Explain the actual problem (which withSpan callback owns an app-level
call, and why the first await makes that unprovable in browsers) before
presenting the two modes as policies for that gap.
- Deleted references to custom events, cross-tier tracing, and framework adapters from the guides.
- Updated the Analytics app documentation to reflect the current structure, reducing the number of areas from five to three and clarifying the features available.
- Adjusted the description to emphasize the core functionalities of the Analytics app, including event tables, SQL queries, and session replays.
Same content as the legacy-docs commit, ported to the live Mintlify app
with its conventions (Hexclave branding, @hexclave/* imports, Analytics
nav group). Also fixes the stale claim here that replay recording is
disabled by default.
Three new concept pages covering the full spans stack (SDK events/spans
API, automatic client→backend session linking, and the tRPC/oRPC/Elysia/
Convex adapters), plus the Traces waterfall section on the Analytics app
page. Also corrects the pre-existing claim that replay recording is
disabled by default — it is on by default with analytics.
- Added a new StickyPageHeader component to enhance the user interface by providing a sticky header that morphs during scrolling.
- Integrated the StickyPageHeader into the metrics page and spans-events analytics page for improved navigation.
- Updated the .gitignore to exclude .gstack/ files.
- Refactored existing header logic to utilize the new sticky header functionality, improving code organization and maintainability.
- Trace list is now a hierarchical span tree: every span (e.g.
$session-replay-segment) is directly reachable and selectable —
selecting focuses it in the waterfall. First 3 levels expand by
default, deeper levels collapse behind carets. The type filter prunes
the tree to matching subtrees with match highlighting and forced
ancestor expansion.
- Timeline scale clamps at 'now': intervals reaching into the future
($refresh-token runs to its expiry a year out) no longer compress the
real activity into a sliver; future ends render as a fading stub with
an 'elapsed →' duration and end-date tooltip.
- Waterfall zoom/pan: cmd/ctrl+scroll zooms around the cursor,
horizontal scroll pans while zoomed, reset button in the ruler;
ew-resize cursor on timeline tracks as affordance. View window math
(zoomViewWindow/panViewWindow) is pure and unit-tested.
- Trace start now includes attached event times (events can predate
their span's started_at since events and replay chunks batch
independently), so nothing renders off-scale at full view.
Every span row (except the current view root) gets a hover focus control
that re-roots the waterfall at that span: the subtree is re-based to
depth 0 and duration/counts/open state are recomputed over the subtree
only, so e.g. a 2-minute $session-replay nested in a year-long
$refresh-token becomes fully readable instead of a sliver. A breadcrumb
in the header climbs back to any ancestor. Focus resets when selecting
another trace.
New analytics page at /projects/[id]/analytics/spans-events: traces
master-detail (trace list + waterfall with spans and event markers,
depth-indented, chronologically interleaved) and an events tab with
type-breakdown chips. Data comes from default.spans/default.events via
the existing admin analytics query endpoint — no backend changes.
Trees are rebuilt client-side from root-first parent_span_ids; rows
re-parent to the nearest FETCHED ancestor so subtrees survive spans
falling outside the time window, with a cycle guard for hand-crafted
parent ids. Custom scope hides $-prefixed system spans/events; All
scope shows full session trees.
Full-SDK adapters: each resolves the authenticated caller onto the framework's
context AND wraps the handler in a request-linked span, so backend telemetry
nests under the caller's client session ($refresh-token/$session-replay/
$session-replay-segment) with ZERO per-route plumbing — consumers never pass
{ request } themselves (they can via the core primitive, but don't need to).
- integrations/adapter-core.ts: shared core — normalizeRequestLike, lazy
memoized per-request user resolution, runRequestSpan (withSpan({request}))
- trpc.ts: createHexclaveTRPC(t, app) → createContext + middleware({ required,
telemetry }); no @trpc/server dep (TRPCError-shaped UNAUTHORIZED rejection)
- orpc.ts: createHexclaveORPC(app, { unauthorized }) → middleware +
wrapFetchHandler (oRPC middleware can't see the Request otherwise); refuses
required:true without a real ORPCError factory (branded class, not faked)
- elysia.ts: createHexclaveElysia(app) → resolveUser (.resolve), requireUser
(beforeHandle 401 guard), handler (around-wrapper — the only honest way to
put an Elysia handler inside a span's async extent)
- convex.ts: hexclaveConvexFunction(app, handler, { required, kind, name }) —
identity-based attribution (Convex is WebSocket-transported; no Request, so
segment linking is a documented follow-up)
Span types stay low-cardinality (trpc.procedure/orpc.procedure/elysia.route/
convex.function) with path/type/kind in span data. New subpath exports ./trpc
./orpc ./elysia ./convex on all four SDK packages. 14 unit tests.
Browser withSpan frames can mix across concurrently interleaved async flows
(no async-context primitive exists there — the reason for TC39 AsyncContext).
Support BOTH policies via analytics.ambientParenting:
- "exact" (default): a frame is ambient only when provably from the current
flow — an exact primitive (ALS on servers/edge; AsyncContext in browsers
once shipped), or the callback's SYNCHRONOUS window on the browser fallback
(single-threaded sync execution cannot interleave, so prologue-open frames
are exact by construction). Never wrong; can only under-attach.
- "best-effort": frames also stay ambient across browser awaits (zero-glue),
accepting the documented cross-flow caveat.
Handle kit on every Span (browser, server, inert) — exact everywhere, both
modes: span.withSpan (nested child), span.run (re-bind for callbacks/timers),
span.getPropagationHeaders (non-fetch transports), span.fetch (pinned
cross-tier header, same origin policy as the auto wrapper, explicit wins).
Internals: sync-stack frames carry a prologueOpen flag (set false the moment
the callback returns its promise); context entry runs synchronously once the
ALS probe settles so sync nesting composes; getAmbientSpanRefs gains an
includeSuspendedSyncFrames read policy; servers fail closed when ALS is
missing. Design validated by Codex (gpt-5.5 xhigh) + a 3-lens design panel.
The browser sync-stack fallback can mix ambient frames from overlapping async
flows into the propagated context (fundamental until TC39 AsyncContext ships —
the span-context loader will pick it up like ALS when it does). Give affected
callers an exact override:
- the auto fetch wrapper never overwrites an explicitly-set
x-hexclave-span-context header — caller intent beats ambient
- getSpanPropagationHeaders({ parentIds, root }): root drops ambient parents
(same semantics as TrackOptions.root), so one request can be pinned to
exactly one span; the segment id (identity, not parenting) always rides
The header's customParentSpanIds now carry the SAME ambient ancestry a
locally-tracked client event would get — global spans (setGlobalSpan) first,
then enclosing withSpan() frames, each contributing its full frozen chain —
flattened root-first and deduped, instead of only the withSpan frames' leaf
ids. Overflow keeps the NEAREST ancestors (tail), mirroring resolveParentIds.
- event-tracker: expose getAmbientParentRefs() (globals + frames)
- span-propagation: pure flattenParentRefsToIds(refs, extraParents) + nearest-10 cap
- client-app-impl: shared _getSpanPropagationContext + public
getSpanPropagationHeaders({ parentIds }) escape hatch for non-fetch
transports (XHR/sendBeacon/WebSocket) and explicit per-request parents
- server-app-impl: client-propagated custom parents now merge BEFORE server
ambient frames (client ancestry is the outer context), so the root-first
order reads client chain -> server frames -> explicit parents
Auto-links a customer's BACKEND span to the caller's CLIENT session with zero glue.
The browser attaches one `x-hexclave-span-context` header to same-origin outgoing
fetches; `serverApp.withSpan(type, { request }, fn)` resolves the caller's refresh
token (from the session) plus that header, so the span — and everything created
inside the callback — parents under $refresh-token → $session-replay →
$session-replay-segment, exactly like a browser event.
- span-propagation.ts: versioned base64url header codec, same-origin fetch
auto-wrapper (idempotent, native header semantics preserved), origin predicate
- server-request-context.ts: AsyncLocalStorage ambient request context
- server-app-impl: withSpan/trackEvent({ request }) resolution; server telemetry
buffer re-keyed from userId to the full (user, refresh, replay, segment) context;
propagated custom parents flow as ambient frames
- backend events/batch: accept forwarded refresh_token_id/session_replay_id on
SERVER auth (rejected on client auth); compose [rti-,sri-,srsi-,cs-] + stamp
the scalar columns
- client-app-impl: install the fetch wrapper on init behind analytics.spanPropagation
(same-origin default + exact-origin allowlist); reads the live per-tab id
- tests: 24 propagation unit tests + 2 e2e (server-auth ancestry + client-auth reject)
The framework adapters (tRPC, Convex, oRPC, ElysiaJS) build on this primitive in a
separate stacked branch; with an adapter you never pass { request } yourself.
- app.withSpan(type, [options,] fn): auto-start, ambient parent for everything
created inside fn (client + server via virtual startSpan dispatch), auto-end
on settle; on throw records data.error and rethrows (telemetry never fails fn)
- span-context.ts: AsyncLocalStorage from the BUILT-IN node:async_hooks via a
bundler-opaque dynamic import (works on Node/Bun/Deno/Workers/Edge; browsers
fall back to a module-level sync stack with documented interleaving limits)
- root: true drops all ambient parents; excludeParentIds filters the FINAL
merged parent list (an excluded span stays excluded even when re-entering via
a kept child's chain) — both on TrackOptions and StartSpanOptions
- analytics.waitUntil constructor option: every batch-send promise (browser
tracker + server coalescer) is passed to it so un-awaited sends survive
serverless teardown; stripped from serialized app options (toClientJson)
Tests: 5 span-context cases (ALS nesting across awaits, parallel isolation,
sync-stack cleanup), 4 event-tracker cases (withSpan auto-end + ambient
parenting, error path, root/exclude final-list semantics, waitUntil hook)
Client SDK (browser):
- app.trackEvent(type, data?, {parentIds?}) and app.startSpan(type, {data?, parentIds?, startedAtMs?}) on StackClientApp
- Span handles: setData (merge + re-write), end (idempotent), trackEvent/startSpan (inherit full chain), ref() for cross-tier parenting
- setGlobalSpan/unsetGlobalSpan (ambient parents for spans AND events, auto-unset on end), app.flush()
- Per-item promises settle on batch ack; pre-caught (no unhandled rejections); nothing throws synchronously
- Spans are versioned upserts: written open on start, re-written on setData/end with per-handle monotonic updated_at_ms
- Batch-poisoning protection: full item validation at enqueue (name regex, 16KB data cap, uuid parents, depth<=10)
- clearBuffer rejects pending + inert-ifies live spans (sign-out privacy, pairs with segment-id rotation)
Server SDK:
- serverApp.trackEvent/startSpan with explicit {userId}; same-tick coalescing per userId (one POST per sync burst)
- sendAnalyticsEventBatchAsServer on HexclaveServerInterface (secret server key -> analytics base URL)
Backend:
- events batch route accepts custom event types, events[*].parent_span_ids, spans[], and server/admin auth with body user_id (verified in tenancy)
- buildCustomSpanRows: cs- prefixed ids, system ancestry (rti-/sri-/srsi-, same gating as events) prepended to client chain, version clamped to now+5min
- custom spans insert off the response path (waitUntil), quota debits events+spans
- stripLoneSurrogates moved to lib/clickhouse for reuse
Tests: 8 new spans.test.ts cases, 9 new event-tracker.test.ts cases, 17 new e2e cases (e2e not executed: dev stack down; 2 pre-existing snapshots updated for intentional schema-message changes)
- span version is now the span's own end time (data-tied, never regresses)
- replay/segment span inserts moved off the response path (waitUntil)
- session_replay_segment_id rotates on sign-out to prevent cross-user correlation
- new SessionReplayChunk index narrowing per-segment bounds aggregation
Enhanced the description for the "/emails/send-email" endpoint to clarify the options for sending emails, including user IDs, all users, and arbitrary email addresses. Additionally, modified the parameters to remove the requirement for "user_id" and added an optional "email" field.
Add a generic Spans telemetry surface alongside events, written directly to
ClickHouse (spans are siblings of events, not routed through ext-db-sync).
- analytics_internal.spans table (ReplacingMergeTree, versioned) + default.spans
view, isolated by a row policy and granted to limited_user exactly like
default.events.
- Span types: $session-replay and $session-replay-segment (per browser tab),
written directly from the session-replays batch route and re-written per batch
so their end advances; $refresh-token, derived in the view from the synced
refresh_tokens dimension.
- events.parent_span_ids: additive column holding the root-first ancestor span-id
list ([rti, sri, srsi]), stamped on each event insert. Typed-id prefixes live
only on span ids / parent_span_ids; scalar id columns stay raw so existing
customer SQL keeps working (additive, backwards compatible).
- SDK: EventTracker and SessionRecorder now share one per-tab
session_replay_segment_id (each previously minted its own, so events could not
be correlated to a replay recording).
- Dashboard: Spans tab on the analytics tables page.
- Tests: spans unit tests + e2e for session replays, analytics events batch, and
the analytics query surface.
<!--
Make sure you've read the CONTRIBUTING.md guidelines:
https://github.com/hexclave/hexclave/blob/dev/CONTRIBUTING.md
-->
## Summary
This PR tightens dashboard consistency across the project list, team
invite flow, hosted auth previews, and shared overlay surfaces. It also
includes smaller visual fixes for charts/tooltips, globe sizing, sticky
data-grid surfaces, and dashboard/hosted component light-mode parity.
**Base:** `dev@2c5b2d02c` -> **Head:**
`more-consistent-dashboard@b1eae206`
**Capture notes:** screenshots were captured locally at `1440x1000`
after signing in through GitHub -> `admin@example.com` on the mock OAuth
server. Each page/state was allowed to settle for 5 seconds immediately
before capture. Dark mode is used by default; light mode is included
only for the light-specific overlay/dialog surfaces changed in this PR.
## Visual Comparison
### Hosted Auth Preview - Dark Mode
| Before | After |
| --- | --- |
| 
| 
|
### Team Invite Dialog - Light-Specific Surface
| Before | After |
| --- | --- |
| 
| 
|
## What's Changed
- Reworks the team invite dialog with clearer loading/retry states, seat
information, capacity warnings, pending invites, revoke actions, and
refresh behavior.
- Adds a dashboard-local hosted auth preview used by Auth Methods and
onboarding-style flows, showing sign-in/sign-up methods with project
display names and disabled sign-up behavior.
- Standardizes overlay and popover surfaces across dashboard, hosted
components, and shared UI primitives for better light/dark consistency.
- Adjusts chart, tooltip, globe, sidebar drawer, data-grid, and
email-log presentation details to reduce clipping, bleed, and stacking
issues.
## Notes For Reviewers
- The visual comparison focuses on places where this PR has visible UI
differences: project cards/actions, the team invite modal, overlay menu
styling, and the hosted auth preview.
- Lower-level primitive background/ring changes are represented through
the menu/dialog screenshots rather than separate screenshots for every
consuming component.
## Test Plan
- Captured the before screenshots on `dev@2c5b2d02c` and the after
screenshots on `more-consistent-dashboard@b1eae206`.
- Waited 5 seconds on each page/state before capturing screenshots.
- Uploaded exactly 10 screenshots to a fresh gist and embedded them
above.
- Not run in this pass: code tests, because this pass only updates the
PR description/screenshots.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added a live hosted authentication preview for project setup and auth
methods, including sign-in/sign-up states and multiple login options.
* **Bug Fixes**
* Improved invite dialog behavior so seat limits, plan upgrades, loading
states, retries, and invitation refreshes stay in sync.
* Fixed chart tooltip and globe display behavior so overlays, sizing,
and labels render more reliably.
* **Style**
* Refreshed dialogs, menus, tooltips, charts, and sidebars with improved
light/dark theme styling and spacing.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: armaan <armaan@stack-auth.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
## Summary
The RDE dashboard was bundled into the `@hexclave/cli` npm tarball
(~165MB). That made the `npx @latest` auto-update re-exec pull a huge
tarball, which tripped download firewalls (Socket Firewall
false-positives on Replit → `Lock compromised`).
This decouples the dashboard from the CLI: it's published as a **GitHub
Release artifact** (a zip + `manifest.json` on a floating
`dashboard-latest` release) and fetched at runtime, cached on disk. The
dashboard now rolls forward **independently of the CLI version**
(assuming newest dashboards stay compatible with older CLIs — no compat
gate yet, by design for now).
## How it works
- **Runtime** (`packages/cli/src/lib/dashboard-release.ts`): on
`hexclave dev`, fetch `manifest.json` (`{version, sha256, url}`) from a
stable, no-API GitHub URL → cache under
`<dirname(devEnvStatePath())>/dashboards/<version>/` → streamed
download, **sha256 verify**, extract via the existing `extract-zip` dep,
atomic publish. Offline → fall back to the newest cached build.
- **Wiring** (`dev.ts`): `startDashboardIfNeeded` resolves/launches the
cached release; the restart decision now compares the **dashboard
release** version (not the CLI version). `HEXCLAVE_DASHBOARD_DIR` runs a
local build with zero network; a custom dev dashboard command bypasses
releases entirely.
- **No more npx self-update**: the `npx @latest` re-exec in `hexclave
dev` is removed entirely (along with `self-update.ts`, the
`--no-auto-update` flag, and the binName plumbing that only built the
npx call). It existed to keep the bundled dashboard fresh; now the
dashboard self-updates from GitHub, so the re-exec only added an npx
download on every run — the firewall surface. Users update the CLI via
npm/npx themselves.
- **Publishing** (`scripts/package-dashboard-release.mjs` +
`.github/workflows/dashboard-release.yaml`): build standalone → zip →
write manifest → publish immutable `dashboard-v<version>` release +
clobber the `dashboard-latest` manifest. **Fails loudly** if an existing
tag's asset sha differs from a fresh build (dashboard changed without a
version bump → would otherwise advertise a hash the served zip doesn't
match).
- **Build**: CLI build is now `tsdown`-only (no bundled dashboard);
`turbo.json` drops the dashboard standalone build from the CLI build
graph.
## Testing
- **Unit**: 135 CLI tests pass (manifest parsing incl. path-safe version
validation, URL resolution, version picking, dir override).
- **End-to-end (no GitHub)**: a localhost HTTP server + fixture zip
exercised the full fetch path — download, sha256 verify, cache-hit,
roll-forward, **corrupt-sha rejection**, offline→cache fallback (9/9).
- **Real dashboard, no GitHub**: built the standalone, served the real
76MB zip over `python -m http.server`, and confirmed `hexclave dev`
downloads → verifies → extracts → **boots** the dashboard and registers
an RDE session against the hosted backend. Also validated
`HEXCLAVE_DASHBOARD_DIR` boot.
## Review
Three review agents (runtime correctness, packaging/workflow, code
quality) ran; valid findings fixed: path-traversal hardening on
`version`, a concurrent cache rm/rename race, the workflow
sha-mismatch/atomicity gap, plus `errorMessage` dedup and added tests.
## Notes / follow-ups
- No CLI↔dashboard compatibility gate yet (assumes always-compatible);
easy to add later via a min-CLI field in the manifest. This matters more
now that the CLI no longer auto-updates.
- The publish workflow has not run yet, so the first push to `main` will
create the `dashboard-latest` release.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added automated, immutable versioned releases for the standalone
dashboard plus a continuously updated “latest” manifest.
* CLI now fetches, verifies, caches, and supports offline fallback for
the dashboard runtime.
* **Bug Fixes**
* Improved dashboard restart behavior to update only when a strictly
newer published version is available.
* **Chores**
* Removed the CLI’s automatic update/re-launch behavior for more
predictable startup.
* **Tests**
* Added unit tests for dashboard manifest validation and latest-version
selection.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Problem
Session replay batches are uploaded as **raw, uncompressed JSON**. A
single large rrweb full snapshot (>900 KB) is dropped client-side in
`SessionRecorder._flush`, and any batch near the server's 1 MB body
limit is rejected. Meanwhile the **analytics events** path right next
door already gzips on the wire (`encodeAnalyticsBody` + a server-side
gunzip with zip-bomb guards) — replays just never got the same
treatment.
Notably, `dev` was already **red** on the SDK test `sends a single
oversized event alone without dropping it` (the code dropped at 900 KB
while the test expected the event to be sent). This change makes that
test pass.
## What this does
Ports the proven analytics compression pattern to session replays:
- **`packages/shared` (client transport):** generalize
`encodeAnalyticsBody` → `encodeGzipJsonBody` and wire it into
`sendSessionReplayBatch`. Replays now gzip via native
`CompressionStream` (`application/octet-stream`), with the same
fallbacks: keepalive flushes send plain JSON (so the request survives
page teardown), and a missing/throwing `CompressionStream` (Safari <
16.4) degrades to plain JSON.
- **`packages/template` (SDK recorder):** replace the 900 KB hard-drop
with `MAX_SINGLE_EVENT_BYTES = 8 MB`, kept in sync with the server's
decompressed cap. The wire is gzipped downstream, so an oversized single
event is now sent compressed instead of discarded; only an event that
could never fit *decompressed* is dropped.
- **`apps/backend` (route):** add `maybeDecodeBinaryBody` (gunzip with a
1 MB compressed cap + 8 MB `maxOutputLength` zip-bomb guard) via a
`.transform` on the batch route body schema. Plain-JSON bodies pass
through untouched.
## Why gzip / CompressionStream
Benchmarked on a real rrweb capture (Puppeteer, rrweb 1.1.3): a **1.09
MiB full snapshot → 134 KiB gzip (~8×) in ~6.6 ms**. Native
`CompressionStream` adds **0 bundle bytes** and was faster than fflate.
This mirrors what PostHog (gzip, native `CompressionStream` + fflate
fallback) and Sentry (fflate deflate in a worker) both ship — both rely
on compression, not splitting, as the primary mechanism for large
events.
## Tests
- **Client SDK** (`client-interface.test.ts`): gzip+octet-stream on
non-keepalive (round-trip + size assert), plain JSON on keepalive, and
both `CompressionStream`-unavailable fallbacks.
- **SDK recorder** (`session-replay.test.ts`): existing oversized-event
test now passes; new test that a >8 MB event is dropped while a
following normal event still sends.
- **Backend e2e** (`session-replays.test.ts`): gzipped happy path
(`compressed < 1MB < raw`), invalid gzip → 400, oversized compressed →
413, zip-bomb → 400. Verified passing against a live local backend.
## Verification notes
- `@hexclave/shared` + `@hexclave/template` typecheck pass; touched
files lint clean; SDK unit tests pass.
- Two **pre-existing, unrelated** local failures (each confirmed by
stashing this change and re-running on clean `dev`): a backend typecheck
error in `oauth/ssrf-protection.test.ts` (`NODE_ENV` read-only), and 3
session-replay **quota** e2e tests failing in the payments
`setSessionReplayItemQuantity` setup helper.
## Follow-up (out of scope)
The keepalive/unload path still sends uncompressed (64 KB browser
keepalive cap), so a very large final batch on page exit can still be
lost — not addressed here.
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Gzip session replay batches so large rrweb snapshots are sent instead of
dropped. Keepalive flushes cap single-event size to avoid 413s on page
unload, and limits now include a 1 KiB envelope margin to match server
caps.
- **Bug Fixes**
- Client (`@hexclave/shared`): generalized `encodeAnalyticsBody` →
`encodeGzipJsonBody`; gzip via native `CompressionStream` for
`/session-replays/batch`; keepalive and missing/throwing
`CompressionStream` fall back to JSON.
- SDK recorder (`@hexclave/template`): normal flushes allow a single
event up to `MAX_SINGLE_EVENT_BYTES = 8 MiB - 1 KiB`; keepalive flushes
cap single-event size at `MAX_BATCH_UNCOMPRESSED_BYTES = 900 KB`; send
oversized events alone and rely on gzip; drop only above those
thresholds; log a distinct 413 warning with the count of buffered events
dropped and stop the loop.
- Backend: gunzip `application/octet-stream` bodies with caps (`1 MB`
compressed, `8 MiB` decompressed); invalid gzip → 400; oversized
compressed → 413; plain JSON passes through.
- Tests: added client encoding tests for session replays; SDK tests for
>8 MiB drop, keepalive drop, and 413 warning; backend e2e for gzip,
invalid gzip, oversized compressed, and zip-bomb. The oversized-event
SDK test now passes.
- **Refactors**
- Hoisted shared test helpers and trimmed comments.
<sup>Written for commit 855cdb052e.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1675?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Session replay batch uploads can now be automatically gzipped (when
supported) to reduce request size; analytics batch uploads use the same
behavior.
* **Bug Fixes**
* Improved handling of gzipped upload bodies, with stricter validation
and safer size/decompression limits to reject invalid payloads and “zip
bomb” scenarios.
* Updated session replay buffering so oversized events are dropped only
when they exceed decompression constraints, while allowing later events
to upload.
* **Tests**
* Added client, server, and end-to-end coverage for gzip/keepalive
behavior and limit enforcement.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What
The `/internal/analytics/query` route accepts an `include_all_branches`
flag (defaults to `false`). Previously, passing `true` threw a
`HexclaveAssertionError`. Since the flag is reachable from user input,
that assertion surfaced as a 500 / error-tracking noise rather than a
clean response.
This makes the flag a **no-op** for now and documents the intended
behavior with a TODO.
## Behavior
- Regardless of `include_all_branches`, queries remain scoped to the
current branch via the ClickHouse row policy that filters on
`SQL_branch_id` (set from `auth.tenancy.branchId`).
- Callers passing `include_all_branches=true` will still only receive
data for the **current branch** until cross-branch filtering is
implemented.
## Note / follow-up
Because the flag is silently ignored, a caller asking for "all branches"
gets single-branch results with no signal. The TODO tracks implementing
real cross-branch querying. Flagging in case we'd rather surface a
proper client-facing error in the interim instead of a silent no-op.
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Made the `include_all_branches` flag on `/internal/analytics/query` a
no-op so `true` no longer throws and avoids 500s from user input.
Branching isn’t implemented yet, so queries always scope to the current
branch via the ClickHouse `SQL_branch_id` row policy; TODO added for
real cross-branch querying.
<sup>Written for commit a07e1237e9.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1678?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Analytics requests now accept the “include all branches” option
without failing.
* Branch-scoped analytics queries continue to run as before, while the
option is treated as a no-op for now.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
Two related email changes:
### 1. Send to arbitrary email addresses
The `sendEmail` SDK function and `POST /api/v1/emails/send-email` now
accept a third recipient selector, **`emails: string[]`**, mutually
exclusive (XOR) with `user_ids` and `all_users`. These recipients don't
have to belong to a user — they're mapped to the existing internal
`custom-emails` recipient type (no associated user object, can't
unsubscribe, so intended for transactional mail).
- Response `results` items are now `{ user_id?, email? }`; the `emails`
path returns `{ email }`.
- Validation message updated to `Exactly one of user_ids, all_users, or
emails must be provided`.
- Recipient addresses are validated with the shared `emailSchema`.
### 2. Run Emailable on managed custom domains
`LowLevelEmailConfig.type` is widened `'shared' | 'standard'` →
`'shared' | 'managed' | 'standard'`, and `getEmailConfig` returns
`'managed'` for the managed-Resend branch. The Emailable deliverability
check now runs for **shared** and **managed**, and is skipped for
**`standard`** (a customer's own SMTP server or Resend API key).
**Why:** managed domains send through our single Resend *account*
(per-domain scoped keys, but shared account → account-level bounce
penalties are collateral across managed customers), so we own that
reputation. With a custom SMTP server or the customer's own Resend key,
the customer owns their deliverability, so we don't second-guess their
recipients or spend Emailable checks on their volume. The shared
dev-email wrapper remains `shared`-only.
## Tests
- `send-email.test.ts`: added arbitrary-recipient send
(single/multiple), empty-array, and validation cases (both-selectors,
no-selector, invalid email); updated the exactly-one-of snapshot.
**24/24 pass.**
- New `emails/deliverability-gating.test.ts`: shared & managed skip an
undeliverable address (`LIKELY_NOT_DELIVERABLE`), custom SMTP still
sends. **3/3 pass.**
> Note: `email-queue.test.ts` has 2 pre-existing snapshot failures (a
`margin:0rem`→`margin:0` CSS normalization in rendered email HTML) that
are unrelated to this change — confirmed by reproducing them on `dev`
with these changes stashed.
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Adds support for sending emails to arbitrary addresses and enables
Emailable checks for managed domains to protect shared sending
reputation. The API/SDK now accept `emails: string[]` as a recipient
selector, and deliverability checks run on shared and managed sending.
- New Features
- API/SDK: `emails: string[]` recipient selector (XOR with
`user_ids`/`all_users`) in `sendEmail` and POST
/api/v1/emails/send-email. Results return `{ email }` for this path;
addresses validated via `emailSchema`. Recipients map to
transactional-only “custom-emails” (no user, no unsubscribe).
- Deliverability: email config adds `'managed'`; `getEmailConfig`
returns it for managed Resend domains. Emailable check runs for `shared`
and `managed`; skipped for `standard` (custom SMTP or customer Resend
key).
<sup>Written for commit e48cf2ddda.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1681?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Emails can now be sent to arbitrary email addresses, not just existing
users.
* Email sending APIs and client options now accept an email list as a
recipient choice.
* **Bug Fixes**
* Improved recipient validation so exactly one target type must be
chosen.
* Fixed response results to consistently show either a user ID or an
email address.
* Corrected deliverability checks for managed email setups and updated
skip behavior for undeliverable addresses.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Problem
The QA reviewer (`apps/backend/src/lib/ai/qa-reviewer.ts`) extracted the
model's JSON verdict from free-form text with a greedy `/\{[\s\S]*\}/`
regex followed by `JSON.parse`. Whenever Haiku wrapped its JSON in prose
or tool-call text, the captured slice was malformed and `JSON.parse`
threw:
```
SyntaxError: Expected property name or '}' in JSON at position 1 (line 1 column 2)
```
This is Sentry issue **STACK-BACKEND-18F** — **218 occurrences in the
last 30 days** (262 all-time), still ongoing. It's caught (background QA
task, 0 users impacted) but noisy, and on failure the review is silently
marked failed.
### Root cause
The brace-matching heuristic assumes the first `{` in the response
starts the JSON object and the last `}` ends it. Any stray brace in
surrounding prose (a preamble like `"Here's my review {note}: {...}"`,
unquoted/single-quoted keys, or `<function_calls>` tool text) breaks the
extraction. The reviewer output is non-deterministic, so a low steady
fraction of reviews fail.
## Fix
Switch to schema-enforced **structured output** via `generateText`'s
`output: Output.object({ schema })` (AI SDK v6), which keeps the
existing tool-calling loop while constraining the final result to a
validated object. The verdict is read directly off `result.output`,
removing the regex, `JSON.parse`, and manual `typeof` shape validation.
- Net **−12 lines**.
- Trimmed the now-redundant "respond with ONLY valid JSON … {inline
schema}" block from the system prompt (the schema enforces structure);
kept the semantic guidance (flag types, scoring, `needsHumanReview`
rule).
- Kept the existing `overallScore` clamp.
## Verification
- **Typecheck** (`tsc --noEmit`): clean.
- **Lint** (`eslint`): clean.
- **Live end-to-end**: ran the real Haiku-4.5 reviewer against the exact
prompt/input that failed **11/15** on the old path → **15/15** return a
validated object on the new path, zero parse failures.
## Follow-ups (out of scope, noted during investigation)
- These Sentry events leak a live access token + publishable key into
the request-context payload; the OpenRouter key sits in plaintext
(commented) in `apps/backend/.env.local`. Worth rotating/scrubbing.
- Consider capturing `result.text` on the failure path so any residual
structured-output failures are inspectable (currently discarded).
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Replaced regex/JSON.parse extraction in the QA reviewer with
schema-enforced structured output using `ai`’s `generateText` +
`Output.object`. Also defaulted `improvementSuggestions` to an empty
string to avoid false failures when the model omits suggestions.
- **Bug Fixes**
- Added a `zod` verdict schema and read result from `result.output`.
- Removed greedy JSON extraction and manual shape checks.
- Defaulted `improvementSuggestions` to "" to tolerate omission and
prevent false `needsHumanReview` verdicts.
- Slimmed prompt by dropping “respond with ONLY JSON” block; kept review
guidance.
- Kept `overallScore` clamp; tool-calling loop unchanged.
- Verification: typecheck/lint clean; 15/15 successful runs on a
previously failing case.
<sup>Written for commit 72ae71017f.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1688?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* QA reviews now use structured output, making review results more
reliable and consistently formatted.
* Review flags now include clearer details such as type, severity, and
explanation.
* **Bug Fixes**
* Improved handling of review output by removing fragile text parsing
and manual JSON extraction.
* Added safer defaults for missing review suggestions.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
- Standardize the unauthenticated `smart` and `smartest` model tiers in
`apps/backend/src/lib/ai/models.ts` on `z-ai/glm-5.2:nitro` (previously
a mix of `z-ai/glm-5.2` and `z-ai/glm-5:nitro`).
- Document the `npx @hexclave/cli exec <javascript>` command (runs JS
with a pre-configured `hexclaveServerApp`) in the agent reminders.
- Note that the config file is automatically synced when using the local
dashboard/dev environment via `npx @hexclave/cli dev --config-file
<path>`.
These reminder updates are reflected across `docs-mintlify/llms.txt`,
`docs-mintlify/snippets/hexclave-agent-reminders.jsx`,
`docs-mintlify/snippets/home-prompt-island.jsx`, and
`docs-mintlify/guides/getting-started/setup.mdx`.
## Test plan
- [ ] Verify unauthenticated AI requests resolve to `z-ai/glm-5.2:nitro`
for smart/smartest tiers.
- [ ] Confirm docs render the new agent-reminder bullets correctly.
Made with [Cursor](https://cursor.com)
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Standardized unauthenticated `smart` and `smartest` tiers to
`z-ai/glm-5.2:nitro` for both slow and fast paths to ensure consistent
model selection. Updated docs to include `npx @hexclave/cli exec
<javascript>` (preloads `hexclaveServerApp`) and to note config
auto-sync when using `npx @hexclave/cli dev --config-file <path>`.
<sup>Written for commit 3363c8d9f3.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1693?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Updated unauthenticated AI model selection to use a newer variant for
faster and more reliable responses in supported smart modes.
* **Documentation**
* Added clearer CLI guidance for running JavaScript with a
pre-configured app context.
* Expanded reminder text to better explain local config syncing during
development and dashboard runs.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Co-authored-by: Cursor <cursoragent@cursor.com>
## Summary
Switches the `ask_hexclave` MCP tool's model from Gemini 3.5 Flash to
GLM 5.2 by changing the speed parameter from `"fast"` to `"slow"` in the
query to the AI backend.
The model selection matrix maps `smart`/`slow`/`unauthenticated` →
`z-ai/glm-5.2` (previously `smart`/`fast`/`unauthenticated` →
`google/gemini-3.5-flash`).
Link to Devin session:
https://app.devin.ai/sessions/f710490916754099916f9899c2700a5c
Requested by: @mantrakp04
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Switched the `ask_hexclave` MCP tool from Gemini to GLM by updating the
model selection matrix. Unauthenticated smart/fast now routes to
`z-ai/glm-5:nitro`, unauthenticated smartest/fast to
`z-ai/glm-5.2:nitro`, and `dumb/slow/authenticated` uses
`z-ai/glm-4.5-air` (removed `:free`).
<sup>Written for commit 6a51ba6358.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1677?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Adjusted AI request behavior to favor more deliberate responses, which
may improve result quality and consistency.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mantra <mantra@stack-auth.com>
## Problem
Sentry is reporting an `AI_APICallError` on the AI proxy path:
> This model is unavailable for free. The paid version is available now
- use this slug instead: `z-ai/glm-4.5-air`
OpenRouter discontinued the **free** tier for GLM 4.5 Air. The model
selection matrix still pinned the `:free` variant for the **dumb / slow
/ authenticated** path, so every request routed there fails.
## Fix
Drop the `:free` suffix in `apps/backend/src/lib/ai/models.ts` so it
uses the paid slug OpenRouter recommends:
```diff
- authenticated: { modelId: "z-ai/glm-4.5-air:free" },
+ authenticated: { modelId: "z-ai/glm-4.5-air" },
```
`ALLOWED_MODEL_IDS` is derived from the matrix, so the new slug is
picked up automatically.
## Notes
- This was the only `:free` slug in the matrix; nothing else is exposed
to the same break.
- This moves a previously-free path onto paid inference (cheap model,
but no longer $0).
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Switch the dumb/slow/authenticated path to the paid `z-ai/glm-4.5-air`
slug, replacing the retired `:free` variant to stop “This model is
unavailable for free” errors. Restores successful requests after
OpenRouter ended the free tier.
- **Bug Fixes**
- Updated model matrix in apps/backend/src/lib/ai/models.ts to use
`z-ai/glm-4.5-air`; `ALLOWED_MODEL_IDS` picks up the new slug
automatically.
<sup>Written for commit c354b5fcf2.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1685?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Updated the model selection for authenticated requests in one
quality/speed setting, so the app now uses the correct model ID and
accepts the updated option in allowed model lists.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
DB migration compat / Back-compat — Current branch migrations with ${{ needs.check-migrations-changed.outputs.base_branch }} branch code (push) Has been cancelled
DB migration compat / Forward-compat — Current branch code with ${{ needs.check-migrations-changed.outputs.base_branch }} branch migrations (push) Has been cancelled
## Summary
Replace `parseHexclaveConfigFileContent` /
`evaluateStaticConfigExpression` (Babel AST walker) with
`evalConfigFileContent` using `jiti.evalModule()`. Move
`renderConfigFileContent` from `hexclave-config-file.ts` →
`config-rendering.ts`.
Added `jiti` dep to `@hexclave/shared` (already used in shared-backend,
dashboard, backend, cli).
Link to Devin session:
https://app.devin.ai/sessions/cb098b1fb62b4dfeaf3324bc2e1377f1
Requested by: @mantrakp04
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Migrates trusted config evaluation to `jiti` and moves GitHub config
edits to a server‑side repo agent running in a Vercel Sandbox with an
apply → review → commit flow. Adds run tracking, safer defaults, and a
dashboard diff review with clear, user‑facing errors.
- **New Features**
- Two‑phase flow and endpoints: POST `/internal/config/github/apply`,
`.../commit`, `.../cancel`, plus GET `.../run`; each run tracked by
`run_id` in `ConfigAgentRun` (status, stage, progress, diff, base
commit, sandbox id). Run ids validated as UUIDs.
- Repo agent runs in a fresh sandboxed clone; warm‑boot via base
snapshot (`apps/backend/scripts/config-agent/build-image.ts`,
`HEXCLAVE_CONFIG_AGENT_BASE_SNAPSHOT_ID`). Captures a unified diff and
base commit, stops the sandbox at review, then rebuilds files from the
stored diff on commit. Returns `commitSha`, uses a safe conflict error,
and strips OAuth tokens from git remotes.
- Dashboard: non‑dismissible progress and diff preview using
`@pierre/diffs` with a cross‑tab run watcher; blocks conflicting edits
and supports cancel/commit review flow. Adds an RDE “apply” path with
progress UI.
- AI proxy defaults to `/api/latest/integrations/ai-proxy` (production
passthrough via `PRODUCTION_AI_PROXY_BASE_URL`); adds
`anthropic/claude-haiku-4.5`.
- **Refactors and Fixes**
- Trusted eval via `@hexclave/shared` `config-eval` using `jiti`;
browser‑safe parsing for untrusted GitHub content; rendering remains in
`config-rendering`. Clear separation of Node‑only code into
`config-eval`.
- Shared agent/updater logic moved to `@hexclave/shared-backend`;
removed deterministic fast path so all writes go through the agent to
preserve authoring. CLI and emulator updated to use `config-eval`.
- Defaults/renames: config file `hexclave.config.ts` (CLI `config pull`
defaults to this path), workflow `hexclave-config-sync.yml`; env
prefixes standardized to `HEXCLAVE_*`.
- Integrity and UX: commit advancement gated to the current linked
repo/branch; cancel clears any captured diff; elapsed timer handles late
starts and the not‑started sentinel; loader vs invalid config export
errors separated for accurate messaging.
- Onboarding and seeds: wizard now uses environment‑based OAuth provider
setup with updated tests; corrected GitHub owner in dummy project
seeding.
<sup>Written for commit 6cf0e899a0.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1661?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Refactor**
* Improved configuration file parsing/validation by evaluating config
modules, supporting both string and object-based `config` exports and
ensuring the expected `config` export is present.
* Updated config rendering and import-package detection to consistently
generate the `config` export and handle legacy package entrypoints.
* Tightened handling of non-statically-resolvable forms during update
flows.
* **Tests**
* Updated and extended config parsing/validation tests to reflect the
new evaluation behavior and edge cases.
* **Chores**
* Added a Jiti-based dependency to support runtime evaluation.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: mantra <mantra@stack-auth.com>
## Summary
In `UserDialog`, duplicate-email errors (`UserWithEmailAlreadyExists`,
`ContactChannelAlreadyUsedForAuthBySomeoneElse`) were shown via
`window.alert()`. Replaced with a controlled `DesignDialog` using
`WarningCircleIcon`, so the error appears as a styled modal instead of a
native browser popup.
Link to Devin session:
https://app.devin.ai/sessions/99d73091c47a4a58b33d8724df5a7ce8
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Replaced `window.alert()` in the user create/edit flow with a controlled
`DesignDialog` for duplicate-email errors (`UserWithEmailAlreadyExists`,
`ContactChannelAlreadyUsedForAuthBySomeoneElse`). Uses
`WarningCircleIcon` with an OK action, preserves form data via
`prevent-close-and-prevent-reset`, clears dialog state when the form
closes, and fixes a max-statements-per-line lint warning.
<sup>Written for commit 274fcb8e9b.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1680?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
---------
Co-authored-by: vedanta.gawande <vedanta.gawande@gmail.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: Armaan Jain <84474476+Developing-Gamer@users.noreply.github.com>
## Summary
Two changes to the user data export dialog on the project Users page:
1. The export scope now defaults to **"Export only filtered/searched
users"** instead of "all users".
2. The all-users option label is now **"Export all users in the project
(includes Anonymous)"**.
To keep this scoped to the Users table (the shared export dialog is
reused by other tables), the dialog's default scope is made configurable
rather than changed globally:
- `DataGridExportOptions` gains `defaultScope?: DataGridExportScope`
(defaults to `"all"`).
- The dialog initializes `useState(exportOptions?.defaultScope ??
"all")`.
- `user-table.tsx` passes `defaultScope: "filtered"` and the updated
`allScopeLabel`.
Other tables (teams, transactions, emails) are unaffected — they keep
the `"all"` default.
Link to Devin session:
https://app.devin.ai/sessions/4996678b2b944090b6eef2f64f0a62a1
<!-- This is an auto-generated description by cubic. -->
---
## Summary by cubic
Default the Users export dialog to filtered scope and clarify that the
"all users" option includes Anonymous; scope resets to the per-table
default only when the dialog reopens, and other tables keep "all".
- **New Features**
- Added `defaultScope` to export options and initialized scope from it;
Users table sets `defaultScope: "filtered"` and updates the all-users
label.
- Reset scope to `defaultScope` only on a closed→open transition to
avoid changing it while the dialog is open.
- **Bug Fixes**
- Stubbed `NODE_ENV` via `vi.stubEnv` in
`apps/backend/src/oauth/ssrf-protection.test.ts` to fix lint and prevent
env mutation.
<sup>Written for commit 3aa670b6d2.
Summary will update on new commits.</sup>
<a
href="https://cubic.dev/pr/hexclave/hexclave/pull/1679?utm_source=github"
target="_blank" rel="noopener noreferrer"
data-no-image-dialog="true"><picture><source
media="(prefers-color-scheme: dark)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source
media="(prefers-color-scheme: light)"
srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img
alt="Review in cubic"
src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"></picture></a>
<!-- End of auto-generated description by cubic. -->
---------
Co-authored-by: vedanta.gawande <vedanta.gawande@gmail.com>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
