`ColumnWidths` now stores a cached bounds width using
on_children_prepaint which is used to calculate the correct resize to do
on double click.

Double clicking on a column resets each column to it's default size
right now. This behavior isn't ideal, but resizing an individual column
directly requires knowing the element's bounds.

I'm hoping to find a way to get an element's bounds tomorrow morning,
once that is done I can based the calculations needed for the correct
size adjustment on the on_drag_move function that already exists and
does what we want

Co-authored-by: Remco Smits <djsmits12@gmail.com>

Co-authored-by: Remco Smits <djsmits12@gmail.com>

Supersedes: #34242

Creates an `ActionArgumentsEditor` that implements the required logic to
have a JSON language server run when editing keybinds so that there is
auto-complete for action arguments.

This is the first time action argument schemas are required by
themselves rather than inlined in the keymap schema. Rather than add all
action schemas to the configuration options we send to the JSON LSP on
startup, this PR implements support for the
`vscode-json-language-server` extension to the LSP whereby the server
will request the client (Zed) to resolve URLs with URI schemes it does
not recognize, in our case `zed://`. This limits the impact on the size
of the configuration options to ~1KB as we send URLs for the language
server to resolve on demand rather than the schema itself. My
understanding is that this is how VSCode handles JSON schemas as well. I
plan to investigate converting the rest of our schema generation logic
to this method in a follow up PR.

Co-Authored-By: Cole <cole@zed.dev>

Release Notes:

- N/A *or* Added/Fixed/Improved ...

Release Notes:

- N/A

Release Notes:

- N/A

Simplify docs for mac/linux/windows by consolidating the backend
dependencies (collaboration) docs into local-collaboration.md. Most
users building zed will not need to do this -- streamline them into
getting setup to build the zed client app first.

Release Notes:

- N/A

Fix a spurious error in Zed logs from the OpenRouter svg Logo introduced
in https://github.com/zed-industries/zed/pull/29496:

```log
WARN  [usvg::parser::svgtree] Failed to parse clip-path value: 'url(#clip0_205_3)'.
```

Release Notes:

- N/A

Closes #32675

Exactly the same changes as in #33640 by @sviande

The PR has been in WIP state for 3 weeks with no activity, and the issue
basically makes Mistral models unusable. I have tested the changes
locally, and it does indeed work. Full credit goes to @sviande, I just
want this feature to be finished.

Release Notes:

- agent: Fixed an issue with tool calling with the Mistral provider
(thanks [@sviande](https://github.com/sviande) and
[@armyhaylenko](https://github.com/armyhaylenko))

Co-authored-by: sviande <sviande@gmail.com>

Closes #27118
Closes #34165

Fix a small issue after we landed
https://github.com/zed-extensions/ruby/pull/113+ where we introduced
`HTML/ERB` and `YAML/ERB` language IDs to improve user experience. Sorry
about that. Thanks!

Release Notes:

- N/A

This PR adds a component level state API to GPUI, as well as a few
utilities for simplified interactions with entities

Release Notes:

- N/A

Release Notes:

- N/A

---------

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

Added [rhysd/actionlint](https://github.com/rhysd/actionlint/) a static
checker for GitHub Actions workflow files.
Install locally with `brew install actionlint` the run with
`actionlint`.

Inspired by: https://github.com/zed-industries/zed/pull/34704 which
yielded this observation:

> In github actions: 
> 1. strings are truthy
> 2. `${{ }}` will become a string if it doesn't wrap the whole value.
>
> So `if: false && true` becomes `false`
> and `if: ${{ false && true }}` becomes `false`
> but `if: false && ${{ true }}` becomes `"false && true"` which
evaluates true
> The reason you sometimes need `${{ }}` is because YAML doesn't like
`!`
> so `if: !false` is invalid yaml
> and `if: ${{ !false }}` works just fine.

Changes:
- Add `actionlint` job
- Refactor `job_spec` job to be more readable
- Fix all `actionlint` and `shellcheck` errors in Actions workflows (62
in all)
- Add `self-mini-macos` and `self-32vcpu-windows-2022` labels to
self-hosted runners. Not strictly related, but useful if you need to
take a runner out of the rotation (since `macOS`, `self-hosted`, and
`ARM64` are auto-set and cannot be added/removed).
- Change ci.yml macos_relase to target `self-mini-macos` instead of
`bundle` which was previously deprecated.

This would've caught the error fixed in
https://github.com/zed-industries/zed/pull/34704. Here's what that [job
failure](https://github.com/zed-industries/zed/actions/runs/16376993944/job/46279281842?pr=34729)
would've looked like.

Release Notes:

- N/A

Broken by #34664

Release Notes:

- N/A

Release Notes:

- N/A

Release Notes:

- N/A

Release Notes:

- N/A

---------

Co-authored-by: Ben Kunkle <ben@zed.dev>

When `use_modifier_to_send` is turned on, holding `cmd`/`ctrl` is
necessary to send a message in the agent panel. Text threads already use
`cmd-enter` by default to submit a message, and it was done this way to
have the usual text editing bindings not taken over when writing a
prompt, sort of stimulating more thoughtful writing. While `enter` to
send is still somewhat a huge pattern in chat-like LLM UIs, it still
makes sense to allow this for the new agent panel... hence the existence
of this setting now!

Release Notes:

- agent: Added the `use_modifier_to_send` setting, which makes holding a
modifier (`cmd`/`ctrl`), together with `enter`, required to send a new
message.

This will be useful for both the current agent panel and some other
onboarding stuff we're working on. Also ended up removing the
`SwitchWithLabel` as it was unused.

Release Notes:

- N/A

Release Notes:

- N/A

This PR fixes an issue where the topmost header in a multibuffer would
jump when the corresponding buffer was folded.
The issue arose because for the topmost header, the offset within the
scroll anchor is negative, as the corresponding buffer only starts below
the header itself and thus the offset for the scroll position has to be
negative.
However, upon collapsing that buffer, we end up with a negative vertical
scroll position, which causes all kinds of different problems. The issue
has been present for a long time, but became more visible after
https://github.com/zed-industries/zed/pull/34295 landed, as that change
removed the case distinction for buffers scrolled all the way to the
top.

This PR fixes this by clamping just the vertical scroll position upon
return, which ensures the negative offset works as expected when the
buffer is expanded, but the vertical scroll position does not turn
negative once the buffer is folded.

Release Notes:

- Fixed an issue where folding the topmost buffer in a multibuffer would
cause the header to jump slightly.

Closes #33510 https://github.com/zed-extensions/php/issues/29

If certain language servers do not provide an insert/replace range, we
use `surrounding_word` as a fallback for that range, which internally
uses `word_characters`. It makes sense to use
`completion_query_characters` instead of `word_characters` to get that
range, because we use `completion_query_characters` to query completions
in the first place.

That means, for some hypothetical reason (e.g., if the Tailwind server
stops providing insert/replace ranges), we would correctly fall back to
the range "bg-blue-200^" instead of "200^", because
`completion_query_characters` includes "-" in this case.

For this particular fix, right now the default PHP language server
`phpactor` does not provide an insert/replace range, and hence
completion query character is used, which is `$` in this case.

Note that `$` isn't in word characters for reasons mentioned here:
https://github.com/zed-extensions/php/issues/14

Release Notes:

- Fixed an issue where accepting variable completion in PHP would result
in a double $ sign in the prefix.

These should have been part of
https://github.com/zed-industries/zed/pull/30785 but I forgot to add
them there.

Release Notes:

- N/A

This PR removes the `screen-capture` feature from `gpui` when depending
on it in `collab`.

Release Notes:

- N/A

This includes making sure that both the agent panel and Zed's edit
prediction have a consistent narrative when it comes to onboarding users
into the AI features, considering the possible different plans and
conditions (such as being signed in/out, account age, etc.)

Release Notes:

- N/A

---------

Co-authored-by: Bennet Bo Fenner <53836821+bennetbo@users.noreply.github.com>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>

Reverts zed-industries/zed#29718

We've noticed some issues with Zed on Intel-based Macs where typing has
become sluggish, and git bisect has seemed to point towards this PR.
Reverting for now, until we can understand why it is causing this issue.

Now ! means "no ancestors matches this", and > means "any descendent"
not "any child".

Updates #34570

Co-authored-by: Ben Kunkle <ben@zed.dev>

Release Notes:

- *Breaking change*. The context predicates in the keymap file now
handle ! and > differently. Before this change ! meant "this node does
not match", now it means "none of these nodes match". Before this change
> meant "child of", now it means "descendent of". We do not expect these
changes to break many keymaps, but they may cause subtle changes for
complex context queries.

---------

Co-authored-by: Ben Kunkle <ben@zed.dev>

This PR adds a new `POST /users/:id/refresh_llm_tokens` endpoint to
Collab so that we can refresh LLM tokens from Cloud.

Release Notes:

- N/A

Release Notes:

- N/A

@cole-miller found a root cause of our struggles with attach scenarios;
we did not fetch .so files necessary for attaching to work,
as we were downloading DebugPy source tarballs from GitHub.

This PR does away with it by setting up a virtualenv instead that has
debugpy installed.

Closes #34660
Closes #34575

Release Notes:

- debugger: Fixed attaching with DebugPy. DebugPy is now installed
automatically from pip (instead of GitHub), unless it is present in
active virtual environment. Additionally this should resolve any startup
issues with missing .so on Linux.

Actual default is 45


https://github.com/zed-industries/zed/blob/fd05f17fa7e0253bc98698e84a1f8939bcb6616f/assets/settings/default.json#L1402

Release Notes:

- N/A

Release Notes:

- agent: Fixed an issue where clicking on "Start New Thread" in the
agent configuration would not always switch to the correct
provider/model

Release Notes:

- N/A

Small improvement for navigating inside the component preview.

Release Notes:

- N/A

- **Fix cancellation of tool calls**
- **Make tool_call test more resilient**
- **Fix tool call confirmation test**

Release Notes:

- N/A

Currently, we're not able to run Go sub-tests that have a raw string
(e.g. we're using multi-line names a lot) via the UI. I added the
changes that are needed, plus a handful of tests to cover the basics.

Quick comparison:

Before:
<img width="901" height="370" alt="before"
src="https://github.com/user-attachments/assets/4e5cadeb-9a0c-49e2-b976-2223e1010f85"
/>



After:
<img width="901" height="505" alt="after"
src="https://github.com/user-attachments/assets/994fc69b-f720-488c-a14b-853a3ca2f53c"
/>


Release Notes:

- Added support for Go subtest runner with raw string names

Use the `static` keyword to actually make the `LazyLock` static, which
previously would reinitialize on every call to `SvgRenderer::new`.

Related: #26335 

Release Notes:

- N/A

Before this, indentation did not automatically increase after
if/for/while/do/else statements in C++, and only increased after if/for
in C. This led to Zed using last line logic when inserting lines *after*
the indented statement, as well as not indenting the statement itself,
resulting in irregular indentation during typing.

Just adding indentation (similar to C) creates a new problem: now if a
scope is started with a brace on a new line, that brace is indented.
Thus we need to deindent it.

Using else_clause in the indent guide results in the else statement
being indented forward as well, so we need to deindent that too.

Note: the most significant issue for me is the one where indentation
jumps forward when inserting lines after indented lines. Unfortunately,
it appears that fixing that issue requires all of these other changes. I
would have preferred a simpler fix, but I'm not sure if disabling last
line behavior for C/C++ is appropriate as it probably breaks something
else, like cases where the file is incomplete and the statements can't
be parsed properly.

Editing flow before this change:

[Screencast From 2025-07-16
08-31-36.webm](https://github.com/user-attachments/assets/3dea86c5-47bd-47c2-aee8-b0aa613948e6)

Editing flow after this change:

[Screencast From 2025-07-16
08-35-36.webm](https://github.com/user-attachments/assets/7ef23e60-1ee3-49fd-90f9-d53f909ca674)

(note: the "else" snippet is completely breaking the flow here, but I
think that comes from clangd by default? Unfortunately I haven't found a
way to disable it cleanly but that is a separate problem that happens
right now too.)

Release Notes:

- Improve indentation during typing for C/C++ around if/for/while/do
blocks

This is now supported.

Release Notes:

- N/A

Closes #34111

In Helix mode, the `;` key should collapse the current selection without
moving the cursor. I've added a new action `vim::HelixCollapseSelection`
to support this behavior.


https://github.com/user-attachments/assets/1a40821a-f56f-456e-9d37-532500bef17b

Release Notes:

- Added `;` key binding to collapse the current text selection in Helix
mode

Quick follow-up to https://github.com/zed-industries/zed/pull/34669 so
that the buttons don't look so big in comparison to the callout.

Release Notes:

- N/A