-logged.
* #4 v35 NOT NULL/DEFAULT — schema v35→v36 re-applies them on
store_entities.visibility_status. CHECK constraint enforced
application-side (DuckDB ADD CHECK on existing column unsupported).
* #7 stuck-review reaper — reap_stuck_llm_reviews flips pending_llm
rows older than guardrails.stuck_review_grace_seconds (default
1800) to review_error. Scheduler runs every 15 min via new
/api/admin/run-reap-stuck-reviews. Set knob to 0 to disable.
* #9 quota counter — count_blocked_for_submitter_since now counts
blocked_inline + blocked_llm + review_error so a submitter
triggering only LLM-blocked verdicts is bounded.
* #10 missing risk_level — surfaces as review_error with
error='missing_risk_level' instead of silently defaulting to
'medium' (which looked like a model-decided block).
* #11 archived_at clear — set_visibility nulls archived_at +
archived_by when transitioning out of 'archived' so a future
read doesn't show stale archive forensics on an approved row.
Maintainability:
* #12 FSM doc comment — accurate insert/transition/lifecycle
description in src/db.py near store_submissions schema.
* #23 sort-key whitelist — admin queue rejects unknown sort keys
with 400 invalid_sort_key; substring-replace footgun removed.
Deferred (separate PRs):
* #5 quota race — proper fix requires asyncio.Lock spanning the
full pipeline; threading.Lock blocks event loop, DuckDB MVCC
doesn't help. API-level slowapi bounds worst case for now.
* #6 part 3 (AST static scan), #8 (enum split), #13 (import
bundle docs), #14 (factory consolidation), #15-#22 (maint).
Tests:
* New: tests/test_store_guardrails_prompt_injection.py (corpus +
trust-boundary invariants), tests/test_store_put_atomic.py,
tests/test_store_guardrails_reaper.py.
* Extended: test_store_guardrails_llm.py (system param, missing
risk_level, BG race), test_admin_store_submissions.py (quota
counter widening, sort whitelist 400), test_store_repositories.py
(un-archive metadata clear), test_db_schema_version.py (v36).
* Full suite: 3738 passed; 17 pre-existing baseline failures
unchanged (db migration tests, cli binary rename, catalog export,
user mgmt v5 backfill — confirmed by stash + rerun on clean tree).
* Extract session pipeline framework, refactor verification, add UsageProcessor skeleton
Pluggable framework under services/session_pipeline/ (contract + lib + per-processor
runner) so multiple processors can read /data/user_sessions/<key>/*.jsonl on their
own cadence with full failure isolation. Verification flow becomes the first plugin;
a no-op UsageProcessor reserves the second slot pending a separate brainstorm on
extraction logic + storage shape.
Schema v28→v29: rename session_extraction_state → session_processor_state with
composite PK (processor_name, session_file). Existing rows copied over with
processor_name='verification'; legacy table dropped. Migration is idempotent and
no-ops the copy step on fresh installs that came up at the new schema.
Endpoint: /api/admin/run-verification-detector replaced by parametrized
/api/admin/run-session-processor?processor=<name>. Audit action format follows.
Scheduler JOBS: verification-detector entry split into session-processor:verification
+ session-processor:usage. SCHEDULER_VERIFICATION_DETECTOR_INTERVAL retained for
operator compatibility (drives both cadence and health-check grace window);
SCHEDULER_USAGE_PROCESSOR_INTERVAL added.
* Address PR #232 review: scan dead branch + per-processor lock
- `SessionProcessorStateRepository.scan_unprocessed_for` dead else: both
branches surfaced every jsonl, the SELECT was unused, runner MD5-rehashed
every stable session per tick. Replaced with an mtime precheck — stable
sessions (mtime <= processed_at) are filtered at scan; modified files
still surface for the runner's authoritative `file_hash` invalidation.
Naive-local comparison matches the existing health-check idiom (DuckDB
TIMESTAMP strips tz on storage).
- Per-processor advisory lock around `_run_processor` in
`/api/admin/run-session-processor`. Scheduler tick + manual admin POST
could otherwise both run, both call create_evidence on overlapping
detections, and accumulate duplicate verification_evidence rows (the
dedup short-circuit only covers create+contradiction, not evidence per
ADR Decision 3). Non-blocking acquire → 409 Conflict on concurrent
invocation; release in finally so a runner exception doesn't wedge the
processor.
Tests: two new scan unit tests (mtime filter + post-mark mtime bump), 409
endpoint test, lock-released-on-exception test. Two existing tests updated
for the new "filtered at scan" stat shape (previously asserted skipped == 1,
now scanned == 0).
* Address PR #232 review #2: parallel scheduler tick + last_run on terminal state
Two pre-existing scaffold bugs in services/scheduler/__main__.py amplified
by adding more session-pipeline jobs:
1. Serial for-loop over jobs with synchronous httpx.post(timeout=900) — a
10-minute verification run blocked every other job (data-refresh,
health-check, usage, corporate-memory) for the whole window. The PR's
stated isolation guarantee held inside the runner but broke at the
scheduler dispatch layer.
2. last_run advanced only when _call_api returned True. Permanent-failure
jobs hot-looped on every tick (30s) instead of cadence (15min).
Fix: ThreadPoolExecutor.submit per due job + per-job in_flight set so a
long-running job can't be re-launched on subsequent ticks. last_run
advances unconditionally in finally; errors still surface via _call_api
logging + audit_log on the receiving side.
_run_job extracted to module-level for unit testing. New tests:
- TestRunJobBookkeeping: advances on success / failure / unhandled raise
- TestRunLoopParallelism: in_flight protection prevents duplicate
launches across ticks for a single slow job
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
* feat(home+news): state-aware /home + /news + admin-edited news section
Squash of the vr/home-page feature work for clean rebase onto main.
Original 18-commit history preserved in branch backup/vr-home-page-pre-rebase.
What's in this PR:
**State-aware /home page**
- New `/home` route with hero + auto-mode + connectors (Asana / GWS /
Atlassian) + lookarounds. Onboarded vs not-onboarded state-machine
branches a single template (`home_not_onboarded.html`); the install
steps, "Setup a new Claude Code" CTA (90-day PAT mint), and per-
connector setup prompts hide once `users.onboarded=TRUE`. A
completion badge replaces them.
- "Mark me as offboarded" button reverses the flag without an SQL UPDATE.
- `users.onboarded BOOLEAN` column added; default FALSE; flipped by the
CLI's `agnes init` post-success POST and the `/admin/users` API.
- Connector setup prompts pre-check whether the tool is already
installed/connected before re-running setup.
- GWS scope set widened to include Google Chat (`chat.spaces`,
`chat.messages`).
**Single template + design tokens**
- `dashboard.html` now extends `base.html` via the new
`{% block layout %}` opt-out (full-width pages skip the 800px
`.container`). Net: every page shares one shell.
- `style-custom.css` `:root` extended with `--space-{7,9,10,12}`,
`--radius-2xl`, `--shadow-{card,elevated}`, `--text-{muted,disabled}`,
`--focus-ring`, `--transition-*`, `--width-{narrow,app,wide}` so
inline page styles can migrate incrementally.
**Auth redirects honor AGNES_HOME_ROUTE**
- `safe_next_path` resolves the configured home route when no `default=`
is passed; OAuth callbacks, magic-link clicks, password form, and
LOCAL_DEV_MODE shortcuts now land on `/home` (or whatever the operator
picked) instead of always /dashboard.
**News section + /news permalink + /admin/news editor**
- Schema-bumped `news_template` table (single versioned entity, draft +
publish gate). `published BOOLEAN` distinguishes draft from public;
monotonically-increasing `version` per save; rows >30d pruned on
save except the currently-displayed published version.
- `/home` bottom-of-page renders the latest published intro with a
"Read more →" link to `/news` (which renders the full body).
- `/admin/news` editor with sandboxed live preview, versions table,
per-row Unpublish, Format-help cheatsheet.
- `agnes admin news show / draft / edit / publish / unpublish /
versions / export` (CLI). Talks to the live server via the
`/api/admin/news/*` endpoints (PAT-authed) — no direct DB access
so it coexists with a running uvicorn.
- **Optimistic-lock guard**: `agnes admin news publish --version N` and
PUT/PATCH endpoints accept `expected_version` and 409 with structured
`{error: "version_conflict", expected, actual, actual_by}` when a
concurrent admin replaced the draft. Edit refuses to overwrite a
draft authored by someone else without `--force` or
`--expect-version`.
- nh3 (Rust-backed ammonia) HTML sanitizer; iframe pre-pass strips
any iframe whose src is not on the YouTube/Vimeo/Loom allowlist;
javascript:/data: schemes blocked everywhere.
- Author CSS vocabulary: `.news-hero` (blue gradient hero block),
`.callout`/`.callout-{info,warn,success,danger}`,
`.video-embed`, `.news-section`, `.news-grid-{2,3}`, `.news-cta` —
all consolidated in `style-custom.css` under "News content
vocabulary (shared)" so /home perex, /news body, and /admin/news
preview share one source of styling.
- Code-inside-`<pre>` contrast fix (was unreadable amber-on-silver).
- `.news-content` table styling (border, header band, row-hover).
**`scripts/dev/run-local.sh`** — local uvicorn launcher. Pulls Google
OAuth client id/secret from GCP Secret Manager
(`AGNES_OAUTH_GCP_PROJECT`-driven, no vendor defaults), points
`AGNES_CLI_DIST_DIR` at `./dist` so the wheel endpoint resolves, and
`--dev` flips `LOCAL_DEV_MODE=1` + `AGNES_HOME_ROUTE=/home` for one-
command iteration. `LOCAL_DEV_MODE=1` also enables the FastAPI debug
toolbar.
**CLAUDE.md "Run tests before every push" section** codifies
`pytest tests/ -n auto -q` as non-negotiable before each push.
**Tests**: 51 + 14 + 8 = 73 new tests across news-template repo,
sanitizer, API, web, CLI; plus updated home/auth/template tests for
the new shared-shell architecture.
Origin docs (gitignored, customer-fork content):
docs/brainstorms/home-page-requirements.md,
docs/plans/2026-05-07-001-feat-home-page-plan.md.
* feat(cli): agnes onboarded {on,off,status} — self-scoped flag toggle
User-facing equivalent of the in-page "Mark me as (off)boarded" button
on /home. POSTs /api/me/onboarded with {onboarded, source}; --source
overrides the audit-log marker so flips made from the CLI vs the web
button vs agnes init automation stay distinguishable.
`status` reads via /api/me/profile (when present); falls back to a
quick body-marker scan of /home so the read path doesn't write an
audit_log row. PAT-authed via cli.client.api_post — same convention
as agnes admin news / agnes admin add-user etc.
Tests: 5 covering on/off/status round-trip, idempotency, and
audit-log source recording. Full suite holds at 12 pre-existing
failures (same set as before).
* ui(nav+home): primary nav reorg + green What's new band + /marketplace link fix
Primary nav (post-rebase audit + per-user feedback):
- Items: Home → Marketplace → Data Packages → Memory. Admin dropdown
for admins only. The "Dashboard" label was renamed Home — point still
resolves through `home_route` so customer instances on /dashboard
still land there.
- Activity Center moved into the Admin dropdown. Per-team adoption
analytics is admin-consumed in practice; the route still allows
any authed user for direct deep-links so existing /home tile +
bookmarks keep working.
- Memory link added (→ /corporate-memory) — was previously buried in
the /home "Look around" tiles.
- Setup local agent + My Stack dropped from main nav. Setup is the
/home install flow's home now; My Stack lives as a tab inside
/marketplace.
/home tweaks:
- Plugin marketplace tile now points at /marketplace (was /store —
legacy from before the marketplace rebrand landed in #230).
- "What's new" section header gets a green band (success-flavored
D1FAE5 background, A7F3D0 border, darker green title) so the
bottom-of-page news block visibly distinguishes from the blue
install-hero at the top. Header strip only — body stays white.
Test fix: test_home_route_resolution renamed `dashboard_link_uses_home_route`
→ `home_link_uses_home_route` and asserts `href="/home">Home` instead
of `href="/home">Dashboard` after the label change.
* fix(home): decouple Step 3 + Connect-tools collapse from server onboarded flag
The server-side `users.onboarded` flip happens through two paths:
1. Explicit user click on "Mark me as onboarded" or `agnes onboarded on`.
2. Implicit `agnes init` POST → /api/me/onboarded on success.
Path 2 produced a UX surprise: an analyst running `agnes init` mid-flow
reloaded /home and saw Step 3 (auto-mode) + Connect-your-tools auto-
collapse to summary bars. They were actively working through those
sections — the install POST never signalled "I'm done with the rest
of setup", just "Agnes itself is installed".
Decouple the section-collapse decision from the server flag:
- Step 1 + Step 2 install blocks: still hidden on `onboarded=TRUE`
(their completion is a hard server signal — Agnes IS installed).
- Step 3 + Connect-your-tools: render flat by default in BOTH states.
Wrapped in `<details class="setup-collapsible" open>` so the
browser's native disclosure handles per-section toggle without JS,
but the `<summary>` is CSS-hidden until the page-level
`data-setup-minimized="1"` attribute is set on `.home-mock`.
- New "Minimize setup view" toggle inside the blue install-hero,
rendered only when onboarded. Click flips the data-attr on
`.home-mock` AND removes the `open` attribute from each
`<details>`. State persists in `localStorage["agnes_home_setup_minimized"]`
so the choice survives reloads but is per-device.
- "Show full setup view" (the same button when minimized) re-opens
both `<details>` and clears localStorage.
When minimized, each `<details>` still has its own native expand/
collapse — click the gray summary bar to peek at one section without
toggling the page-level minimize off.
Tests:
- test_step3_and_connectors_render_flat_when_onboarded_by_default —
asserts `<details class="setup-collapsible" ... open>` for both
sections post-onboarding and the absence of any server-rendered
`data-setup-minimized` attribute on the `.home-mock` root.
- test_minimize_toggle_visible_only_when_onboarded — toggle button
rendered only when onboarded.
Full pytest holds at 12 pre-existing failures (same set).
* feat(observability): optional PostHog integration (errors, LLM traces, replay, flags)
Off by default. Activates when POSTHOG_API_KEY is set in env. Defaults
to PostHog Cloud EU; override host for US Cloud or self-hosted.
Coverage:
- FastAPI 500 handler captures unhandled exceptions
- src/orchestrator.py rebuild + rebuild_source failures
- services/scheduler/ HTTP-job failures
- cli/main.py uncaught CLI errors (Typer.Exit/SystemExit/KeyboardInterrupt
skipped; flushes before re-raise so short-lived CLI invocations don't
drop events)
- connectors/llm/anthropic_provider.py + openai_compat.py emit
$ai_generation events with provider, model, latency, token counts
(prompt/completion bodies stay off unless POSTHOG_LLM_PAYLOADS=1
because LLM prompts here routinely include customer SQL/data)
- Browser snippet injected into every text/html response by
PosthogInjectionMiddleware — registered inside the GZip layer so it
sees uncompressed HTML before compression. Many templates are
standalone (their own DOCTYPE) and never extend base.html, so a
per-template include would miss them.
- Frontend: $pageview, $pageleave, JS error capture via window.error
and unhandledrejection handlers, masked session replay
(maskAllInputs: true plus CSS-selector mask for known data surfaces),
feature flags (browser posthog.isFeatureEnabled + server-side
feature_enabled with fallback for older SDKs).
Identification mode operator-configurable: none / id / email / full.
Default email ships user.id + email but never name. CLI entry point
moves from cli.main:app to cli.main:main (Typer wrapper).
Files:
- src/observability/posthog_client.py — lazy singleton, no network
when disabled, single-process flush on shutdown
- src/observability/llm_tracing.py — trace_generation context manager
- app/middleware/posthog_inject.py — HTML rewrite middleware
- app/web/templates/_posthog.html — browser snippet template
- docs/observability.md — operator guide
- config/.env.template — documented POSTHOG_* knobs
- tests/test_posthog_disabled.py + tests/test_posthog_client.py +
tests/test_llm_tracing.py — 18 tests covering disabled state,
identify-mode payloads, $ai_generation shape, error variant.
CHANGELOG entry under [Unreleased] Added.
* feat(observability): tag every PostHog event with environment + release
Splits PostHog dashboards cleanly between localhost / dev / staging /
production without manual tagging on every capture call.
- POSTHOG_ENVIRONMENT explicit override; auto-resolves to "local" when
LOCAL_DEV_MODE=1, else RELEASE_CHANNEL, else AGNES_DEPLOYMENT_ENV,
else "unknown".
- AGNES_VERSION → RELEASE_CHANNEL fallback feeds the `release` property
for "is this error new in this release?" cohorting.
- Backend gets both via the PostHog SDK's super_properties constructor
arg (every captured event picks them up automatically).
- Browser snippet calls posthog.register({environment, release}) inside
the loaded callback so $pageview, $exception, autocapture, etc. all
carry the same labels.
- request.state.user now populated by auth dependencies so the snippet
can actually call posthog.identify(user_id, {email}) for logged-in
users (previously the user block always resolved to None because
nothing wrote to request.state.user).
4 new tests cover env resolution: explicit > LOCAL_DEV_MODE > channel
> unknown, plus super-properties forwarding into the SDK constructor.
* feat(observability): inline user attrs on every PostHog event + debug throw route
PostHog's UI shows person properties on the Person profile page, not
inline on each event — so a reviewer triaging an exception couldn't tell
which user hit the bug without clicking through. Fix it on both sides.
- Backend capture_exception merges user_id / user_email / user_name into
the event properties (gated by POSTHOG_IDENTIFY_PII: none/id/email/full).
Backed by a new _user_props_for_event helper on PosthogClient.
- Browser snippet registers user_id + user_email + user_name as super-
properties via posthog.register({...}) so every $exception, $pageview,
and custom event coming from posthog.captureException() carries them
inline. Mirrors the backend so cross-referencing client/server events
doesn't require a person-profile lookup.
- /api/debug/throw — debug-only endpoint gated by DEBUG=1 (404 in prod).
Runs Depends(get_current_user) first so request.state.user is set when
the unhandled-exception handler captures the event. Lets operators
exercise the full observability path end-to-end without hand-rolling
a TestClient script. Configurable via ?kind=ValueError&msg=...
7 new tests cover: backend user-attr merge across identify modes,
anonymous request fall-through, browser snippet super-prop emission for
logged-in / anonymous / id-only / full-name cases.
* fix(observability): address minasarustamyan PR #231 review
Two bugs caught in review.
1. PosthogInjectionMiddleware dropped Response.background on every
return path. BaseHTTPMiddleware materialises the body and asks
subclasses to return a fresh Response — three paths in dispatch()
omitted background=, silently cancelling any BackgroundTask /
BackgroundTasks the route attached (audit logging, async webhooks,
email sends) with no log line. Fix: route every return through a
_passthrough() helper that forwards background.
Also adds a _MAX_BUFFER_BYTES (4 MB) cap so a streamed-HTML response
can't balloon RSS during buffering. Bigger bodies short-circuit
through with a warning rather than being injected.
Regression tests in tests/test_posthog_inject_middleware.py exercise
four return paths (snippet present, render-fail, double-injection
guard, non-HTML passthrough) plus the streaming-guard short-circuit.
2. $ai_input / $ai_output_choices were emitted without truncation, so
POSTHOG_LLM_PAYLOADS=1 silently dropped events past PostHog's ~32 KB
per-event ingest limit — exactly the calls (large prompts with
schemas / sample rows / SQL) an operator would want to inspect.
Fix: clip both at POSTHOG_LLM_PAYLOAD_MAX_CHARS (default 30000) with
an explicit "…[truncated N chars]" marker so readers don't mistake
truncated captures for complete ones. Metadata (provider, model,
tokens, latency, error) flows regardless. Three new tests cover
default-cap clipping, env-override, and pass-through under the cap.
37 PostHog tests pass.
* Add /marketplace browse page + Model B opt-in stack composition
New /marketplace browse surface unifies the curated marketplaces
(admin-managed git mirrors) and the community Flea Market behind
three tabs — Curated / Flea / My Stack — with per-tab category
filter, search across both sources with scope checkboxes, and
numeric pagination, all driven by URL query state. Plugin detail
at /marketplace/curated/<slug>/<plugin> and /marketplace/flea/<id>;
nested skill / agent detail at /marketplace/curated/<slug>/<plugin>/
{skill,agent}/<name> and the flea-side single-page detail.
Model B opt-in: an RBAC grant on a curated plugin is now only
*eligibility*. The user must click "Add to my stack" for it to
enter their served Claude Code marketplace. Composition flips
from (rbac ∖ opt_outs) ∪ store_installs to
(rbac ∩ subscriptions) ∪ store_installs. The legacy
user_plugin_optouts table is renamed user_curated_subscriptions
(schema v27) — same table shape, inverted semantic, repository
methods become subscribe / unsubscribe / is_subscribed.
UX vocabulary: Install → Add to my stack, Installed → In your
stack, card "Installed" badge → "In stack" (amber pill), tab
"My Subscriptions" → "My Stack". Bridges the two-step model
(server-side bookmark vs. on-laptop install) the previous label
hid. Click triggers an inline post-add hint panel under the
description with the agnes refresh-marketplace recipe + Copy
chip, dismissible per-browser via localStorage.
Per-tab info blocks above the filter row:
- Curated: trust signal — "Each plugin here has a named curator
accountable for it." (blue accent + See-all-curators link)
- Flea: open-shelf signal — "Anyone in the company can upload
here." (purple accent + Tips-for-sharing link)
- My Stack: personal-shelf orientation — "Your AI stack —
everything you've added." (slate accent, no link)
Tabs carry per-tab Heroicons (shield-check / building-storefront
/ rectangle-stack) tinted to match each tab's accent; flips white
when the tab is active for contrast.
Hero illustration anchored to the right of the blue hero panel
(absolute, 47% wide, behind the search row content). Hidden
under 900px viewport.
Action-row CTAs realigned to publication intent: curated
"How to add new content" → "Submit a plugin" (links to the
guide page); flea button removed since +Upload sits next to it.
Empty-state CTAs match. /marketplace/guide/{curated,flea}
routes now host publication-flow guide pages with placeholder
ledes — full copy to be authored separately.
Categories: Heroicons-based icons mapped per category in
src/category_icons.py (zero new dependencies; SVG path strings
inlined). Marketplace cards, filter pills, and detail pages
read from the same source.
API endpoints under /api/marketplace:
- GET /items per-tab listing (curated / flea / my)
- GET /categories per-tab non-zero counts
- GET /curated/{slug}/{plugin} plugin detail
- POST/DELETE /curated/{slug}/{plugin}/install subscribe toggle
- GET /curated/{slug}/{plugin}/{skill,agent}/{name} inner item
The tab=my branch reads directly from
user_curated_subscriptions ∪ user_store_installs (not
resolve_user_marketplace, which bundles flea skills/agents into
a single store-bundle synthetic entry useful for serving the
Claude Code marketplace ZIP/git but wrong for browsing where
each item should appear as its own card).
Detail pages: plugin detail surfaces inner skills/agents as
clickable nested cards; commands/hooks/MCPs render as plain
name lists. Skill/agent detail mirrors the plugin layout with
kind-tinted accents (skill = green, agent = purple), Description
+ Details sidebar, Files + Docs sections, and the "How to call
it" copy-able invocation chip showing /<plugin>:<inner-name>
exactly as Claude Code namespaces it post-install. Curated
nested has no install button — links back to the parent plugin.
Navbar: standalone "My AI Stack" relabelled "My Stack" and
points at /marketplace?tab=my; "Store" link removed (Store
flow is reachable via the Flea Market tab's +Upload button).
The standalone /my-ai-stack and /store routes still work for
old bookmarks.
Tests cover the new browse / categories / install / RBAC paths
under tests/test_marketplace_api.py; existing marketplace and
store tests updated for Model B (explicit subscribe in fixtures).
Schema bumped v26 → v27 with idempotent migration that wipes
existing user_plugin_optouts rows on flip and adds
marketplace_plugins.created_at with registered_at backfill.
* Fix v28 migration + post-rebase test fallout
v28 ALTER TABLE marketplace_plugins ADD COLUMN created_at conflicted with
_SYSTEM_SCHEMA's earlier CREATE that already includes the column on fresh
installs (test fixtures starting at any pre-v28 version trip on it).
Switch to ADD COLUMN IF NOT EXISTS — same idiom as the upstream v27
Keboola sync-strategy migration on the same ladder.
Two test patches needed after the rebase bumped SCHEMA_VERSION 27 → 28:
- test_keboola_v27_migration.py: test_schema_version_constant_is_27 was
pinning ==27. Loosened to >=27 (the test's purpose is to verify the
v27 Keboola migration, not to pin the current SCHEMA_VERSION).
- test_setup_page_unified.py: was monkeypatching resolve_allowed_plugins
but compute_default_agent_prompt now reads from resolve_user_marketplace
(Model B-aware). Stub the right function so the test exercises the
v28 served-set path.
* Harden curated skill/agent inner endpoints against path traversal
`_read_inner`, the `skill_dir` walk in `curated_skill_detail`, and the
`agent_path.stat` in `curated_agent_detail` joined URL path-params onto
`plugin_root` without verifying the resolved candidate stayed inside it.
Starlette's `[^/]+` on `{skill_name}` / `{agent_name}` blocks the direct
URL exploit (encoded `/` 404s before the handler), but a curator-planted
symlink inside a curated marketplace's git mirror could still dereference
outside the plugin tree on read.
Adds `_safe_join(plugin_root, *parts)` doing
`Path.resolve(strict=True)` + `relative_to(plugin_root.resolve())`, used
by all three call sites so the boundary is enforced once and consistently.
Tests cover the helper directly (normal path resolves, escaping `..`
returns None, escaping symlink returns None, missing file returns None)
plus an end-to-end check that the symlink case actually 404s on the
HTTP endpoint. Symlink tests skip on Windows where symlink creation
needs elevated permissions; they run on Linux CI.
---------
Co-authored-by: Minas Arustamyan <arustamyan.minas@gmail.com>
## Summary
Brings the Keboola connector to feature parity with the legacy internal data-analyst's per-table sync strategies. Closes the four documented gaps from the spec branch (`zs/keboola-connector-specs`):
- **Typed parquet** in the legacy SDK extraction path — column types from Keboola Storage metadata (provider cascade `user > ai-metadata-enrichment > keboola.snowflake-transformation`) survive the CSV → parquet roundtrip; invalid date strings (`'0000-00-00'`) and invalid numeric strings (`'Non-Manager'`) become NULL while keeping the column's typed schema. Pre-fix everything was VARCHAR.
- **Incremental sync** via Storage API `changedSince` — opt-in per table; pulls only delta rows, merges into the existing parquet by `primary_key` (drop_duplicates with keep='last'). Cuts daily extraction from O(full table) to O(delta).
- **Partitioned sync** — flat per-partition layout `data/<table>/<key>.parquet` (e.g. `2026_05.parquet`), per-affected-partition merge for daily updates, chunked initial load with 1-day overlap and 2-empty-chunk stop heuristic.
- **`where_filters`** — server-side row filter with date placeholders (`{{today}}`, `{{last_3_months}}`, `{{start_of_3_months_ago}}`, etc.) resolved at sync time. Force the SDK path; reject `incremental + where_filters` combination at API layer (changedSince already filters temporally).
## Architecture
- **Schema migration v25 → v26**: 7 new columns on `table_registry`. Existing `sync_strategy` column reused (pre-v26 it was inert catalog metadata; post-v26 the extractor dispatches off it).
- **Per-table dispatcher** in `extractor.run()` routes to one of `_extract_via_extension` (full_refresh + extension), `_extract_via_legacy` (full_refresh + filters or extension fallback), `extract_incremental`, or `extract_partitioned`.
- **API conflict policy**: `incremental + where_filters` → 422; `partitioned + query_mode='remote'` → 422; `partitioned ⇒ partition_by required`.
- **Admin UI**: third "Direct extract (Storage API)" radio in the Keboola Register / Edit modals, alongside existing "Whole table (extension)" and "Custom SQL". When selected, exposes a v26 sync-strategy panel with conditional fields per strategy.
## Test plan
- [x] **Unit + module** — 134 v26 tests covering migration, repo, parquet_io, where_filters, incremental (compute_changed_since + merge_parquet + extract_incremental E2E), partitioned (key derivation + merge_partition + chunked windows + extract_partitioned E2E), extractor dispatcher, admin API validators, PUT field clearing, registry-shape → dispatcher bridge
- [x] **HTML form structure** — all v26 inputs + visibility classes + JS payload fields verified in rendered template
- [x] **Real Keboola roundtrip** — registered a small test table as `sync_strategy='incremental'` against a test Storage project, triggered two syncs:
- Sync 1: `changedSince=None` → full pull → 9 rows typed parquet
- Sync 2: `changedSince=last_sync - 1d window` → 9 delta rows merged with 9 existing → 9 after dedup on primary_key (PK merge confirmed)
- [x] **Browser UX** — agent-browser session against a local uvicorn: login → admin/tables → register modal → switch radios → verify field visibility per strategy → submit → edit existing row → switch to Direct/Incremental → save → confirm DB persistence
- [x] **Regression** — no regressions in the broader 3252-test suite (3 pre-v26 tests updated for the deprecation-marker removal + schema-version bump; 2 pre-existing environment-sensitive test failures unrelated to this change)
## Bugs caught + fixed during E2E
The browser + real-Keboola roundtrip exposed four bugs the unit tests missed:
1. **JS visibility race** — two competing `forEach` loops set `display=''` then `display='none'` on form elements sharing `kb-strategy-incremental kb-strategy-partitioned` classes (window_days + max_history_days are reused across strategies). Fix: single-pass selector with class-based visibility resolver.
2. **PUT cannot clear field** — pre-v26 `updates = {k: v ... if v is not None}` collapsed "omitted from body" and "sent as null" into the same case, so admin couldn't switch a partitioned row back to full_refresh and have stale `partition_by` clear. Fix: `model_dump(exclude_unset=True)`.
3. **Subprocess DB lock conflict** — `_read_last_sync` reopened `system.duckdb` while the parent server held the write lock (subprocess contract at `app/api/sync.py:_run_sync` line 260). Fix: parent injects `__last_sync__` into table_config before subprocess spawn.
4. **Wrong KBC table_id** — `extract_incremental` / `extract_partitioned` built the Storage API table_id from the registry row's slugified `id` (`circle_inc`) instead of `bucket.source_table` (`in.c-finance.circle`), producing 404s. Fix: prefer `bucket+source_table`; fall back to `id` only when bucket empty.
## Operator notes
- Existing tables stay on `full_refresh` after migration; admins opt individual tables in via `agnes admin register-table --sync-strategy ...`, the Keboola Edit modal, or `POST/PUT /api/admin/registry`.
- `merge_parquet` and `merge_partition` use `pd.concat + drop_duplicates`, loading both existing and delta into pandas RAM. For tables in the multi-million-row range this may OOM — switch to `partitioned` strategy for those (per-partition merge keeps memory bounded). Documented in `### Internal` of the changelog entry.
- Date placeholders are resolved at **sync time**, not register time — a typo'd `{{lasst_week}}` is accepted at register and surfaces only when the next sync runs. By design (rolling windows need late-binding).
## Spec source
The four corresponding plans on the `zs/keboola-connector-specs` branch under `docs/superpowers/plans/2026-05-07-0[1-4]-*.md` capture the design rationale and link back to internal repo references for each subsystem.
<!-- devin-review-badge-begin -->
---
<a href="https://app.devin.ai/review/keboola/agnes-the-ai-analyst/pull/217" target="_blank">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://static.devin.ai/assets/gh-open-in-devin-review-dark.svg?v=1">
<img src="https://static.devin.ai/assets/gh-open-in-devin-review-light.svg?v=1" alt="Open in Devin Review">
</picture>
</a>
<!-- devin-review-badge-end -->
## Summary
- Catalog enrichment for `query_mode='remote'` rows: `rows`, `size_bytes`, `partition_by`, `clustered_by` per table (BQ + Keboola providers).
- `/api/v2/schema/{id}` cache miss: 2 BQ jobs → 1 (-50%) via shared `fetch_bq_columns_full`.
- All four catalog/schema/sample/metadata caches flush on registry change; single-row re-warm scheduled.
- Automatic cache warmup at server startup (bounded concurrency, opt-out via `AGNES_SKIP_CACHE_WARMUP=1`).
- SSE-driven freshness toolbar on `/admin/tables` with progress bar, log, and per-row badge.
- New admin doc `docs/admin/query-modes.md` — single source of truth on `local` / `remote` / `materialized` choice.
Closes#155.
Closes#156.
## Test plan
- [x] 65+ targeted tests pass across 11 new test modules + 3 modified ones.
- [x] No DB migration; no wire-break; `MIN_COMPAT_CLI_VERSION` unchanged.
- [ ] Reviewer: register a remote BQ table via `/admin/tables`, observe the toolbar populates within ~2 s and the per-row badge transitions warming → fresh.
- [ ] Reviewer: trigger `Re-warm all`, verify SSE log scrolls and `cacheWarmupBar` progresses.
- [ ] Reviewer: edit a registered row's bucket, verify `agnes schema <id>` returns updated columns immediately (no 1-hour staleness).
- [ ] Reviewer: confirm `agnes admin register-table --query-mode remote` prints the new IAM-smoke-check hint.
## Notable design decisions
- BigQuery `INFORMATION_SCHEMA.TABLE_STORAGE` is the only valid scope for size+rows (verified live 2026-05-07; dataset-scoped doesn't exist). Region resolved from `instance.yaml.data_source.bigquery.location` → `bq.client().get_dataset(...)` → fall back to legacy `__TABLES__`.
- VIEW handling: TABLE_STORAGE returns no rows for views, fall through to `__TABLES__` (also empty) → `TableMetadata(rows=None, size_bytes=None, partition_by=..., clustered_by=...)`. Null size signals analyst Claude to apply existing CLAUDE.md guidance.
- `size_bytes` is `active_logical_bytes + long_term_logical_bytes` — full BQ scan reads both; reporting only active undercounts aged partitioned tables.
- Source-agnostic provider seam: per-source `connectors/<source>/metadata.py:fetch(MetadataRequest)`; dispatcher in `app/api/v2_catalog.py:_metadata_provider_for` lazily imports per source_type so a Keboola-only deployment doesn't pay the BQ-extension import cost.
- Warmup non-blocking: FastAPI `lifespan` schedules `asyncio.create_task(_warm_catalog_caches_bg)` before `yield`. Per-row failures isolated.
## Out of scope
- Profile / column histograms / dimension cardinality for remote tables (separate issue).
- Onboarding nudge ("you have 0 remote tables, consider registering some BQ ones") — separate UX call.
- Provider plug-in registration via entry-points (the dispatch table is a hardcoded if-tree today; one line per future source).
## Release
Bumps `pyproject.toml` 0.46.1 → 0.47.0 (main shipped 0.46.0 + 0.46.1 during this PR — see commit `d98976ec`). New CHANGELOG section under `## [0.47.0] — 2026-05-07`.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- devin-review-badge-begin -->
---
<a href="https://app.devin.ai/review/keboola/agnes-the-ai-analyst/pull/223" target="_blank">
<picture>
<source media="(prefers-color-scheme: dark)" srcset="https://static.devin.ai/assets/gh-open-in-devin-review-dark.svg?v=1">
<img src="https://static.devin.ai/assets/gh-open-in-devin-review-light.svg?v=1" alt="Open in Devin Review">
</picture>
</a>
<!-- devin-review-badge-end -->
The 'Add to group' dropdown on /admin/users/{id} silently filtered out
every Google-Workspace-managed group (rightly — the API would 409 on
POST). On deployments where Admin and Everyone are both Workspace-mapped
via AGNES_GROUP_{ADMIN,EVERYONE}_EMAIL and no custom Agnes groups exist
yet (FoundryAI prod + dev today), the picker showed only the literal
'— Pick a group —' option with the 'Add' button disabled. Operator had
no indication that they needed to create a custom group first.
Three states surface a hint below the picker now:
- user is already in every group (literally nothing left)
- every remaining group is Google-Workspace-managed (link to
/admin/groups + admin.google.com explainer)
- no groups exist at all
The skip-google-managed logic stays — POST would still 409 on those
rows, this just stops the empty-state from being a silent dead end.
Vojtech
