supabase

38,871 Commits 1,789 Branches 62 Tags

3 Commits

Include renames

Author	SHA1	Message	Date
Saxon Fletcher Claude Opus 5.5 Joshen Lim	7994191e49	feat(studio): group consecutive assistant tool calls (#50893 ) <img width="913" height="572" alt="image" src="https://github.com/user-attachments/assets/e8112085-508a-49e4-abb1-7550248e611e" /> ## Problem A single Assistant response often produces 10+ reasoning and lookup rows ("Reasoned", "Ran search_docs", …). They push the answer down the chat, use raw tool names, and a fast tool call flashes past before the row goes back to "Thinking...". ## Solution Consecutive reasoning and lookup rows fold into one collapsible group. - Running: the header shows a tool only while it executes ("Checking policies in public..."). Between calls it reads "Thinking...", however long that lasts. Each header label stays up for at least 1 second, so quick calls no longer flash. - Finished: the header lists what the tools did, e.g. "Searched docs and checked policies", or "…, and 2 more". - Expanded (any time): every call is listed under a vertical rule. Rows still in progress shimmer, including several at once for parallel calls. ## How to test 1. Run `pnpm dev:studio` and open the Assistant on a project with a few tables. 2. Ask something that needs several lookups, e.g. "What RLS policies do I have and what do the docs recommend for them?" 3. While it streams, check the collapsed header: - It shows each tool while it runs, then goes back to "Thinking..." between calls. - Labels don't flash. Each stays up for about a second. - Only the shimmer marks progress, with no blinking cursor underneath. 4. Expand the group mid-stream. Rows read like "Checking policies in public..." rather than tool names, and only rows still in progress shimmer. 5. When it finishes, the header lists the actions ("Searched docs and checked policies") and stops shimmering. 6. Press Stop while a group is running. The unfinished row reads "Response interrupted" and stops spinning. 7. Reload the chat. Older groups show their collapsed summaries. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * New Features * AI assistant tool activity is grouped into collapsible sections with progress labels while work is underway and summaries when complete. * Expand grouped activity to review reasoning and tool details. Active tools and reasoning are highlighted, while completed reasoning without text is hidden. * Progress labels remain visible briefly during transitions, and active responses display a shimmer effect. * Bug Fixes * The loading indicator no longer appears while the assistant is processing a tool group. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com> Co-authored-by: Joshen Lim <joshenlimek@gmail.com>	2026-09-28 16:07:20 +08:00
Saxon FletcherandClaude Opus 5.5	5e59b6047e	chore(studio): extend Assistant response time and handle timeouts (#50892 ) ## Problem - Assistant responses were capped at 120 seconds and 10 steps, which is too short for longer reasoning or multi-step tool work. - When the hosting platform ended a request at that limit, the connection just dropped. The user got no explanation, and "Thinking…" and tool rows kept spinning. - Studio's own tools ignored the request's abort signal, so a stop, disconnect or deadline couldn't cancel their in-flight requests. - Aborted responses never closed their Braintrust span. Under TanStack Start, the remote MCP client was only released on `res.on('close')`, which the adapter never emits. ## Solution Uses AI SDK options instead of custom stream handling: - `maxDuration` goes to 300s and the step limit to 20. `streamText({ timeout: { totalMs } })` stops the response at 270s, leaving time to finish the stream before the platform cutoff. - `toUIMessageStream({ messageMetadata })` marks an aborted response `timedOut: true`. `Chat` ignores `abort` chunks, so the client reads this flag instead and shows a timeout alert with Retry. The flag is saved with the message, so the alert survives a reload. - `toUIMessageStream({ onEnd })` aborts the request whenever the stream ends, releasing the MCP client on both runtimes. `streamText({ onAbort })` ends the Braintrust span. - Studio tools pass the SDK's `abortSignal` to their fetches. MCP tools already did. - Reasoning and server-tool rows that never finished show "Response interrupted" instead of a spinner or "Ran X ✓". There's no per-tool timeout. Approved SQL and migrations can legitimately run longer, and aborting the HTTP request doesn't stop the query in Postgres. ## Review instructions 1. Run the unit tests: `cd apps/studio && pnpm vitest run lib/api/generate-v4.test.ts lib/ai components/ui/AIAssistantPanel` 2. To see a timeout without waiting 4.5 minutes, temporarily set `ASSISTANT_TIMEOUT_MS` in `apps/studio/lib/ai/assistant-timeout.ts` to `15_000` and run `pnpm dev:studio`. 3. Ask the Assistant something that needs several tool calls or long reasoning, for example "Audit my schema for missing indexes and RLS gaps, then write the fixes." 4. After 15 seconds, check that: - the response stops and a "Assistant response timed out" alert appears with Retry - any in-progress reasoning or tool row shows "Response interrupted" instead of spinning - Retry starts a new response - reloading the page still shows the alert on that chat 5. Stop a response with the Stop button before the deadline. It should stop without the timeout alert. 6. With the default 270s, confirm that a normal response completes as before. ## Checklist Check all before review: - [ ] I have read [CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md) - [ ] If I wrote a new docs topic or edited an existing topic, I used the `/write-the-docs` or `/edit-the-docs` skill, which references [WORD_LIST](https://github.com/supabase/supabase/blob/master/apps/docs/WORD_LIST.md) and the docs [CONTRIBUTING](https://github.com/supabase/supabase/blob/master/apps/docs/CONTRIBUTING.md) guide <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * Improvements * AI assistant responses can now run for up to five minutes, supporting longer requests. * When a response times out, the assistant displays a message suggesting you retry or ask for a smaller change. * Incomplete responses now show a “Response interrupted” notice, and loading indicators stop when generation ends. <!-- end of auto-generated comment: release notes by coderabbit.ai --> --------- Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>	2026-09-28 12:44:39 +10:00
Saxon Fletcher	d0135231fb	chore(studio): improve assistant feed performance (#50888 ) ## Problem Assistant conversations with multiple query and Edge Function blocks repeatedly render expensive content while streaming. Scrolling past the feed boundary can also move the surrounding layout. ## Solution Memoize unchanged messages, blocks, and code highlighting; batch streaming UI updates; and skip off-screen query layout while keeping block state mounted. Preserve streamed status updates and contain scrolling in the message viewport. The changes are shared by Next and TanStack. ## Review instructions 1. Compare the base branch and this branch using the same saved conversation containing 10–20 query, result/chart, and Edge Function blocks. Keep the browser, viewport, and conversation identical. 2. In Chrome DevTools, record Performance with 4× CPU throttling while streaming a follow-up, typing in the composer, and scrolling through the feed. Compare scripting/layout time and long tasks. React DevTools Profiler should show unchanged completed blocks avoiding renders during subsequent text updates. 3. Scroll away from query blocks and return. Confirm results, display settings, selections, and controls retain their state. Run a read-only query such as `select 1` and check its results still update. 4. Confirm “Thinking…” finishes, Stop retains the latest streamed text, and approval/skip, copy, edit, and branch actions still work. Repeated scrolling at the feed boundary must leave the outer layout/composer stationary; jump-to-latest and following new messages should still work. 5. Repeat in both runtimes: `STUDIO_FRAMEWORK=next pnpm dev:studio` and `STUDIO_FRAMEWORK=tanstack pnpm dev:studio`. Also check the assistant sidebar, which shares the feed. ## Validation - 170 assistant/Explorer tests and one shared CodeBlock test passed; formatting and Studio lint passed (two existing warnings). - Browser checks covered both route entry points, viewport state/geometry, and scroll behavior. - Review fixes: 23 focused tests, lint, formatting, and full Studio typechecking passed. Full production builds were not verified. ## Checklist - [x] I have read [CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md) - [x] No docs content changed; docs authoring skills are not applicable. <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * Performance * Improved responsiveness in the AI assistant by reducing unnecessary updates while messages stream and conversation history is displayed. * Optimized query previews, message rendering, and code blocks to keep the interface smoother during use. * Bug Fixes * Improved handling of message edits and deletions during generation, and preserved the latest response when generation is stopped. * Improved conversation scrolling behavior while keeping conversation content and scroll areas working as expected. <!-- end of auto-generated comment: release notes by coderabbit.ai -->	2026-09-25 18:09:34 +10:00

Author

SHA1

Message

Date

7994191e49

feat(studio): group consecutive assistant tool calls (#50893 )

<img width="913" height="572" alt="image"
src="https://github.com/user-attachments/assets/e8112085-508a-49e4-abb1-7550248e611e"
/>


## Problem

A single Assistant response often produces 10+ reasoning and lookup rows
("Reasoned", "Ran search_docs", …). They push the answer down the chat,
use raw tool names, and a fast tool call flashes past before the row
goes back to "Thinking...".

## Solution

Consecutive reasoning and lookup rows fold into one collapsible group.

- **Running:** the header shows a tool only while it executes ("Checking
policies in public..."). Between calls it reads "Thinking...", however
long that lasts. Each header label stays up for at least 1 second, so
quick calls no longer flash.
- **Finished:** the header lists what the tools did, e.g. "Searched docs
and checked policies", or "…, and 2 more".
- **Expanded (any time):** every call is listed under a vertical rule.
Rows still in progress shimmer, including several at once for parallel
calls.

## How to test

1. Run `pnpm dev:studio` and open the Assistant on a project with a few
tables.
2. Ask something that needs several lookups, e.g. "What RLS policies do
I have and what do the docs recommend for them?"
3. While it streams, check the collapsed header:
- It shows each tool while it runs, then goes back to "Thinking..."
between calls.
   - Labels don't flash. Each stays up for about a second.
- Only the shimmer marks progress, with no blinking cursor underneath.
4. Expand the group mid-stream. Rows read like "Checking policies in
public..." rather than tool names, and only rows still in progress
shimmer.
5. When it finishes, the header lists the actions ("Searched docs and
checked policies") and stops shimmering.
6. Press Stop while a group is running. The unfinished row reads
"Response interrupted" and stops spinning.
7. Reload the chat. Older groups show their collapsed summaries.




<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* AI assistant tool activity is grouped into collapsible sections with
progress labels while work is underway and summaries when complete.
* Expand grouped activity to review reasoning and tool details. Active
tools and reasoning are highlighted, while completed reasoning without
text is hidden.
* Progress labels remain visible briefly during transitions, and active
responses display a shimmer effect.
* **Bug Fixes**
* The loading indicator no longer appears while the assistant is
processing a tool group.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Joshen Lim <joshenlimek@gmail.com>

2026-09-28 16:07:20 +08:00

Saxon FletcherandClaude Opus 5.5

5e59b6047e

chore(studio): extend Assistant response time and handle timeouts (#50892 )

## Problem

- Assistant responses were capped at 120 seconds and 10 steps, which is
too short for longer reasoning or multi-step tool work.
- When the hosting platform ended a request at that limit, the
connection just dropped. The user got no explanation, and "Thinking…"
and tool rows kept spinning.
- Studio's own tools ignored the request's abort signal, so a stop,
disconnect or deadline couldn't cancel their in-flight requests.
- Aborted responses never closed their Braintrust span. Under TanStack
Start, the remote MCP client was only released on `res.on('close')`,
which the adapter never emits.

## Solution

Uses AI SDK options instead of custom stream handling:

- `maxDuration` goes to 300s and the step limit to 20. `streamText({
timeout: { totalMs } })` stops the response at 270s, leaving time to
finish the stream before the platform cutoff.
- `toUIMessageStream({ messageMetadata })` marks an aborted response
`timedOut: true`. `Chat` ignores `abort` chunks, so the client reads
this flag instead and shows a timeout alert with Retry. The flag is
saved with the message, so the alert survives a reload.
- `toUIMessageStream({ onEnd })` aborts the request whenever the stream
ends, releasing the MCP client on both runtimes. `streamText({ onAbort
})` ends the Braintrust span.
- Studio tools pass the SDK's `abortSignal` to their fetches. MCP tools
already did.
- Reasoning and server-tool rows that never finished show "Response
interrupted" instead of a spinner or "Ran X ✓".

There's no per-tool timeout. Approved SQL and migrations can
legitimately run longer, and aborting the HTTP request doesn't stop the
query in Postgres.

## Review instructions

1. Run the unit tests: `cd apps/studio && pnpm vitest run
lib/api/generate-v4.test.ts lib/ai components/ui/AIAssistantPanel`
2. To see a timeout without waiting 4.5 minutes, temporarily set
`ASSISTANT_TIMEOUT_MS` in `apps/studio/lib/ai/assistant-timeout.ts` to
`15_000` and run `pnpm dev:studio`.
3. Ask the Assistant something that needs several tool calls or long
reasoning, for example "Audit my schema for missing indexes and RLS
gaps, then write the fixes."
4. After 15 seconds, check that:
- the response stops and a "Assistant response timed out" alert appears
with Retry
- any in-progress reasoning or tool row shows "Response interrupted"
instead of spinning
- Retry starts a new response
- reloading the page still shows the alert on that chat
5. Stop a response with the Stop button before the deadline. It should
stop without the timeout alert.
6. With the default 270s, confirm that a normal response completes as
before.

## Checklist

Check all before review:

- [ ] I have read
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
- [ ] If I wrote a new docs topic or edited an existing topic, I used
the `/write-the-docs` or `/edit-the-docs` skill, which references
[WORD_LIST](https://github.com/supabase/supabase/blob/master/apps/docs/WORD_LIST.md)
and the docs
[CONTRIBUTING](https://github.com/supabase/supabase/blob/master/apps/docs/CONTRIBUTING.md)
guide

## Summary by CodeRabbit

* **Improvements**
* AI assistant responses can now run for up to five minutes, supporting
longer requests.
* When a response times out, the assistant displays a message suggesting
you retry or ask for a smaller change.
* Incomplete responses now show a “Response interrupted” notice, and
loading indicators stop when generation ends.

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

2026-09-28 12:44:39 +10:00

Saxon Fletcher

d0135231fb

chore(studio): improve assistant feed performance (#50888 )

## Problem

Assistant conversations with multiple query and Edge Function blocks
repeatedly render expensive content while streaming. Scrolling past the
feed boundary can also move the surrounding layout.

## Solution

Memoize unchanged messages, blocks, and code highlighting; batch
streaming UI updates; and skip off-screen query layout while keeping
block state mounted. Preserve streamed status updates and contain
scrolling in the message viewport. The changes are shared by Next and
TanStack.

## Review instructions

1. Compare the base branch and this branch using the same saved
conversation containing 10–20 query, result/chart, and Edge Function
blocks. Keep the browser, viewport, and conversation identical.
2. In Chrome DevTools, record Performance with 4× CPU throttling while
streaming a follow-up, typing in the composer, and scrolling through the
feed. Compare scripting/layout time and long tasks. React DevTools
Profiler should show unchanged completed blocks avoiding renders during
subsequent text updates.
3. Scroll away from query blocks and return. Confirm results, display
settings, selections, and controls retain their state. Run a read-only
query such as `select 1` and check its results still update.
4. Confirm “Thinking…” finishes, Stop retains the latest streamed text,
and approval/skip, copy, edit, and branch actions still work. Repeated
scrolling at the feed boundary must leave the outer layout/composer
stationary; jump-to-latest and following new messages should still work.
5. Repeat in both runtimes: `STUDIO_FRAMEWORK=next pnpm dev:studio` and
`STUDIO_FRAMEWORK=tanstack pnpm dev:studio`. Also check the assistant
sidebar, which shares the feed.

## Validation

- 170 assistant/Explorer tests and one shared CodeBlock test passed;
formatting and Studio lint passed (two existing warnings).
- Browser checks covered both route entry points, viewport
state/geometry, and scroll behavior.
- Review fixes: 23 focused tests, lint, formatting, and full Studio
typechecking passed. Full production builds were not verified.

## Checklist

- [x] I have read
[CONTRIBUTING.md](https://github.com/supabase/supabase/blob/master/CONTRIBUTING.md)
- [x] No docs content changed; docs authoring skills are not applicable.


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Performance**
* Improved responsiveness in the AI assistant by reducing unnecessary
updates while messages stream and conversation history is displayed.
* Optimized query previews, message rendering, and code blocks to keep
the interface smoother during use.
* **Bug Fixes**
* Improved handling of message edits and deletions during generation,
and preserved the latest response when generation is stopped.
* Improved conversation scrolling behavior while keeping conversation
content and scroll areas working as expected.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

2026-09-25 18:09:34 +10:00