The tab queried the functions log source with a worker_name filter that source
does not have. It now points at the CLI until the workers logs endpoint ships.
## Context
As per PR title: recently updated list just comprises of notebooks and
chats
Note: known API issue that save a notebook doesn't update its
`updated_at` value, so you'll notice that if you save a notebook it
doesn't move up the recently updated list
<img width="281" height="270" alt="image"
src="https://github.com/user-attachments/assets/f637a593-09a7-4df4-85b7-43637f04738d"
/>
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added a home view displaying recently updated notebooks and chats.
* Recent items are sorted by update time and limited to five entries.
* Added relative timestamps in minutes, hours, or days.
* Chat entries open directly, while notebook entries link to their
explorer view.
* Added an empty state when no recent items are available.
* **Bug Fixes**
* Excluded unsupported chat types from recent-item results.
* Improved handling of unavailable timestamps.
* **Tests**
* Added coverage for sorting, filtering, limits, and relative-time
formatting.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
- Added ~11 new eval cases for Notebooks AI assistant evals, including:
basic notebook cell creation, multi-cell composition
(markdown+database+log), log cell time ranges, row_limit defaults, chart
config, destructive SQL safety warnings, and hallucination guards for
nonexistent tables
- Extended `toolUsageScorer` with deterministic `forbiddenTools` field
to score both required and forbidden tool usage, enabling eval cases to
assert tool choice (e.g., `execute_sql` vs `create_notebook`) without
LLM-as-judge
- Extended SQL validators
(`sqlSyntaxScorer`/`sqlIdentifierQuotingScorer`) to validate SQL inside
`create_notebook` database cells (log cells deliberately excluded as
they use ClickHouse dialect)
- Fixed two real assistant issues in `NOTEBOOKS_PROMPT`: (a) reuse
`CLICKHOUSE_LOGS_COMPLETION_INSTRUCTIONS` and schema section to prevent
incorrect BigQuery-style SQL in logs queries, (b) require schema
verification before writing `database_cell` to prevent queries against
nonexistent tables
Resolves FE-4087
## Test plan
- All 11 new eval cases run live against OpenAI via Braintrust; traces
inspected and validated
- Existing unit tests, lint, and typecheck pass
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added support for creating notebooks with database, Markdown, chart,
and log cells.
* Improved handling of reusable notebook requests versus one-off SQL
queries.
* Added guidance for modern ClickHouse SQL and absolute log time ranges.
* **Bug Fixes**
* Improved SQL validation, row-limit enforcement, and destructive-query
safety.
* Prevented invalid or nonexistent-table queries from being accepted.
* Improved validation of notebook cell types and tool usage.
* Improved handling of ClickHouse log queries and database-cell SQL.
* Improved evaluation reliability by limiting concurrent test execution.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
The `logs.all` Management API endpoint runs BigQuery SQL and is being
retired next month. The Platforms guide was the only hand-written doc
still pointing at it.
Repoints the debugging example at `GET
/v1/projects/{ref}/analytics/endpoints/logs`, which serves the same data
as a single `logs` table keyed by `source`, with structured fields in
the `log_attributes` map, and converts the query to the ClickHouse
dialect.
Verified by running the example's exact SQL and curl shape against a
real project on the OTEL logs endpoint: 100 rows, with `status_code` and
`path` populated.
The generated API specs still list `logs.all`; those regenerate from the
platform side.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Documentation**
- Updated the Supabase for Platforms integration guide’s
debugging-projects example.
- Revised the example to use the analytics logs endpoint and unified
logs table.
- Added ClickHouse SQL filtering for edge logs, structured log
attributes, and HTTP errors.
- Improved the example’s alignment with current log query and analytics
capabilities.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What
The Workers list page at `/project/[ref]/workers`, behind
`useFlag('workers')`. Reads `GET /v2/projects/{ref}/workers`.
- Sidebar and command-menu entries, both hidden when the flag is off
- Name search, state and access filters, pagination
- Read-only
Gating, in order: flag off redirects to the project home; a 404 from the
API means the project is outside the alpha allow-list ("not enabled for
this project"); a 403 means the caller lacks the permission
(`NoPermission`); anything else is an `AlertError`.
`parseWorker` in `data/workers/workers.utils.ts` is the only place the
API shape becomes the view model. It validates with zod, so a drifted
response fails the query instead of half-rendering a row.
## How to test
Only on the **Mockamaster** project in staging — it is the one project
in the alpha allow-list, and standing a worker up anywhere else is
involved right now.
1. Staging dashboard → Mockamaster → **Compute** in the sidebar
2. Expect the `dashboard-test` worker: state `Active`, runtime Deno,
private, US West, 2 GB · 1 vCPU · 1 inst
3. Open any other project's `/workers` URL → "Compute is not enabled for
this project"
4. Turn the `workers` flag off → the sidebar entry disappears and the
URL redirects to the project home
Closes FE-4188
The 47 BigQuery-era logs queries across these 20 pages error on the
ClickHouse-backed logs engine ("Backend error! Retry your query."). This
converts them per the rules in `apps/studio/lib/ai/clickhouse-logs.ts`
and repoints every Logs Explorer link at the SQL Editor with the query
source set to **Logs**, since the Logs Explorer is being retired. Also
fixes two stale PostgreSQL 12 links in the tables guide.
Each of the 14 prefilled links was verified to decode back to exactly
the SQL shown on its page. One caveat for review:
`response.headers.proxy_status` in `postgrest-error-codes.mdx` is
unverified — it isn't in the published field reference, and the test
project had no `edge_logs` traffic to confirm against.
Fixes DOCS-1331
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **Documentation**
- Updated database, storage, API, and Edge Function logging guides to
use the SQL Editor and current Logs interface.
- Replaced legacy Log Explorer and BigQuery examples with current query
syntax and structured log fields.
- Refreshed troubleshooting queries for error diagnosis, filtering,
aggregation, and performance analysis.
- Improved examples with clearer source filters, status handling,
request details, joins, and result limits.
- Updated PostgreSQL documentation links and clarified how API error
codes appear in responses.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Jordi Enric <jordi.err@gmail.com>
Currently, sessions timeouts and reuse interval inputs accepted any
values. This PR caps:
- absolute session timeout to 1 year
- inactivity timeout to 1 year
- refresh token reuse interval to 300 seconds
Since these maximums are introduced _after_ some projects have values
that exceed the new limits, we allow the users to save the form if their
values exceed the max but are unchanged. However, if they decide to
change the value, it must fit within the limits.
<img width="1195" height="402" alt="Screenshot 2026-08-20 at 15 45 49"
src="https://github.com/user-attachments/assets/192420e8-4878-4e4b-9d82-0d1cc4074728"
/>
<img width="1194" height="512" alt="Screenshot 2026-08-20 at 15 46 06"
src="https://github.com/user-attachments/assets/bb333c54-daa3-46ac-b144-96263428f4d7"
/>
<img width="1168" height="323" alt="Screenshot 2026-08-20 at 15 46 35"
src="https://github.com/user-attachments/assets/ae38fcf6-bc73-453f-a61b-2d6f2d0ecfee"
/>
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Improvements**
* Added clear maximum-value guidance for session and refresh-token
settings.
* Existing projects with previously configured values above new limits
can retain those values while making unrelated changes.
* Removed session-related settings from the protection authentication
form.
* **Bug Fixes**
* Improved validation for session timeouts, JWT expiration, and
refresh-token reuse intervals.
* Added clearer validation messages and support for reducing previously
over-limit values.
* **Tests**
* Expanded coverage for boundary values, invalid inputs, saved settings,
and submitted configuration updates.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Context
API changes are ready so hooking up the endpoints for full CRUD UX E2E
- Can create notebooks
- Can load notebooks
- Can delete notebooks
- Can update notebooks
Informational banner for the `logs.all` Management API removal on Sept
23, in the Logs and Observability sections.
* Untargeted. Whether a project calls the endpoint is behaviour that no
API response carries, so precise targeting needs a mgmt-api change (we
aimed for speed and less complexity here). Copy is informational rather
than "action required" since most viewers won't be affected.
* Uses `BannerStack` (bottom-right card) rather than the top header
banner, at priority 4 so it renders as the front card. Note this pushes
`database-connections-banner` (p2) and `index-advisor-banner` (p3) into
peek slivers on Observability.
* Short Notice card: title, one line of copy with `logs.all` inline, and
a Learn more link to the changelog.
* Waits for localStorage before showing, and BannerStack ignores stale
dismiss timers when a banner is revived (avoids flash-then-disappear on
refresh).
* Dismiss is browser-level; self-expires Sept 24 via
`LogsAllDeprecationExpiry`.
* Cleanup tracked in GROWTH-1104.
* Tested in staging.
Check in:
- /project/_/logs (unified logs)
- /project/_/logs/explorer
- /project/_/observability
| After |
| --- |
| <img width="626" height="528" alt="CleanShot 2026-08-20 at 12 29
49@2x"
src="https://github.com/user-attachments/assets/2044966d-bc83-4f88-ac75-1b8ff80be08d"
/> | <img width="622" height="440" alt="CleanShot 2026-08-20 at 12 28
33@2x"
src="https://github.com/user-attachments/assets/1dc768cd-837c-4a5a-a3fa-7b6986fe5876"
/> |
Resolves GROWTH-1093.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
## New Features
* Added a dismissible notice about the `logs.all` endpoint retirement on
September 23, 2026.
* The notice appears on relevant Logs and Observability pages with
streamlined migration guidance.
* Clarified that dashboard logs remain unchanged.
* Dismissal preferences are saved, and notices remain visible or are
removed reliably during navigation.
## Telemetry
* Added tracking for notice display and dismissal interactions.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Danny White <3104761+dnywh@users.noreply.github.com>
Co-authored-by: Cursor Agent <cursoragent@cursor.com>
Co-authored-by: Danny White <dnywh@users.noreply.github.com>
Two fixes for the [Prisma
guide](https://supabase.com/docs/guides/database/prisma):
- `StepHikeCompact.Code` marked its whole subtree `not-prose`, so the
labels and admonitions that steps interleave with their code samples
rendered at 16px with zero margins, flush against the samples and tab
bars. Dropping `not-prose` restores body typography and spacing;
back-to-back samples now get a gap too, since they have no prose between
them.
- The guide's three outer tab groups omitted `type`, so they fell back
to pill styling — the only pills among 395 `<Tabs>` in the content tree.
Fixes DOCS-1327
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Style**
* Improved spacing and prose behavior for code samples in the
documentation.
* Preserved existing code margin customizations.
* Updated Prisma guide tabs with a consistent compact, underlined
appearance.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Bug fix
## What is the current behavior?
In the SQL Editor, when two snippets are both named "Untitled query" and
one is renamed, the rename modal's state is not reset afterwards.
Opening the rename modal for the second snippet prefills the input with
the first snippet's new name, and the second snippet can't be renamed at
all because the "Rename query" button stays disabled.
`RenameQueryModal` fed the snippet to react-hook-form through the
`values` option, which only re-runs its reset when the values object
deep-changes. Two snippets with the same name (and no description)
produce a deep-equal object, so switching between them never resets the
form — it keeps the previously renamed name and stays non-dirty.
## What is the new behavior?
The form is mounted per snippet (`key={snippet.id}`) with plain
`defaultValues`, so no form state can carry over between snippets
regardless of name collisions. `SQLEditorNav` derives modal visibility
from the selected snippet and clears it on cancel/complete, matching
`SearchList`.
Covered by a new component test in `RenameQueryModal.test.tsx` that
renames one "Untitled query", reopens the modal for a second one, and
asserts the field resets and the second rename submits.
## Additional context
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Fixed the rename dialog retaining input from a previously renamed
snippet.
* Ensured the rename form resets correctly after successful submission
and when switching between snippets.
* **Tests**
* Added regression coverage for renaming multiple untitled snippets with
the same original name.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Changes to the audit logs UI
## What is the current behavior?
Partner related fields in the audit logs are not shown
## What is the new behavior?
- Shows all partner related fields in the audit logs
- Also uses the new fields to compute the user name
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Audit log entries now display partner names, installation IDs, user
emails, and user IDs when available.
* Partner identity and email are shown when standard actor details are
unavailable.
* Partner names are consistently formatted for clearer display.
* Entries without partner information continue to display cleanly
without blank or confusing actor details.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
<img width="1510" height="860" alt="image"
src="https://github.com/user-attachments/assets/36a748b7-bdeb-4685-8bb1-da911711874b"
/>
Introduces a new OAuth consent block in preparation for offering more
MCP focused blocks that require authentication and consent. The general
approach for this is to decouple consent block from authentication block
but provide guidance on how to use both. The alternative is to add auth
as a dependency to consent but apps may already have their own
authentication UI / flows.
The block is also positioned as a general OAuth Consent vs MCP Consent
as it can be put to use for other use cases outside of MCP on projects
who want to make use of the OAuth 2.1 Server offering.
A couple of changes outside of the block itself were required:
- Updated the Auth blocks to allow for a `next` param to redirect users
to after signing in
- Updated middleware so next param is correctly passed through to sign
in
## How to test
Requires Docker and a Supabase CLI recent enough to support
`[auth.oauth_server]` (verified on 2.109.0 / GoTrue v2.192.0).
### 1. Local Supabase with the OAuth server enabled
In your `supabase/config.toml`, edit the existing `[auth.oauth_server]`
section — `supabase init` already writes one, and adding a second fails
with `table oauth_server already exists`:
```toml
[auth.oauth_server]
enabled = true
authorization_url_path = "/oauth/consent"
allow_dynamic_registration = true
```
Set `site_url` to wherever your test app runs (e.g.
`http://localhost:3100`), then `supabase start`. Grab the API URL and
publishable key from `supabase status`.
### 2. A consumer app with the blocks installed
The consent block ships no login route by design, so pair it with an
auth block:
```bash
npx create-next-app@latest consent-test --ts --tailwind --app --yes
```
```bash
cd consent-test && npx shadcn@latest init -d -y && npx shadcn@latest add https://supabase.com/library/r/password-based-auth-nextjs.jsonhttps://supabase.com/library/r/oauth-consent-nextjs.json
```
To test this branch before it deploys, run `pnpm --filter ui-library
dev` and use `http://localhost:3004/library/r/...` instead. If you
changed anything under `registry/default/blocks/oauth-consent/**`, run
`pnpm --filter ui-library build:registry` first — shadcn fetches the
generated `public/r/*.json`, not the source.
Put `NEXT_PUBLIC_SUPABASE_URL` and
`NEXT_PUBLIC_SUPABASE_PUBLISHABLE_KEY` in `.env.local` and start the app
on the port you set as `site_url`.
### 3. Register an OAuth client and start a real authorization request
```bash
curl -s -X POST http://127.0.0.1:54321/auth/v1/oauth/clients/register -H "Content-Type: application/json" -d '{"client_name":"Test Client","redirect_uris":["http://localhost:3100/callback"],"grant_types":["authorization_code"],"response_types":["code"],"scope":"openid profile email"}'
```
Then open the authorize URL in a browser (not curl — you need the
redirect chain and cookies):
```
http://127.0.0.1:54321/auth/v1/oauth/authorize?client_id=<id>&response_type=code&redirect_uri=http://localhost:3100/callback&scope=openid+profile+email&state=xyz&code_challenge=<challenge>&code_challenge_method=S256
```
Auth mints the `authorization_id` and redirects to
`<site_url>/oauth/consent?authorization_id=…`. An MCP client pointed at
your app is an even better driver, since that's the real consumer shape.
### 4. Cases to walk
| Case | Expected |
| --- | --- |
| Signed out, hit the authorize URL | Lands on
`/auth/login?next=%2Foauth%2Fconsent%3Fauthorization_id%3D…`; after
login, returns to the consent screen |
| Consent screen | Shows client name, redirect URI, signed-in email, and
requested scopes from `getAuthorizationDetails` |
| Allow access | Redirects to `redirect_uri` with `code` and your
original `state`; the code exchanges at `/oauth/token` for a real access
token |
| Deny | Redirects with `error=access_denied` and your `state` |
| Re-run the same authorize URL after approving | Skips the screen,
straight to callback with a new code |
| Visit `/oauth/consent` with no `authorization_id` | "This page needs
an authorization_id" |
| Stale or bogus `authorization_id` | Error shown, buttons still usable
|
| Double-click Allow | Exactly one `POST
/oauth/authorizations/<id>/consent` |
Test the react, react-router, or tanstack variant the same way if you're
touching them — the hook is duplicated per framework, so a fix in one
doesn't carry.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added an OAuth 2.1 consent experience with client details, requested
scopes, redirect URI, and approve/deny actions.
* Added OAuth consent examples and registry blocks for Next.js, React,
React Router, and TanStack Start.
* Added OAuth documentation, navigation, and framework support across
the UI library.
* **Bug Fixes**
* Login flows now safely preserve valid same-origin redirect
destinations while rejecting unsafe URLs.
* OAuth routes can handle consent flows before authentication and
redirect safely to sign-in.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
<img width="840" height="507" alt="image"
src="https://github.com/user-attachments/assets/d0f4667f-a7bb-4afe-95b2-a9e224adcbb5"
/>
<img width="848" height="597" alt="image"
src="https://github.com/user-attachments/assets/beb0c239-d36c-4103-ab07-8a3872ba3f30"
/>
Updates how we display Notebooks in Assistant to be more in line with
our AssistantQueryCell.
## To test:
- Open Assistant and ask it to create a test notebook and note the new
styling
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added unified, read-only notebook previews for create and update
proposals.
* Preview cells now support expandable content, clearer type icons,
metadata, and “Show more” controls.
* Added before-and-after metadata comparisons for replaced cells.
* Integrated previews into confirmation cards with approval, skip, and
refresh actions.
* Added skip-only confirmation flows when approval is unavailable.
* **Bug Fixes**
* Improved handling of parse failures, stale notebooks, invalid changes,
and loading errors.
* **Style**
* Refined confirmation card layouts, borders, spacing, and footer
presentation.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## What kind of change does this PR introduce?
Chore. Stops committing generated design-system registry output
([DEPR-647](https://linear.app/supabase/issue/DEPR-647/generate-and-ignore-design-system-registry-output)).
## What is the current behavior?
`apps/design-system/__registry__` is build output from `registry/`, but
the chart snapshots and index are committed. That makes reviews noisy,
and a forgotten `build:registry` leaves `master` out of date until
someone else regenerates it.
## What is the new behavior?
`pnpm dev` and `pnpm typecheck` generate `__registry__` automatically.
The directory is gitignored, and the previously tracked snapshots are
removed. `pnpm build` still generates it as before.
## To test
- From the repo root, run `pnpm --filter=design-system
generate:registry` and confirm
`apps/design-system/__registry__/index.tsx` is created locally and is
untracked.
- Run `pnpm dev:design-system`, open
[http://localhost:3003](http://localhost:3003), and open any component
docs page with a live preview (for example Charts). Previews and source
panels should still load.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Documentation**
* Clarified when the component registry is generated and how to
regenerate it.
* Expanded component documentation guidance, including content sources
and generated-file editing restrictions.
* **Chores**
* Improved registry generation across development, type checking,
builds, and cleanup.
* Generated registry files are now excluded from version control and
linting.
* Improved reliability when creating and refreshing generated registry
files.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
11 of the generated `MD_PAGES` entries are blog slugs whose HTML pages
308-redirect away via `apps/www/lib/redirects.js` before any `<head>`
renders. They can never carry an alternate tag and Accept negotiation
never fires (Next.js `redirects()` runs before middleware), so their
`.md` siblings are orphaned content reachable only by guessing the
suffixed URL. Six of them duplicate live, correctly-tagged pages
(`/customers/*`, `/pricing`).
**Changed:**
- `generateMdContent.mjs` derives an exclusion set from
`lib/redirects.js` at generation time: any unconditional exact-match
redirect source (wildcard/param patterns and conditional `has`/`missing`
redirects are skipped) drops the matching slug from both `MD_CONTENT`
and `MD_PAGES`. The build log names every excluded slug, currently the
11 known ones.
- Self-maintaining by design (per the decision recorded on the issue): a
future redirected post auto-excludes on the next build, and removing a
redirect brings its `.md` sibling back. The MDX sources stay in the
repo; nothing is deleted.
- Effect on the 11 slugs: alternate tags stay absent (nothing rendered
them anyway), and explicit `.md` URLs go from serving orphaned markdown
to 404, the same external effect deletion would have had.
## To test
Tested locally:
- [x] `node scripts/generateMdContent.mjs` logs `🚫 Excluded 11
redirected slugs: ...` naming exactly the 11 known slugs; output drops
483 → 472 pages
- [x] Generated file carries no MD_CONTENT/MD_PAGES key for any excluded
slug (raw URL mentions inside other posts' bodies remain, as expected)
- [x] `apps/www` vitest: 71/71 (GROWTH-1013 drift tests unaffected)
Post-merge:
- [ ] `https://supabase.com/blog/case-study-xendit.md` returns 404
(previously 200 orphaned markdown); `https://supabase.com/pricing.md`
still 200
## Linear
- fixes GROWTH-1022
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Excluded content with valid exact-path redirects from generated
documentation.
* Preserved content associated with conditional or wildcard redirects.
* Updated generated page counts and output statistics to reflect the
filtered content.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Context
Adds the inline AI completion functionality into Explorer QueryEditor,
similar to what we've got for the existing SQL editor
- Shifts the `ResizableAIWidget` and `InlineWidget` components out of
the SQL Editor folder into `components/ui/AiEditor` to be used by both
SQL Editor and Query Editor
- Consolidates the "proposal" logic that was initially set up for the
Clickhouse Migration functionality with this Inline AI stuff
- Also added the prompt into the proposal header (Refer to the
screenshots below)
- SQL Editor didn't have this - but figured this is useful as context
for the user
<img width="935" height="352" alt="image"
src="https://github.com/user-attachments/assets/3f71539a-dda1-4763-be08-a850bdc8aec6"
/>
Source selected: Database
<img width="922" height="357" alt="image"
src="https://github.com/user-attachments/assets/81e772e6-dcc9-440d-83db-49d0d487dd13"
/>
Source selected: Logs
<img width="920" height="345" alt="image"
src="https://github.com/user-attachments/assets/83f1dae3-cf62-4424-be42-b06e70cb366d"
/>
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **New Features**
- Added AI-assisted SQL generation with contextual prompts and
OS-specific guidance.
- Review generated SQL changes in a diff, then accept, reject, or cancel
suggestions.
- Added inline, resizable AI prompt controls with loading and submission
states.
- Added the Ctrl/Cmd+Shift+K shortcut to run AI SQL generation.
- **Improvements**
- Added options to disable query execution and run custom actions.
- Renamed “Recent” to “Recently updated.”
- Improved editor widget positioning and display behavior.
- Added clearer error notifications when AI generation fails.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Context
Just realised that chats in the new Explorer UI have no delete
functionality so this patches it
<img width="296" height="184" alt="image"
src="https://github.com/user-attachments/assets/90a79b6b-55a8-4122-8cd8-05fc13e9f4a5"
/>
Also added a confirmation modal for deletion
<img width="473" height="266" alt="image"
src="https://github.com/user-attachments/assets/56a1e400-2a1c-48b7-b6b8-0104af49b1a9"
/>
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
- **New Features**
- Added an option to delete Explorer chats from the chat toolbar.
- Added a confirmation prompt before permanently deleting chat history.
- Added success feedback after deletion is completed.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
## Summary
- Regenerates `packages/api-types` for the content endpoints now that
the Platform API's `notebook` content type has landed (list/get/upsert
`type` enums, plus `UpsertContentBody`'s notebook cell shape with
`_id`/`y_series`). Unrelated schema drift from the same regen
(Warehouse, SSO, notification exceptions, etc.) is excluded — only the
content-endpoint hunks are applied.
- Removes every local widening cast added while the API support was
pending (`content-query.ts`, `content-infinite-query.ts`,
`notebook-query.ts`, `notebook-upsert-mutation.ts`,
`sql-folders-query.ts`).
- What remains is scoped and renamed to match: draft ids
(`generateDraftId`/`isDraftId`), used only for cells created client-side
in the editor before their first save, dropped before they'd ever reach
the backend as a fake `_id`.
## Test plan
- [x] `pnpm typecheck` — clean
- [x] `pnpm --filter studio test` — full suite passes (518 files / 5471
tests)
- [x] `pnpm --filter studio run lint:ratchet` — no new warnings
- [x] `pnpm format` / prettier — clean
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Bug Fixes**
* Improved notebook cell tracking during editing, reordering, insertion,
and deletion.
* Preserved existing cell identifiers while removing temporary draft
identifiers before saving.
* Improved chart configuration for selecting and displaying multiple
Y-axis series.
* Strengthened notebook validation and content persistence behavior.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Joshen Lim <joshenlimek@gmail.com>
## Context
Resolves FE-4209
Client crash occurs when re-ordering a QueryCell in the new explorer UI
with the error "InstantiationService has been disposed"
Investigated this with Claude which eluded that it's a bug that's within
the Monaco package which `4.8.0-rc.3` actually patched hence opting to
upgrade the package. Verified that monaco still functions as expected +
re-ordering query cells in the explorer UI no longer crashes
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Chores**
* Updated the Monaco Editor integration to release candidate version
4.8.0-rc.3.
* No visible end-user functionality changes.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
<img width="1512" height="861" alt="image"
src="https://github.com/user-attachments/assets/404c9a27-dc10-497e-a5ec-003cd4b9705a"
/>
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Feature.
## What is the current behavior?
Assistant `execute_sql` tool parts and markdown SQL fences render
through `DisplayBlockRenderer`. The confirm footer is gated to the last
part of the last message, so a pending SQL approval can disappear if the
assistant keeps writing.
## What is the new behavior?
SQL tool parts and markdown fences use `AssistantQueryCell` inside
`Confirm`. The footer follows the same manual-approval helpers as Edge
Functions. `DisplayBlockRenderer` is removed.
## Additional context
Top of stack #49171. Base: `feat/assistant-query-cell` (#49169).
Does not wrap notebook create/update proposals. That depends on
[#49159](https://github.com/supabase/supabase/pull/49159) merging first.
## Test plan
- [ ] `execute_sql` approval shows Run query / Skip on the Confirm card
under the editor
- [ ] Footer still shows if the assistant writes text after the SQL tool
part
- [ ] Markdown SQL fences render as AssistantQueryCell without a confirm
footer
- [ ] After skip, the query cell remains so the user can run it locally
- [ ] Edge Function confirm from #49168 still works
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Feature.
## What is the current behavior?
Notebooks and query tabs use `QueryEditor`. Assistant SQL still uses
`DisplayBlockRenderer` / `QueryBlock`.
## What is the new behavior?
Adds `AssistantQueryCell`, a local-state wrapper around the shared
`QueryEditor` (`variant="viewport"`, `isRunDisabled` while confirming).
Nothing is wired into the conversation yet — that is #49170 — so this PR
is the reusable cell plus the small editor/report-container hooks it
needs.
## Additional context
Part of stack #49171. Base: `feat/assistant-confirm` (#49168).
## Test plan
- [ ] `AssistantQueryCell.utils.test.ts` passes
- [ ] Query editor still runs in Explorer notebooks / query tabs
- [ ] No assistant conversation UI change in this PR (still
DisplayBlockRenderer)
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
<img width="1512" height="862" alt="image"
src="https://github.com/user-attachments/assets/79a6d4dc-dcd2-489f-97d7-3ee7a0196b7d"
/>
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Feature / UI refactor.
## What is the current behavior?
Assistant Edge Function approval nests `ConfirmFooter` under the
function block. `addToolApprovalResponse` is wired whenever state is
`approval-requested`, including automatic approvals.
## What is the new behavior?
Introduces a `Confirm` card that owns the frame, with the footer
attached below the body. Edge Function approval uses that card.
Interactive Approve/Deny only runs for manual `approval-requested` parts
(`!approval.isAutomatic`), matching the [AI SDK tool-approvals `useChat`
guidelines](https://ai-sdk.dev/docs/agents/tool-approvals).
SQL still uses `DisplayBlockRenderer` until #49170. `ConfirmFooter` is
inlined into `Confirm` so SQL can keep importing the named footer until
that PR.
## Additional context
Part of stack #49171. Base: `chore/ai-sdk-7` (#49167).
Notebook proposal Confirm wrapping is **not** in this stack — that file
lives on [#49159](https://github.com/supabase/supabase/pull/49159).
Follow up after that stack merges.
## Test plan
- [ ] Deploy-edge-function tool part shows Confirm with Skip / Deploy
- [ ] Existing-function replace warning still requires the second
confirm
- [ ] After approve, footer morphs to loading and buttons disable
- [ ] `Confirm.utils.test.ts` and `EdgeFunctionRenderer.test.tsx` pass
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added confirmation cards for AI-assisted actions, including approve
and cancel controls.
* Improved handling of manual approval requests for SQL execution,
notebook changes, and Edge Function deployment.
* Added support for customizing report and Edge Function block styling.
* **Bug Fixes**
* Automatic approvals no longer appear as pending manual confirmations.
* Skipped SQL actions now provide clearer messaging.
* **Tests**
* Expanded coverage for approval states, confirmation controls, and
automatic decisions.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
<!-- ccr-slack-attribution -->
_Requested by **Sean Oliver** · [Slack
thread](https://supabase.slack.com/archives/C07P3AU3J2D/p1787036390117589?thread_ts=1787036390.117589&cid=C07P3AU3J2D)_
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Chore. One entry added to the www sitemap generator.
## What is the current behavior?
`https://supabase.com/evals` is missing from `sitemap_www.xml`, so
search crawlers are never told the page exists. `robots.txt` doesn't
block it, they just have no way to find it from the sitemap.
The reason is that `/evals` is served by a separate Vercel project and
only reaches supabase.com through a proxy rewrite in
`apps/www/lib/rewrites.js`:
```js
{
source: '/evals',
destination: 'https://supabase-evals.vercel.app',
},
```
`apps/www/internals/generate-sitemap.mjs` builds its URL list by
globbing local route source files (`pages/**`, `_blog/*.mdx`,
prerendered `.next/server/pages/**`, etc.) and never resolves rewrites.
There is no page file behind `/evals`, so the globs can't discover it.
Closes GROWTH-1113.
## What is the new behavior?
`https://supabase.com/evals` appears once in the generated
`sitemap_www.xml`, with the same `<changefreq>weekly</changefreq>` and
`<priority>0.5</priority>` as every other entry in the file (no entry in
this sitemap carries a `<lastmod>`).
The entry is a small named const spread into the final `urlset` join,
next to `changelogDetailUrls` — the existing precedent in this file for
URLs with no page file behind them. Nothing else in the script changed,
and the sitemap index output (`sitemap.xml`) is byte-identical.
```diff
+ // /evals is a separate app proxied onto supabase.com via a rewrite in lib/rewrites.js,
+ // so it has no page file for the globs above to find. Hardcode it here.
+ const proxiedAppUrls = [
+ `
+ <url>
+ <loc>https://supabase.com/evals</loc>
+ <changefreq>weekly</changefreq>
+ <priority>0.5</priority>
+ </url>
+ `,
+ ]
+
const sitemap = `
<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
- ${[...staticUrls, ...changelogDetailUrls].join('')}
+ ${[...staticUrls, ...changelogDetailUrls, ...proxiedAppUrls].join('')}
</urlset>
`
```
This only makes the URL discoverable. Whether the page content itself is
crawlable is separate work, tracked in the evals repo.
## Additional context
Verification, run locally against this branch. The generator runs
standalone (`node ./internals/generate-sitemap.mjs` from `apps/www`); a
missing `.next` just means the globs match fewer pages, and the missing
changelog RSS is caught internally.
I generated `sitemap_www.xml` from `master` and from this branch and
diffed the two. The added entry is the only difference:
```
3271a3272,3277
>
> <url>
> <loc>https://supabase.com/evals</loc>
> <changefreq>weekly</changefreq>
> <priority>0.5</priority>
> </url>
```
Exactly one occurrence, with its neighbouring entry for context:
```
$ grep -c '<loc>https://supabase.com/evals</loc>' public/sitemap_www.xml
1
<url>
<loc>https://supabase.com/terms</loc>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://supabase.com/evals</loc>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
</urlset>
```
Other checks:
- Both outputs parse as well-formed XML (Python `xml.dom.minidom`):
`sitemap_www.xml` has 545 `<url>` elements, `sitemap.xml` parses OK.
- `sitemap.xml` (the sitemap index) is identical to the pre-change
output; `diff` reports no changes.
- `npx prettier --check internals/generate-sitemap.mjs` → "All matched
files use Prettier code style!"
- Both generated sitemaps are gitignored (`apps/www/.gitignore` lines
29-30), confirmed with `git check-ignore`. `git status` shows only
`apps/www/internals/generate-sitemap.mjs`, so no generated file is in
the commit.
- No test, snapshot, or fixture anywhere in the repo references the
sitemap generator, so there was nothing to run. Its only caller is
`apps/www`'s `postbuild` script.
Not run: `pnpm --filter=www build`. It fails during "Collecting page
data" on a clean `master` checkout in this environment too, so the
failure is pre-existing and unrelated, and this change needs no build to
verify.
Co-authored-by: Claude <noreply@anthropic.com>
## Problem
The guide alternated between context, procedure, and reference on almost
every heading. A reader who wanted to write a policy passed through four
context or reference sections to reach one. A reader who wanted the
model had to skip three procedures.
## Solution
- Group into three sections by information type: `Understand Row Level
Security`, `Secure a table with RLS`, and `RLS reference`, with a
navigation intro.
- Merge the four policy sections. They repeated the same setup block,
burying the clause that differed. One setup block now precedes four
short policy examples.
- Move the auto-enable recipe into `event-triggers.mdx`, whose stub
section's entire body was a link back here.
- Relocate the stranded `auth.uid()` caution into the `auth.uid()`
reference.
- Lift the revoke-and-grant procedure out of the danger admonition and
merge it with the two other places that taught `enable row level
security`.
- Point the Grafana IO chart entry at the performance guide. Its
`#rls-performance-recommendations` anchor went away when tuning split
out in #49016.
765 lines to 582. 30 headings to 25.
Headings are demoted rather than renamed wherever anything links to
them. Every inbound anchor in the repo still resolves; the only one
removed, `#auto-enable-rls-for-new-tables`, was referenced solely by the
`event-triggers.mdx` stub this PR replaces.
## Note on the history
Rebuilt from `master` after #49011, #49015, and #49016 merged. The
branch previously carried those 10 commits plus rebase churn against
them.
Rebasing naively would have reverted review feedback from #49016
(`70fa812`), which removed the benchmarks table and the "This guide"
opener from the performance guide. Those are deliberately not restored
here. The only changes to that file are two missing `await`s and a join
predicate that was a tautology while unqualified.
The three PRs stacked on this one (#49268, #49269, #49270) have been
rebased onto the new base.
## Manual testing
1. Open the [Row Level Security
guide](https://docs-git-docs-rls-restructure-supabase.vercel.app/docs/guides/database/postgres/row-level-security)
on the preview. Three top-level sections appear in the table of
contents.
2. Select each link in the intro. All three jump to their section.
3. Open [Event
triggers](https://docs-git-docs-rls-restructure-supabase.vercel.app/docs/guides/database/postgres/event-triggers).
The auto-enable section holds the full recipe instead of a link.
4. Open the [performance
guide](https://docs-git-docs-rls-restructure-supabase.vercel.app/docs/guides/database/postgres/row-level-security-performance).
No benchmarks table, and the three bullets at the top link into the RLS
guide.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Documentation**
* Reworked the Row Level Security guide with clearer guidance on grants,
policies, permissions, performance, testing, views, and secure
functions.
* Added a complete example for automatically enabling RLS on newly
created public tables.
* Improved SQL examples and clarified table references in RLS
performance guidance.
* Corrected grammar in the Grafana chart troubleshooting documentation.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
<!-- ccr-slack-attribution -->
_Requested by **Ali Waseem** · [Slack
thread](https://supabase.slack.com/archives/C063LNYJJKS/p1787146439389169)_
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Bug fix.
## What is the current behavior?
Opening `/authorize` for an OAuth app whose `name` the platform API
omitted crashed the entire page with `TypeError: Cannot read properties
of undefined (reading 'toLowerCase')`
([SUPABASE-APP-K7E](https://supabase.sentry.io/issues/7679644991/)). The
user got a full-page error instead of a consent screen, and could
neither authorize nor decline.
The same class of crash hit the project-level OAuth apps list
([SUPABASE-APP-JB1](https://supabase.sentry.io/issues/7502074939/)).
Typing in the search box called `.toLowerCase()` on `client_name` for
every app, so one app registered without a name broke search for the
whole list.
The project-claim page crashed the same way, reading the first character
of the name for the fallback avatar.
## What is the new behavior?
The trusted-partner helpers treat a missing name as "no trusted partner
matched" and return `null`. The apps filter treats a missing name or
client ID as "does not match the search string". The claim page falls
back to a placeholder initial instead of indexing into `undefined`.
The authorize page now renders normally, minus the optional
partner-impersonation caution, which cannot be evaluated without a name.
Three changes:
-
`apps/studio/components/interfaces/Organization/OAuthApps/OAuthApps.utils.ts`
— `findTrustedPartnerByName` accepts `string | null | undefined` and
returns `null` early on a falsy name; `getOAuthImpersonationWarning`'s
`name` param widened to match (its existing `if (!namedPartner) return
null` already handles the rest).
- `apps/studio/components/interfaces/Auth/OAuthApps/oauthApps.utils.ts`
— `filterOAuthApps` optional-chains `client_name` and `client_id` before
`.toLowerCase()`, defaulting each match to `false`.
-
`apps/studio/components/interfaces/Organization/ProjectClaim/confirm.tsx`
— `{requester.name?.[0] ?? '?'}` for the fallback avatar initial.
Each is a separate commit so any one can be dropped independently.
## Additional context
### Root cause, not fixed here
`apps/studio/data/api-authorization/api-authorization-query.ts:37`
returns `data as ApiAuthorizationResponse`, an unchecked cast with no
runtime validation, even though the openapi-fetch client already types
the endpoint from the generated schema. Both the generated
`GetOAuthAuthorizationResponse` and the hand-written local type declare
`name: string` as required, so this was invisible to TypeScript.
The durable fix is to derive the type from the schema and drop the cast,
which is the house pattern elsewhere in `apps/studio/data`, and to
correct the OpenAPI spec at source if the API can legitimately omit
`name`. Left out deliberately to keep this cherry-pickable.
### Not in scope
`requester.scopes` is optional in the schema but required in the local
type, and is read unguarded in several places. Defaulting it to `[]`
would tell a user an app requested no permissions on a live consent
screen, so it needs a product decision rather than a drive-by guard.
### Testing
No local checks were run. This clone has no `node_modules` and `pnpm
install` is blocked in the environment, so `npm run build`, typecheck,
lint, Prettier and tests were all left to CI. Please treat CI as the
verification for this PR.
There is also a coverage gap worth noting:
`apps/studio/tests/components/ApiAuthorization.test.tsx:48-62` hardcodes
`name: 'Test App'` in `createMockAuthResponse`, and no test omits the
field, which is why none of these crashes were caught.
---
_Generated by [Claude
Code](https://claude.ai/code/session_01P489vrPdHcJfMfzCGM9rZ5)_
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Ali Waseem <waseema393@gmail.com>
Stacked on #49015, which is stacked on #49011. Review those first.
## Problem
The Row Level Security guide spent 225 lines and 5 benchmark tables on
performance, 29% of the page. The `RLS Performance and Best Practices`
troubleshooting entry already covers the same six tips with the same
numbers, from the same source. Neither page tells you how to check
whether RLS is your bottleneck in the first place.
Four of the six tips are not tuning advice. Indexes, `select`-wrapping,
role scoping, and `security definer` safety change whether a policy is
correct and safe, not just fast.
## Solution
- Add `guides/database/postgres/row-level-security-performance`. It
carries the client-filter rule, the join-rewrite rule, all 5 benchmark
tables merged into one, and a new `Diagnose whether RLS is the
bottleneck` section: toggle RLS off to confirm it's the cost, then read
the plan under an impersonated role. That diagnostic exists in the
troubleshooting entry and has never been in the guide.
- Keep every rule that affects correctness on the RLS guide, grouped
under `Write policies that scale`. These are also the four the
`build-docs-002-rls-guide` eval grades, and an agent reads the guide
top-down.
- Repoint the Grafana IO troubleshooting entry at the new page.
- Rewrite `More resources` as `Related content`. Every link now says
what it is and when to use it. Adds `Advanced pgTAP testing`, the
deepest RLS testing content in the docs, which nothing here linked.
Drops discussion 14576: locked, mislabeled here as "RLS Guide and Best
Practices" when it is "RLS **Performance** and Best Practices", and
superseded by the troubleshooting entry and this new page.
**Ownership rule** so the two pages don't drift: the RLS guide owns the
rule and the correct form. The performance page owns the measurement and
the optimizer explanation. If a sentence on the performance page tells
you what to write, it belongs on the guide.
Scoped out of this PR: `More resources` was assigned to the restructure
PR in the plan, but the 14576 link is what this PR supersedes, so
leaving it would ship a stale pointer.
## Manual testing
1. Open the [RLS performance
guide](https://docs-git-docs-rls-performance-split-supabase.vercel.app/docs/guides/database/postgres/row-level-security-performance)
on the preview. It appears in the left nav under Database, Access and
security, directly below Row Level Security.
2. Select the three rule links in its intro. Each lands on the matching
section of the RLS guide.
3. Open the [Row Level Security
guide](https://docs-git-docs-rls-performance-split-supabase.vercel.app/docs/guides/database/postgres/row-level-security)
and go to `Write policies that scale`. It holds indexes,
`select`-wrapping, and role scoping, with one link out to the
performance page.
4. Open the [Grafana IO troubleshooting
entry](https://docs-git-docs-rls-performance-split-supabase.vercel.app/docs/guides/troubleshooting/interpreting-supabase-grafana-io-charts-MUynDR)
and select the RLS performance guide link. It lands on the new page.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Documentation**
* Added a dedicated guide for diagnosing and improving PostgreSQL Row
Level Security performance.
* Expanded guidance on indexing, query filters, role targeting, function
usage, and avoiding costly policy joins.
* Updated the Row Level Security guide with streamlined, scalable policy
recommendations and links to related resources.
* Added the new performance guide to the Database documentation
navigation.
* Updated troubleshooting guidance to reference the dedicated
performance guide.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Database migration — adds a table for collecting free-form product
feedback submitted from Supabase interfaces (starting with the CLI and
the MCP server), including support for deleting a submission via a
server-issued token.
## What is the current behavior?
There is no destination for feedback submitted from the CLI or MCP
server. The existing `feedback` and `feedback_comments` tables are
scoped to the docs feedback widget, so interface feedback would
otherwise end up as ad-hoc GitHub issues — with no way to revoke
something submitted by accident (e.g. a secret key pasted into the
message).
## What is the new behavior?
Adds `public.interfaces_feedback`:
| Column | Type | Notes |
| --- | --- | --- |
| `id` | `bigint` identity | primary key (not exposed through the API) |
| `created_at` | `timestamptz` | `not null default now()` |
| `feedback` | `text` | `not null`, ≤ 1000 chars — the free-form
feedback |
| `delete_token` | `uuid` | server-generated, `unique not null`;
authorizes deleting the row |
| `user_agent` | `text` | ≤ 255 chars; interface + version, also
identifies the source interface |
| `user_id` | `text` | optional, ≤ 255 chars; unverified,
interface-defined identifier |
| `project_ref` | `text` | optional, ≤ 255 chars |
| `metadata` | `jsonb` | ≤ 8 KB catch-all |
**Submission** happens exclusively through a `SECURITY DEFINER`
function, `submit_interfaces_feedback(...)`, which inserts the row and
returns the server-generated `delete_token` exactly once. There is no
insert grant or policy on the table itself, so clients cannot insert
directly or supply their own token — the function is the only door.
Execute is revoked from `PUBLIC` and granted to `anon` only (both
statements matter: local and hosted databases have different default
function ACLs).
**Deletion** is a hard `DELETE` authorized by presenting the token in an
`x-feedback-token` request header. RLS policies compare the row's
`delete_token` against that header (`current_setting('request.headers',
...)`) — the URL filter is never the security boundary; a request
without the matching header affects zero rows, even with no filter or
someone else's token in the filter. Tokens never expire (the delete
right shouldn't lapse). The header is cast to `uuid` and compared
against the untransformed column, so lookups use the unique index on
`delete_token` even for header-only reads; a malformed token header is
rejected with a `400` (`22P02`), consistent with what a malformed URL
filter value already returns.
**Context gate (defense-in-depth)**: rows submitted with a `project_ref`
and/or `user_id` additionally require the matching
`x-feedback-project-ref` / `x-feedback-user-id` headers — on both reads
and deletes — so a leaked bare token can neither read the submission
text back nor remove the row. A `NULL` column imposes no requirement:
context-free rows keep token-only behavior, and extra headers sent
against them are ignored (this keeps clients that always send their
current context from being locked out of rows submitted without it).
These are client-supplied, unverified values, so the gate is a knowledge
factor rather than an identity check; clients should persist
`{delete_token, project_ref, user_id}` together at submit time and
re-present them byte-exact (`project_ref`/`user_id` are compared as
plain text).
**Reads** are limited to `grant select (feedback, delete_token)` behind
the same token-scoped policy: a token-holder can preview their own
submission text before deleting and confirm the delete matched (`Prefer:
count=exact` → `Content-Range: */1` vs `*/0`). No other columns are
readable by any API role; `delete_token` needs select because PostgREST
requires a WHERE clause on deletes and filter columns require select
privilege.
Verified locally via `supabase db reset` + the local REST API: token
issuance, token-scoped preview and delete, zero-row results for
missing/wrong/malformed tokens (including a victim's token in the filter
without the header), the full context-gate matrix (project+user,
project-only, and context-free rows, incl. lenient extra-header
behavior), denied direct inserts and column reads, length caps enforced
through the function, and no execute for `authenticated`.
## Additional context
Linear tickets: [CLI-1946](https://linear.app/supabase/issue/CLI-1946),
[CLI-1999](https://linear.app/supabase/issue/CLI-1999)
The client-side flows (`supabase feedback add` / `feedback delete` in
the CLI, and the MCP tool) land separately in their respective repos and
will call the RPC / DELETE endpoint described above.
Supersedes #48378 — recreated on a fresh git branch so that the Supabase
preview branch used for testing this table isn't shared with unrelated
work.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Added support for collecting and storing feedback submitted through
interfaces.
* Feedback can include submission source, timestamps, user details,
project references, and additional metadata.
* Added secure feedback submission with controlled access to protect
submitted information.
* Added support for authorized feedback removal using a secure deletion
token.
* Added safeguards to validate feedback content and restrict access to
permitted information.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
UI refactor of the notebook create/update preview in the AI Assistant
panel, plus a small additive prop on the shared `CodeBlock`.
## What is the current behavior?
The assistant's notebook diff renders each cell as its own bordered box
with a gap between them, under a `6 cells` line that is easy to miss.
Cells can't be collapsed, each one carries a repeated `ADDED` badge and
a nested "Show more" toggle, and long markdown scrolls sideways instead
of wrapping.
## What is the new behavior?
<img width="796" height="1076" alt="CleanShot 2026-08-18 at 14 41 30@2x"
src="https://github.com/user-attachments/assets/45e58c6c-48f2-404b-8699-757ee96a4a8d"
/>
- The whole diff is one card: a distinct header row (notebook name,
summary, expand/collapse all) over cells glued together by dividers.
- Every cell is a `Collapsible`. Added and replaced cells open by
default; unchanged, moved, and removed cells stay as single rows but are
now inspectable instead of being content-free.
- The per-row badge is replaced by a colored gutter glyph (`+` `−` `~`
`↕`) with a tooltip naming the change type. The change type reaches the
accessible name via `aria-label` on the row.
- The nested "Show more" toggle inside each cell is gone — the row
itself is the only control.
- `CodeBlock` gains a `wrapLongLines` prop (default `false`, no change
for existing callers), used here so markdown and SQL soft-wrap. The
highlighter sets `white-space` inline on the `<code>` element, so a
class on the `<pre>` can't do this.
## Additional context
Towards FE-4143
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **New Features**
* Notebook previews now display titles, notebook icons, and clearer
bordered layouts.
* Added per-cell expand/collapse controls, including “Expand all” and
“Collapse all.”
* Long code lines can now wrap for improved readability.
* **Improvements**
* Added mode-based fallback labels when notebook titles are unavailable.
* Newly added and replaced cells expand by default, while unchanged
cells remain collapsed.
* Improved change markers, tooltips, removed-cell styling, and notebook
proposal preview spacing.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Part of my onboarding to add myself to humans.txt
## I have read the
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
file.
YES
## What kind of change does this PR introduce?
Docs update to add new joiner (me!)
## What is the current behavior?
N/A
## What is the new behavior?
I am part of the team. :)
## Additional context
Part of the onboarding process.
Closes FE-3811
## Problem
CodeRabbit reviews UI PRs without prompting on accessibility gaps
axe-core cannot judge: live regions, keyboard and hover, reduced motion,
alt quality, focus visibility, color-only state, and vague link names.
## Solution
- Add a path_instruction on `{apps,packages}/**/*.{tsx,jsx,css,mdx}`.
- Keep comments advisory. Skip tests, generated files, Radix/shadcn from
`ui`, and mechanical axe findings.
- Cover live-region lifecycle, pointer-only and hover-only UI, reduced
motion, alt quality including a two-sentence length heuristic, focus
rings, color-only state, and generic link names.
## Manual testing
1. After merge, open a PR that touches a UI or MDX file under `apps/` or
`packages/`.
2. Confirm CodeRabbit comments on at least one of: an unannounced status
change, a live region created with its message, a pointer-only or
hover-only control, animation without reduced motion, generic or
redundant or long alt, `outline-none` without a focus-visible
replacement, color-only status, or a "learn more" link that does not
name its destination.
<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit
* **Accessibility**
* Expanded accessibility review coverage for interface content and
styling.
* Reviews now identify missing focus indicators, color-only status or
selection cues, and unclear link labels.
* Continued checks cover state announcements, pointer-only interactions,
reduced-motion support, and alternative text quality.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
---------
Co-authored-by: Cursor <cursoragent@cursor.com>
Jordi Enric