Closes https://github.com/zed-industries/zed/issues/43728
Release Notes:
- agent: Fixed a bug where the plan and edit files list would consume
the whole space of the thread if they were too long. They're now capped
by a max-height and scrollable.
This is still not perfect, but it reduces the shift that happens when
navigating between pages that have and don't have the table of contents.
It also tries to reduce the theme flicker that happens by moving its
loading to an earlier moment.
Release Notes:
- N/A
Adds clarifying examples and information about the `include_pattern`
parameter for the grep tool, analogous to the `path` parameter of the
read and list tools and others.
The lack of clarity led to unexpected agent behavior described in
#41189. This change is confirmed to improve the precision of Claude
Sonnet 4.5 when searching for code in an empty conversation (without it,
it leaves out the root directory in the query).
Closes#41189.
```
Release Notes:
- Clarify grep tool description to improve agent precision when using it with the `include_pattern` parameter
```
Closes#43040
Release Notes:
- Remove the end-of-support Gemini 1.5 model from the options.
- Remove the older Gemini 2.0 model from the options.
- Please let me know if you think it's better to keep it, as it is still
a usable model.
- Update the incorrect amounts for some input/output tokens.
- Update the default model to Gemini 2.5 Flash-Lite.
- Rename variant `Gemini3ProPreview` to `Gemini3Pro`
When this PR is merged, users will be able to select the following
Gemini models.
- 2.5 Flash
- 2.5 Flash-Lite
- 2.5 Pro
- 3 Pro
This PR adds the ability to collapse section in the docs sidebar (which
are persistent until you close the tab), and some design facelift to the
docs, which makes its design close to the site as well as polishing up
many elements and interactions (like moving the search to a modal and
making the table of content visible in smaller breakpoints).
<img width="600" height="2270" alt="Screenshot 2025-11-28 at 5 26@2x"
src="https://github.com/user-attachments/assets/3a8606c6-f74f-4bd2-84c8-d7a67ff97564"
/>
Release Notes:
- N/A
Release Notes:
- Added a keybinding to open the `settings.json` file when the Settings
UI is in focus: `cmd-,` (macOS) / `ctrl-,` (Windows / Linux).
- A tooltip was added to the `Edit in settings.json` button to show this
keybinding.
- shorten some phrases to (hopefully) retain attention better
- move the required field up to not bore people with scrolling
- call it a bug and not an issue to further emphasize we want feature
requests in Discussions where the product managers are
- remove the default from the WSL question since not everyone does
WSL or even Windows
Release Notes:
- N/A
---------
Co-authored-by: Miguel Raz Guzmán Macedo <miguel@zed.dev>
Co-authored-by: Kunall Banerjee <hey@kimchiii.space>
Closes#9648#9755
Release Notes:
- Added way to configure ESLint's working directories in settings. For
example:
`{"lsp":{"eslint":{"settings":{"workingDirectories":["./client","./server"]}}}}`
---------
Co-authored-by: Piotr Osiewicz <24362066+osiewicz@users.noreply.github.com>
Closes#26649
Release Notes:
- Improved git support for remotes handling. Git will now check pushRemote and pushDefault configurations before
falling back to branch remote.
---------
Co-authored-by: Jakub Konka <kubkon@jakubkonka.com>
There is still room for improvement here as `anchor(s)_in_excerpt` is
generally a bad API here due to it reseeking the entire excerpt tree
from the start on every call which we don't really need. But this at
least cuts the seeks down by a factor of 4 for now.
Release Notes:
- Improved performance of bracket colorization in large multibuffers
Introduces new "mapping point cursors" for the different display map
layers allowing one to map multiple points in increasing order more
efficiently than using the one shot operations.
This is used in the `BlockMap::sync` for `header_and_footer_blocks`
which spends a significant time in sumtree traversal due to repeatedly
transforming points between the different layers. This effectively turns
the complexity of those operations from quadratic in the number of
excerpts to linear, as we only go through the respective sumtrees once
instead of restarting from the start over and over again.
Release Notes:
- Improved performance for editors of large multibuffers with many
different files
These are long running foreground tasks that tend to not have pending
await points, meaning once we start executing them they will never
reschedule themselves, starving the foreground thread.
Release Notes:
- Improved git project diff responsiveness
Replace single quotes with double quotes when referencing the askpass
script name in the helper command. The Windows command processor
(cmd.exe) requires double quotes for proper string handling, as single
quotes are treated as literal characters.
Error I get when trying to open remote project:
<img width="396" height="390" alt="image"
src="https://github.com/user-attachments/assets/1538ee10-8efc-4f80-a867-b367908091b6"
/>
```
Zed Nightly 0.216.0
c2281779af
0.216.0+nightly.1965.c2281779af56bd52c829ccd31aae4eb82b682ebc
```
Release Notes:
- N/A
Without this new flag, the Sweep API will actually produce a combination
of UTF-16 and char indices, leading to incorrect edit positions.
Release Notes:
- N/A
We restructured `struct Repository` a bit so that it now stores a shared
task of `enum RepositoryState`. This way, we can now re-use it outside
of the git job queue when spawning a git job a standalone bg task. An
example here would be `git-blame` that does not require any file locks
and is not susceptible to git job ordering.
As a result of this change, loading (and modifying) a file that contains
a huge git history will no longer block other git operations on the repo
such as staging/unstaging/committing.
Release Notes:
- Improved overall git experience when loading buffers with massive git
history where they would block other git jobs from running (such as
staging/unstaging/commiting). Now, git-blame will run separately from
the git job queue on the side and the buffer with blame hints when
finished thus unblocking other git operations.
Co-authored-by: Cole Miller <cole@zed.dev>
This reduces hangs on windows when we have many tasks queued up on the
main thread that yield a lot.
Release Notes:
- Reduced hangs on windows in some situations
Introducing this little popover here that's aimed at better
communicating what Zed's built-in edit prediction feature is and how
much people can get of it for free by purely just signing in.
<img width="600" height="1914" alt="Screenshot 2025-11-27 at 9 50@2x"
src="https://github.com/user-attachments/assets/a7013292-f662-4cae-9a6f-0e69a4a4fa1d"
/>
Release Notes:
- N/A
- **release_channel: Do not use prerelease channel for build id**
Prerelease channel specifiers always compare as less than to
non-prerelease, which led to 2 auto-update bugs fixed in
https://github.com/zed-industries/zed/pull/43595 and
https://github.com/zed-industries/zed/pull/43611.
We'll use a dot-delimited build specifiers in form:
release-channel.build_number.sha1 instead
- **auto_update: Do not display full build metadata in update
notification**
Release Notes:
- N/A
Sometimes we are unable to receive messages at all from an agent. This
puts on upper bound on the `initialize` call so we can at least give a
message to the user that something is wrong here.
30s might feel like too long, but I wanted to avoid some false positives
in case there was something an agent needed to do at startup. This will
still communicate to the user at some point that something is wrong,
rather than leave them waiting forever with no signal that something is
going wrong.
Release Notes:
- agent: Show an error message to the user if we are unable to
initialize an ACP agent in a reasonable amount of time.
Update the way that both
`search::buffer_search::BufferSearchBar.replace_next` and
`search::buffer_search::BufferSearchBar.replace_all` are registered as
listeners, so that we don't require the replacement editor to be focused
in order for these listeners to be active, only requiring the
replacement mode to be active in the buffer search bar.
This means that, even if the user is focused on the buffer editor, if
the "Replace Next Match" or "Replace All Matches" buttons are clicked,
the replacement will be performed.
Closes#42471
Release Notes:
- Fixed issue with buffer search bar where the replacement buttons
("Replace Next Match" & "Replace All Matches") wouldn't work if search
bar was not focused
Fixes ZED-3P9
We only clamped the end which for a completely wrong input could cause
us to construct a reversed range which will end up underflowing later
on.
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Zeta evals now include a character n-gram metric adapted for multi-edit diffs (“delta chrF”). It works as follows:
1. Reconstruct the original, golden (expected), and actual texts from unified diffs.
- "original": the text before any edits
- "golden": the text after applying the expected edits
- "actual": the text after applying the actual edits
2. Compute n-gram count deltas between original→golden and original→actual.
- n-grams are computed as in chrF (max n=6, whitespace ignored).
3. Compare these deltas to assess how well the actual edits match the expected edits.
- As in standard chrF, classify n-grams as true positives, false positives, and false negatives, and report the F-beta score with beta=2.
Release Notes:
- N/A
There are 3 factors:
1. The Preview channel endpoint does not propagate versions with build
identifier (which we oh-so-conveniently store in pre-release field of
semver).
2. Preview build, once fetched, sees it's version *with* build
identifier (as that's baked into the binary).
3. Auto update logic treats versions with pre-release version as less
than versions without pre-release version.
This in turn makes any Preview client see itself as versioned like
0.214.4-123-asdf1234455311, whereas the latest version on the endpoint
is 0.214.4. Thus, the endpoint version is always more recent than the
client version, causing an update loop.
The fix is to ignore build identifier when comparing versions of
non-nightly channels. This should still let us introduce changes to
auto-update behavior in minor releases in the future.
Closes#43584
Release Notes:
- (Preview only): Fixed an update loop with latest Preview update.
CLOSES #39042
This is a reopening of #34985+.
_Original descrioption_:
In Wayland, the client implement key repeat themself. In Zed this is
ultimately handled by the gpui crate by inserting a timer source into
the event loop which repeat itself if the key is still held down [1].
But it seems the processing of the repeated key event happen
synchronously inside the timer source handler, meaning the effective
rate become slightly lower (since the repeated timer is scheduled using
the 1/rate as delay).
I measured the event processing time on my laptop and it's typically
around 3ms, but sometimes spiking at 10ms. At low key repeat rates this
is probably not _very_ noticeable. I see the default in Zed is set to a
(measly) 16/s, but I assume most systems will use something closer to
25, which is a 40ms delay. So ~3ms is around 7.5% of the delay. At
higher rate the discrepancy become worse of course.
I can visible notice the spikes, and doing some crude stopwatch
measurements using gedit as a reference I can reproduce around 5-10%
slower rates in Zed.
IMO this is significant enough to warrant improving, especially since
some people can get quite used the repeat rate and might feel something
being "off" in Zed.
~~The suggested fix simply subtract the processing time from the next
delay timer.~~
[1] https://github.com/olejorgenb/zed/blob/32df726f3b7fa83e7399f6629c59e0a3f3fff125/crates/gpui/src/platform/linux/wayland/client.rs#L1355
Release Notes:
- Improved Wayland (Linux) key repeat rate precision
Adds a `trigger` field to the zeta1/zeta2 prediction requests so that we
can distinguish between editor, diagnostic, and zeta-cli requests.
Release Notes:
- N/A
Many prediction requests end up being rejected early without ever being
set as the current prediction. Before this change, those cases weren’t
reported as rejections because the `request_prediction_with_*` functions
simply returned `Ok(None)`.
With this update, whenever we get a successful response from the
provider, we will return at least the `id`, allowing it to be properly
reported. The request now also includes a “reject reason,” since the
different variants carry distinct implications for prediction quality.
All of these scenarios are now covered by tests. While adding them, I
also found and fixed a bug where some cancelled predictions were
incorrectly being set as the current one.
Release Notes:
- N/A
---------
Co-authored-by: MrSubidubi <dev@bahn.sh>
Danilo Leal