## Summary
Enable image processing capabilities for GPT-5 series models by updating
the `supports_images()` method.

## Changes
- Add vision support for `gpt-5`, `gpt-5-mini`, and `gpt-5-nano` models
- Update `supports_images()` method in
`crates/language_models/src/provider/open_ai.rs`

## Models with Vision Support (after this PR)
- gpt-4o
- gpt-4o-mini
- gpt-4.1
- gpt-4.1-mini
- gpt-4.1-nano
- gpt-5 (new)
- gpt-5-mini (new)
- gpt-5-nano (new)
- o1
- o3
- o4-mini

This brings GPT-5 vision capabilities in line with other OpenAI models
that support image processing.

Release Notes:

- Added vision support for OpenAI models

Release Notes:

- N/A

Co-authored-by: Michael Sloan <mgsloan@gmail.com>

Release Notes:

- N/A

Co-authored-by: Michael Sloan <mgsloan@gmail.com>

Release Notes:

- N/A

Co-authored-by: Oleksiy <oleksiy@zed.dev>

Closes #34881

For horizontal scroll, we weren't keeping track of the `local` bool, so
whenever the agent tries to autoscroll horizontally, it would be seen as
a user scroll event resulting in unfollow.

Release Notes:

- Fixed an issue where the Follow Agent could unexpectedly stop
following during edits.

Closes #31108

The `num_tokens_from_messages` method we use from `tiktoken-rs` creates
new BPE every time that method is called. This creation of BPE is
expensive as well as has some underlying issue that keeps memory from
releasing once the method is finished, specifically noticeable on Linux.
This leads to a gradual increase in memory every time that method is
called in my case around +50MB on each call. We call this method with
debounce every time user types in Agent Panel to calculate tokens. This
can add up really fast.

This PR lands quick fix, while I/maintainers figure out underlying
issue. See upstream discussion:
https://github.com/zurawiki/tiktoken-rs/issues/39.

Here on fork https://github.com/zed-industries/tiktoken-rs/pull/1,
instead of creating BPE instances every time that method is called, we
use singleton BPE instances instead. So, whatever memory it is holding
on to, at least that is only once per model.

Before: Increase of 700MB+ on extensive use

On init:
<img width="500" alt="prev-init"
src="https://github.com/user-attachments/assets/70da7c44-60cb-477b-84aa-7dd579baa3da"
/>
First message:
<img width="500" alt="prev-first-call"
src="https://github.com/user-attachments/assets/599ffc48-3ad3-4729-b94c-6d88493afdbf"
/>
Extensive use:
<img width="500" alt="prev-extensive-use"
src="https://github.com/user-attachments/assets/e0e6b688-6412-486d-8b2e-7216c6b62470"
/>

After: Increase of 50MB+ on extensive use
On init:
<img width="500" alt="now-init"
src="https://github.com/user-attachments/assets/11a2cd9c-20b0-47ae-be02-07ff876e68ad"
/>
First message:
<img width="500" alt="now-first-call"
src="https://github.com/user-attachments/assets/ef505f8d-cd31-49cd-b6bb-7da3f0838fa7"
/>
Extensive use: 
<img width="500" alt="now-extensive-use"
src="https://github.com/user-attachments/assets/513cb85a-a00b-4f11-8666-69103a9eb2b8"
/>

Release Notes:

- Fixed issue where Agent Panel would cause high memory consumption over
prolonged use.

Closes #35980

Release Notes:

- Fixed Python Debug sessions not starting up when a session is started
up for the first time.

Follow-up of https://github.com/zed-industries/zed/pull/35682

Release Notes:

- N/A

Co-authored-by: Smit Barmase <smit@zed.dev>

Switch our Linux aarch_64 release builds from Linux on Graviton (32
vCPU, 64GB) to Linux running on Apple M4 Pro (8vCPU, 32GB). Builds are
faster (20mins vs 30mins) for the same cost (960 unit minutes;
~$0.96/ea).

<img width="763" height="285" alt="Screenshot 2025-08-08 at 13 14 41"
src="https://github.com/user-attachments/assets/12c45c8b-59f3-40d8-974c-1003b5080287"
/>

Release Notes:

- N/A

This PR fixes an issue where Edit Predictions would not be available in
buffers that were opened when the workspace loaded.

The issue was that there was a race condition between fetching/setting
the authenticated user state and when we assigned the Edit Prediction
provider to buffers that were already opened.

We now wait for the event that we emit when we have successfully loaded
the user in order to assign the Edit Prediction provider, as we'll know
the user has been loaded into the `UserStore` by that point.

Closes https://github.com/zed-industries/zed/issues/35883

Release Notes:

- Fixed an issue where Edit Predictions were not working in buffers that
were open when the workspace initially loaded.

Co-authored-by: Richard Feldman <oss@rtfeldman.com>

This PR makes it so we refresh the LLM token upon receiving a
`UserUpdated` message from Cloud over the WebSocket connection.

Release Notes:

- N/A

This PR updates signed-out state of the Collab panel to show when not
connected to Collab, as opposed to just when the user is signed-out.

Release Notes:

- N/A

This PR makes it so that only Zed staff connect to Collab automatically.

Anyone else can connect to Collab manually when they want to collaborate
(but this is not required for using Zed's LLM features).

Release Notes:

- N/A

---------

Co-authored-by: Richard <richard@zed.dev>

This PR wires up handling for the new `UserUpdated` message coming from
Cloud over the WebSocket connection.

When we receive this message we will refresh the authenticated user.

Release Notes:

- N/A

Co-authored-by: Richard <richard@zed.dev>

This PR adds a new WebSocket connection to Cloud.

This connection will be used to push down notifications from the server
to the client.

Release Notes:

- N/A

---------

Co-authored-by: Marshall Bowers <git@maxdeviant.com>

This PR adds types for the Cloud WebSocket protocol to the
`cloud_api_types` crate.

Release Notes:

- N/A

We used to not respond at all to requests that we didn't have a handler
for, which is yuck. It may have left the language server waiting for the
response for no good reason. The other (worse) finding is that we did
not have a full definition of an Error type for LSP, which made it so
that a spec-compliant language server would fail to deserialize our
response (with an error). This then could lead to all sorts of
funkiness, including hangs and crashes on the language server's part.

Co-authored-by: Lukas <lukas@zed.dev>
Co-authored-by: Remco Smits <djsmits12@gmail.com>

Co-authored-by: Anthony Eid <hello@anthonyeid.me>

Closes #ISSUE

Release Notes:

- Improved reporting of errors to language servers, which should improve
the stability of LSPs ran by Zed.

---------

Co-authored-by: Lukas <lukas@zed.dev>
Co-authored-by: Remco Smits <djsmits12@gmail.com>
Co-authored-by: Anthony Eid <hello@anthonyeid.me>

Follow-up to:
- https://github.com/zed-industries/zed/pull/35826

Release Notes:

- N/A

In response to an ongoing BuildJet outage, consider migrating CI to
GitHub hosted runners.

Also includes revert of (causing flaky tests):
- https://github.com/zed-industries/zed/pull/35741

Downsides:
- Cost (2x)
- Force migration to Ubuntu 22.04 from 20.04 will bump our glibc minimum
from 2.31 to 2.35. Which would break RHEL 9.x (glibc 2.34), Ubuntu 20.04
(EOL) and derivatives.

Release Notes:

- N/A

Follow-up of https://github.com/zed-industries/zed/pull/35682

Release Notes:

- N/A

If we can't find CC in the PATH, we'll now fall back to a known local
install path.

Release Notes:

- N/A

Release Notes:

- N/A

Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>

This uses the `current_user` watch in the `UserStore` instead of looping
every 100ms in order to detect if the user had signed in.

We are changing this because we noticed it was causing the deterministic
executor in tests to never detect a "parking with nothing left to run"
situation.

This seems better in production as well, especially for users who never
sign in.

/cc @maxdeviant 

Release Notes:

- N/A

Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>

Follows https://github.com/zed-industries/zed/pull/35642

Release Notes:

- Fixed accidentally picking a non executable as a rust-analyzer server
when downloading fails

cherry pick changes from #35510 onto latest main

Release Notes:

- N/A

---------

Co-authored-by: Nathan Sobo <nathan@zed.dev>
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Lukas Wirth <lukas@zed.dev>

Closes https://github.com/zed-industries/zed/issues/33668

The recursive case increments both indices by 1, but only one of the two
had a base case check in the function prologue so the other could spill
over into a different matrix row or out of bounds entirely.

Lacking a test as I haven't figured out a test case yet.

Release Notes:

- Fixed out of bounds panic in fuzzy matching

Release Notes:

- Added GitHub artifact digest verification for rust-analyzer and clangd
binary downloads, skipping downloads if cached binary digest is up to
date
- Added verification that cached rust-analyzer and clangd binaries are
executable, if not they are redownloaded

---------

Co-authored-by: Kirill Bulatov <kirill@zed.dev>

Fixes a regression introduced in
https://github.com/zed-industries/zed/pull/34992

### Background

Paths are rendered first to an intermediate MSAA texture, and then
copied to the final drawable. Because paths can have transparency, it's
important that pixels are not copied repeatedly if paths have
overlapping bounding boxes. When N paths have the same draw order, we
infer that they must have disjoint bounding boxes, so that we can copy
them each individually (as opposed to copying a single rect that
contains them all). Previously, the bounding box that we were using to
copy paths was not accounting for the path's content mask (but it is
accounted for in the bounds tree that determines their draw order).


This cause bugs like this, where certain path pixels spuriously had
their opacity doubled:


https://github.com/user-attachments/assets/d792e60c-790b-49ad-b435-6695daba430f

This PR fixes that bug.

* [x] mac
* [x] linux
* [x] windows

Release Notes:

- Fixed a bug where a selection's opacity was computed incorrectly when
it overlapped with another editor's selections in a certain way.

This _should_ allow sentry to associate related panic events with the
same issue, but it doesn't change the issue title. I'm still working on
figuring out how to set those fields, but in the meantime this should at
least associate zed versions with crashes

Release Notes:

- N/A

Release Notes:

- N/A

We broke it in #35263 when we changed the open options to use
`create_new`

Release Notes:

- N/A

Follow up to: https://github.com/zed-industries/zed/pull/35670,
simplifies the List state APIs so you no longer have to worry about
strong vs. weak pointers when rendering list items.

Release Notes:

- N/A

---------

Co-authored-by: Agus Zubiaga <agus@zed.dev>

Fixes https://github.com/zed-industries/zed/issues/33257

Previously, the scale-factor-change-handling logic relied on
`SetWindowPos` enqueuing a `WM_SIZE` window event. But that does not
happen when the window is maximized. So when the scale factor changed,
maximized windows neglected to call their `resize` callback, and would
misinterpret the positions of mouse events.

This PR adds special logic for maximized windows, to ensure that the
size is updated appropriately.

Release Notes:

- N/A

* Straightens out the `*_ext.rs` workflow for clangd and rust-analyzer:
no need to asynchronously query for the language server, as we sync that
information already.
* Fixes inlay hints editor menu toggle not being shown in the remote
sessions

Release Notes:

- Fixed inlay hints editor menu toggle not being shown in the remote
sessions

Release Notes:

- N/A

This PR adds handling for the case where an agent binary exits
unexpectedly after successfully establishing a connection.

Release Notes:

- N/A

---------

Co-authored-by: Agus <agus@zed.dev>

This PR removes some billing-related queries that are no longer used.

Release Notes:

- N/A

Just making sure the switch's dimensions aren't affected by the need to
having an outer border to represent focus.

Release Notes:

- N/A

Use includes centralizing what each plan delivers in one single file
(`plan_definitions.rs`).

Release Notes:

- N/A

Release Notes:

- N/A

> Any user interaction that is equivalent to a click, such as pressing
the Space key or Enter key while the element is focused. Note that this
only applies to elements with a default key event handler, and
therefore, excludes other elements that have been made focusable by
setting the
[tabindex](https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Global_attributes/tabindex)
attribute.

https://developer.mozilla.org/en-US/docs/Web/API/Element/click_event

---------

Co-authored-by: Anthony <anthony@zed.dev>
Co-authored-by: Mikayla Maki <mikayla.c.maki@gmail.com>
Co-authored-by: Umesh Yadav <23421535+imumesh18@users.noreply.github.com>

When you switched away from an ACP thread, the `AcpThreadView` entity
(and thus thread, and subprocess) was leaked. This happened because we
were using `cx.processor` for the `list` state callback, which uses a
strong reference.

This PR changes the callback so that it holds a weak reference, and adds
some tests and assertions at various levels to make sure we don't
reintroduce the leak in the future.

Release Notes:

- N/A