The Dev Container [metadata
reference](https://containers.dev/implementors/json_reference/#port-attributes)
defines all five `PortAttributes` fields (`elevateIfNeeded`, `label`,
`onAutoForward`, `protocol`, `requireLocalPort`) as optional with
defined defaults. The struct required all of them, causing
deserialization failures for any `portsAttributes` entry that omitted a
field.
Add `#[serde(default)]` to each field, wrap `label` and `protocol` in
`Option<T>` (spec default: not set), and derive `Default` on
`OnAutoForward` with `Notify` as the default variant (spec default:
`notify`). The bool fields already default to false via
`serde(default)`, matching the spec.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [ ] Performance impact has been considered and is acceptable
Closes#53686.
Release Notes:
- Made all `PortAttributes` fields optional according to the Dev
Container spec
This changes our approach for creating Zed version bumps with the Zed
Zippy identity:
With these changes, the Zippy version bumps as well as the tag creations
will be verified as per
https://docs.github.com/en/authentication/managing-commit-signature-verification/about-commit-signature-verification#signature-verification-for-bots.
This adds an extra layer of security for version bumps and release
triggers done by the Zippy identity.
Eventually, we can move code more around for this so it happens in the
bot itself, but for now, I think its a good and solid interim solution
to have this here.
Release Notes:
- N/A
There were a few places where you could trigger generation without
causing the `interacted_at` field to be set.
Also renames `message_sent_or_queued` to `interacted`, to be consistent
with the field on `ThreadMetadata`
Release Notes:
- N/A or Added/Fixed/Improved ...
Follow up to #54206
`on_flags_ready` relied on the fact that the `FeatureFlagStore` was only
set once the flags had been received from the server.
However, after #54206 the global gets instantiated earlier, without the
flags being resolved.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
Update how auto indentation is handled by `Editor::newline`, in the case
where auto-indentation is not `AutoIndentationMode::None`, so as to
extend the range of the edits being applied to the beginning of the
line, in case the line only contains the indentation whitespace or
there's only whitespace indentation before the cursor position
This ensures that newlines are not left with whitespace.
Closes#34316
Release Notes:
- Fixed auto indent leaving trailing whitespace on blank indented lines
when creating new lines.
---------
Co-authored-by: dino <dinojoaocosta@gmail.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54193
Additionally, respects `cacheFrom` and `options` args defined in
devcontainer.json's `build` object. Also, addresses a bug which was
hiding these problems by overriding the primary build target with
aliasing.
Release Notes:
- Fixed an image aliasing but, respect build context and build args in
dev containers
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#52559
Release Notes:
- Fixed an issue where holding the modifiers used in
`editor::AcceptEditPrediction` would cause the code action menu to
dissapear
This PR adds support for rendering **Netpbm** image formats (`.pbm`,
`.ppm`, `.pgm`) within Zed's built-in image viewer.
These formats are particularly useful for projects that want minimal
external dependencies, a common scenario in academic environments and
low-level graphics programming.
Since the underlying `image` crate and `GPUI` already provide support
for these codecs, this change explicitly exposes the `Pnm` variant
within `gpui::ImageFormat` by mapping it to `image::ImageFormat::Pnm`.
## Screenshots/Examples
Below is an example of `.pbm`, `.ppm`, and `.pgm` files being rendered
correctly in the image preview (images taken from
https://filesamples.com):
<img width="1917" height="1012" alt="pnm_example"
src="https://github.com/user-attachments/assets/0056133f-908c-4c91-ba9d-53aef0657b05"
/>
Release Notes:
- Added support for PNM image previews (`.pbm`, `.ppm`, `.pgm`).
## Context
`ctrl-n` / `cmd-n` doesn't work on the Welcome tab. The global binding
lives under `Workspace && !Terminal` (macOS) and similar contexts that
don't include Welcome. The fix just adds the same binding to the Welcome
context block on all three platforms — same approach the font-size and
recent-project shortcuts already use there.
Closes#52426
## Demo
### Before:
https://github.com/user-attachments/assets/69becde8-25d2-45e3-9e7c-416b7937bd17
### After:
https://github.com/user-attachments/assets/6d9ede76-7adb-4527-bfef-c18d5b8a4fb4
## How to review
One line added per platform keymap file. Check that `ctrl-n` / `cmd-n`
maps to `workspace::NewFile` in the `Welcome` block of:
- `assets/keymaps/default-macos.json`
- `assets/keymaps/default-linux.json`
- `assets/keymaps/default-windows.json`
## Self-review checklist
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed `ctrl-n` / `cmd-n` (New File) not working on the Welcome tab
Co-authored-by: Danilo Leal <67129314+danilo-leal@users.noreply.github.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54283
Release Notes:
- Fixed: Git graph commit detail header now uses “1 Changed File” when
exactly one file changed, and “N Changed Files” otherwise, instead of
always saying “Changed Files”.
---------
Co-authored-by: Kunall Banerjee <hey@kimchiii.space>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54305
Release Notes:
- Fixed inconsistent keybinding hint casing in the outline panel
(“Toggle Panel With …”) by using standard keybinding rendering.
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Co-authored-by: Danilo Leal <67129314+danilo-leal@users.noreply.github.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#54309
Release Notes:
- Fixed an outline panel issue where the pin/unpin tooltip could show
stale text after toggling.
Follow up to https://github.com/zed-industries/zed/pull/52730.
This PR adds a boolean setting `limit_content_width` in the Agent Panel
settings that allows turning off the content max-width entirely, which
was added for better readability. We had a handful of requests for it,
so it feels fair.
<img width="500" alt="Screenshot 2026-04-20 at 8 57@2x"
src="https://github.com/user-attachments/assets/d2540b35-3fa8-4424-895d-dc499ac4839c"
/>
Release Notes:
- Agent: Added a new `limit_content_width` setting in the agent panel
that allows turning off the content max-width limit.
Stacked on top of #54112
This is part 2 of 3 towards #51197
More details from the original PR #53551
This PR includes the changes from #54112 , im not sure how to avoid
that, my understanding is that after that one is merged, this PR can be
rebased onto main and everything will be correct. You can also view the
version of this that does reflect the changes more directly here:
https://github.com/feitreim/zed/pull/1
## Changes
In this PR I added a more general string matching functionality to
`fuzzy_nucleo`, in order to have proper testing for this, I also changed
the command palette, tab switching picker, branch picker, and recent
projects picker to use this new implementation. I think the command
palette change in particular is awesome, just super nice to vaguely
gesture at the command i want and have it pop right up.
The main change here and departure from
https://github.com/zed-industries/zed/pull/37123 is realizing that the
primary reason for the regressions is actually how nucleo handles smart
case, the old `fuzzy` crate only uses the smart case argument to score
things differently, while nucleo actually filters on the case, eg. with
smart case query "Apple" wouldnt match "apple". To get around this we
always pass `CaseMatching::Ignore` to nucleo and implement the same
score modifications from fuzzy in our code.
There is a performance cost to that, of course, but from my testing it
is fairly static, not growing as the size increases, so maybe a query
takes 35 µs instead of 25 µs, but a query that takes 800 µs will only
take 820 µs.
Benchmark:
| kind | query | size | nucleo | fuzzy | nucleo/fuzzy |
|---|---|---:|---:|---:|---:|
| string | 1-word | 100 | 9.15 µs | 24.6 µs | 0.37× |
| string | 1-word | 1000 | 150.2 µs | 207.2 µs | 0.72× |
| string | 1-word | 10000 | 1.34 ms | 2.07 ms | 0.65× |
| string | 2-word | 100 | 5.16 µs | 2.94 µs | 1.75× |
| string | 2-word | 1000 | 29.0 µs | 11.0 µs | 2.63× |
| string | 2-word | 10000 | 210.6 µs | 55.5 µs | 3.79× |
| string | 4-word | 100 | 2.57 µs | 2.33 µs | 1.10× |
| string | 4-word | 1000 | 6.98 µs | 5.85 µs | 1.19× |
| string | 4-word | 10000 | 20.0 µs | 12.0 µs | 1.66× |
When I added the 4-word queries to the benchmarks I was actually really
concerned that the performance would be awful, making it unsuitable for
the command palette especially. However, I think due to the CharBag
pre-filtering when the query is longer, the performance is actually way
better than the 2 word case.
Video:
https://github.com/user-attachments/assets/3cd7221b-424f-4fd3-8df1-5543dcc340a3
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Improved fuzzy matching in the command palette, branch picker, tab
switcher, and recent projects picker to support multi-word queries.
---------
Co-authored-by: Yara <git@yara.blue>
Fix two issues with reasoning support in the Copilot provider:
- Responses API path: use the user's thinking_effort setting instead of
hardcoding Medium effort
- Chat Completions path: compute and pass thinking_budget when thinking
is enabled, instead of unconditionally setting it to None
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [ ] Performance impact has been considered and is acceptable
Closes#52140
Release Notes:
- Fixed a bug where copilot wouldn't use the thinking level the user's
have set
---------
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Co-authored-by: Bennet Bo Fenner <bennet@zed.dev>
Update all direct calls to `Editor::inline_blame_popover::take` to now
be calls to `Editor::hide_blame_popover` as this ensures that the
popover is also not shown in case its task still hasn't finished.
This fixes a bug where the inline git blame popover could be shown even
after the user had opened a modal.
Release Notes:
- Fixed the inline Git blame popover sometimes appearing after opening a
modal.
---------
Co-authored-by: dino <dinojoaocosta@gmail.com>
This is just a refinement given it's pretty common for menus under
ellipsis icon buttons to also open as a context-menu through the mouse's
right-button click. This should make it slightly more convenient to
interact with this menu.
Release Notes:
- N/A
This change contains a number of fixes to make kept_rate more intuitive.
It also adds a CLI utility to print debug info on how the metric is
computed.
Release Notes:
- N/A
This PR makes Zed only have one worktree picker, as opposed to a flavor
of it in the title bar and another in the agent panel. It then moves it
to the title bar, making it always present, so that its trigger is
separate from the branch picker (which now contains only two views:
branches and stashes). For the worktree picker, I'm mostly favoring the
behavior we've introduced in the agent-panel-flavored version.
It also updates the title bar settings migration to use the JSON
`migrate_settings` helper instead of a shallow Tree-sitter rewrite, so
old `show_branch_icon = true` values are promoted to
`show_branch_status_icon = true` across root, platform, release-channel,
and profile settings scopes.
- [x] Move worktree creation logic to the `git_ui` crate to make this
more generic and less agent-specific
- [x] Double-check the remote use case and ensure nothing broke there
- [x] Improve the UX for the detached HEAD state; better invite people
to create a branch
- [x] Migrate `show_branch_icon = true` to `show_branch_status_icon =
true` across nested settings scopes
Suggested .rules additions
When migrating renamed settings keys that can appear in platform
overrides, release-channel overrides, or profiles, prefer the JSON
`migrations::migrate_settings` helper over shallow Tree-sitter key
rewrites unless tests explicitly cover every nested scope that can
contain the key.
Release Notes:
- Improved migration of the title bar branch status icon setting.
---------
Co-authored-by: Nathan Sobo <nathan@zed.dev>
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Super small update changing two instances where
`DiagnosticSeverity::HINT` was set to use the color for "info".
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#45637
Release Notes:
- Fixed various instances of hint level diagnostics using the color
designated for info
Fixes#42787
## Summary
- When "Go to Definition" navigates into a dependency (e.g. `.venv/`,
`node_modules/`), the save dialog for new files defaulted to that
directory
- `most_recent_active_path` now checks the `read_only_files` setting and
skips matching paths, falling back to the next eligible path, the
worktree root, or the home directory
## Design tradeoffs
We considered three approaches:
1. **Filter by `is_ignored`/`is_hidden`/`is_external` on worktree
entries** — catches `.venv` when gitignored or when it's a dotfile, but
also false-positives on directories like `.github/workflows/` that users
intentionally edit.
2. **Use preview tab status** — "Go to Definition" opens files as
preview tabs, so skipping preview paths targets the right intent. But it
doesn't work when preview tabs are disabled, and the signal is transient
(preview status changes as you interact with tabs).
3. **Use `read_only_files` setting** (this PR) — an explicit user
declaration of "I never want to edit files here." If you can't edit
them, you don't want to save new files next to them either. This is the
clearest signal of intent and respects user configuration. The tradeoff
is that `read_only_files` is empty by default, so users need to
configure it. But the kind of user bothered by the save dialog
defaulting to a dependency directory is the same kind of user who
already configures `read_only_files` (see
[#46827](https://github.com/zed-industries/zed/discussions/46827) for an
example).
## Test plan
- [x] Manual test: configured `read_only_files: ["**/.venv/**"]`, opened
project, Go to Definition into `.venv`, created new file — save dialog
defaults to project root
- [x] Added `test_most_recent_active_path_skips_read_only_paths`
- [x] All existing workspace tests pass
Release Notes:
- Fixed save dialog defaulting to dependency directories (e.g. `.venv/`,
`node_modules/`) after using Go to Definition, when those directories
are configured as `read_only_files`.
---------
Co-authored-by: Lukas Wirth <lukas@zed.dev>
### Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the UI/UX checklist
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
#### Closes#52702
### Video
[Screencast from 2026-03-30
13-17-08.webm](https://github.com/user-attachments/assets/8b0f6927-04d2-406b-b7fd-064a730b2f86)
Release Notes:
- Fixed the Inline Assist button showing in the Project Diagnostics
toolbar when the agent is disabled.
Closes#47926
## Summary
When using a uv (or Poetry/PDM) workspace with multiple subprojects,
Python LSP servers (Pyright, Ruff, ty, etc.) are initialized with the
subproject directory as their workspace folder instead of the workspace
root. This happens because `PyprojectTomlManifestProvider::search()`
returns the first (innermost)
`pyproject.toml` found walking up the directory tree.
For example, in a uv workspace like:
main-project/
├── pyproject.toml # workspace root with [tool.uv.workspace]
├── uv.lock
├── packages/
│ └── project-api/
│ └── pyproject.toml # subpackage
└── projects/
└── project-a/
└── pyproject.toml # subpackage
Opening a file in `packages/project-api/` would register
`packages/project-api/` as the LSP workspace folder instead of
`main-project/`.
## Approach
The fix uses lockfile existence as a heuristic to detect workspace
roots. The updated `search()` method walks all ancestors (similar to
`CargoManifestProvider`) and:
- Tracks the **innermost** `pyproject.toml` as a fallback
- Tracks the **outermost** `pyproject.toml` that has a sibling lockfile
(`uv.lock`, `poetry.lock`, `pdm.lock`, or `Pipfile.lock`)
- Returns the outermost workspace root if found, otherwise falls back to
the innermost
This works within the existing `ManifestDelegate` interface (existence
checks only, no file content reading).
| Scenario | Result |
|---|---|
| uv workspace (root `pyproject.toml` + `uv.lock`) | Returns workspace
root |
| Poetry workspace (root `pyproject.toml` + `poetry.lock`) | Returns
workspace root |
| Simple project (single `pyproject.toml`, no lockfile) | Returns
project dir (unchanged) |
| Independent subprojects (no lockfile at any level) | Returns each
project's own dir (unchanged) |
Since the manifest provider is set at the Python **language** level,
this fix applies to all Python LSP servers (Pyright, Ruff, ty, etc.).
## Test plan
- [x] Added unit tests for `PyprojectTomlManifestProvider` covering all
scenarios above
- [x] Existing integration test
`test_running_multiple_instances_of_a_single_server_in_one_worktree`
passes (independent subprojects without lockfiles)
- [x] `cargo check -p languages` compiles cleanly
- [x] Manual testing with a real uv workspace (Pyright and Ruff both
receive correct workspace root)
Release Notes:
- Fixed Python LSP servers (Pyright, Ruff, etc.) using the wrong
workspace folder in uv/Poetry/PDM workspaces with multiple subprojects.
Hover popovers aren't immediately dismissed when the hover ends, but
this delay mechanism is currently causing popovers to sometimes still
briefly appear when the hover lasted too short to have triggered it.
Fixed by immediately cancelling the popover task when the hover ends
before the popover is displayed.
Most easily reproduced when setting the hover popover delay to a value
lower than the default 300ms (but it still happens sometimes with the
default value), because the popover hide delay is also 300ms. Here's it
with `"hover_popover_delay": 200`:
https://github.com/user-attachments/assets/6415d112-d8e0-4a87-9a79-a7ab559f20f2
After the fix:
https://github.com/user-attachments/assets/34782389-de4c-4a25-bd6e-4858b55028de
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed hover popovers sometimes briefly appearing after the hover
already ended.
---------
Co-authored-by: Lukas Wirth <lukas@zed.dev>
## Context
Red and blue have inherently low perceptual luminance in APCA — pure
default Lc 45 threshold. This caused ensure_minimum_contrast to lighten
them, washing out the color to pink/lavender.
When an application explicitly specifies an RGB value via
ESC[38;2;R;G;Bm,
that color should be rendered as-is. Named and indexed colors continue
to be contrast-adjusted since they come from theme mappings
[zed-colors-before.webm](https://github.com/user-attachments/assets/1adb8c95-5dec-4744-b6f8-38f085602ef4)
[zed-colors-after.webm](https://github.com/user-attachments/assets/1968db26-9c02-48cf-9542-86a6c39899c0)
Closes: #50396
<!-- What does this PR do, and why? How is it expected to impact users?
Not just what changed, but what motivated it and why this approach.
Link to Linear issue (e.g., ENG-123) or GitHub issue (e.g., Closes#456)
if one exists — helps with traceability. -->
## Self-Review Checklist
<!-- Check before requesting review: -->
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed true-color render in terminal
Remove statements that imply that `subtle` edit prediction mode can only
be used if Zed is set as an edit provider, despite this not being the
case.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes#53207
Release Notes:
- N/A
---------
Co-authored-by: morgankrey <morgan@zed.dev>
Bumps the `lsp-types` rev which contains patch for breaking change
introduced by upstream `typescript-go` repo
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed tsgo LSP
This PR revamps our feature flag system, to enable richer iteration.
Feature flags can now:
- Support enum values, for richer configuration
- Be manually set via the settings file
- Be manually set via the settings UI
This PR also adds a feature flag to demonstrate this behavior, a
`agent-thread-worktree-label`, which controls which how the worktree tag
UI displays.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
## Context
Fixes a bug on Windows where the Alt modifier key becomes stuck in the
pressed state after using Alt-Tab to switch away from and back to the
Zed application.
### Root Cause
In `crates/gpui_windows/src/events.rs`, when the window loses focus
during Alt-Tab, the WM_KEYUP event for the Alt key is never delivered to
the application. The pending modifier state in GPUI remains as
`alt=true`. When the window regains focus, there is no synchronization
point to reset this stale state.
## How to Review
In `handle_activate_msg()` (WM_ACTIVATE handler), when the window is
activated:
1. Reset the cached modifier tracking state (`last_reported_modifiers`
and `last_reported_capslock` set to `None`)
2. Query the actual current modifier state from Windows using
`GetKeyState()` APIs
3. Dispatch a `ModifiersChanged` event with the true state to
synchronize GPUI
## Self-Review Checklist
<!-- Check before requesting review: -->
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
I think it's difficult to add tests for this because it's
platform-dependent, but I recorded a video for comparison — sorry I only
have an anime‑style app to show the key presses.
<details>
<summary>Video</summary>
https://github.com/user-attachments/assets/b545178c-b360-4e2b-8d60-d6f95a0b4b79https://github.com/user-attachments/assets/bc4d41eb-6f42-4040-a588-0fc46c576db7
</details>
Closes#45485
Release Notes:
- Fixed Alt modifier key stuck after Alt-Tab on Windows. Modifier state
is now synchronized when the window regains focus, ensuring correct key
interpretation after window switching.
We had an internal report of soft wrap not working in git panel's commit
editor. Given the following settings:
```json
{
"languages": {
"Git Commit": {
"preferred_line_length": 80,
"soft_wrap": "preferred_line_length",
},
},
}
```
We would not soft-wrap in narrow viewports. As it turned out, the
problem was that we were always prefering a `preferred_line_length` as
our soft wrap boundary over the actual width of the editor.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [ ] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- Fixed git commits editor not respecting soft wrap boundaries.
- settings: Removed `"soft_wrap": "preferred_line_length"` in favour of
`"soft_wrap": "bounded"`. Soft wrap now always respects editor width
when it's enabled.
This is a follow up to: #54202
`thread_worktree_archive.rs` had its own
walk-up-and-remove-empty-parents pass that ran after archiving a Zed
managed git worktree. It duplicated logic that #54202 moved into
`Repository::remove_worktree` (which also adds remote support for this)
This PR deletes the dead code path and its associated tests.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A
Hit a flake in this test in
https://github.com/zed-industries/zed/actions/runs/24555985980/job/71792549618.
Ran for 1000 iterations and hit a second race, now also hopefully fixed.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Reversal tracking tests were removed in
https://github.com/zed-industries/zed/pull/53912. This PR re-adds them
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Fixes#54140
When `RulesLibrary::count_tokens` calls
`CloudLanguageModel::count_tokens` for Google cloud models, it does so
inside a `cx.update` closure, which holds a mutable borrow on the global
`AppCell`. The Google provider branch then called
`token_provider.auth_context(&cx.to_async())`, which created a new
`AsyncApp` handle and tried to take a shared borrow on the same
`RefCell` — causing a "RefCell already mutably borrowed" panic.
This only affects Google models because they are the only provider that
counts tokens server-side via an HTTP request (requiring
authentication). The other providers (Anthropic, OpenAI, xAI) count
tokens locally using tiktoken, so they never call `auth_context` during
`count_tokens`.
The fix makes `CloudLlmTokenProvider::auth_context` generic over `impl
AppContext` instead of requiring `&AsyncApp`. This allows the
`count_tokens` call site to pass `&App` directly (which reads entities
without re-borrowing the `RefCell`), while all other call sites that
already pass `&AsyncApp` (e.g. `stream_completion`, `refresh_models`)
continue to work unchanged.
Release Notes:
- Fixed a crash ("RefCell already mutably borrowed") that could occur
when counting tokens with Google cloud language models.
<img width="767" height="428" alt="Screenshot 2026-04-16 at 11 29 13 AM"
src="https://github.com/user-attachments/assets/e8b450fa-aefc-4dec-a286-b211bd492011"
/>
Add Claude Opus 4.7 (`claude-opus-4-7`) to the anthropic, bedrock, and
opencode provider crates.
Key specs:
- 1M token context window
- 128k max output tokens
- Adaptive thinking support
- AWS Bedrock cross-region inference (global, US, EU, AU)
Release Notes:
- Added Claude Opus 4.7 as an available language model
We currently, unfortunately, don't support title summarization and/or
regeneration for external agents, so we should hide that menu for them
so it's not misleading.
Release Notes:
- N/A
Kunall Banerjee