## Context
This PR adds a function to sync up the title bar state if a workspace is
moved to a new window through the threads sidebar. This is needed
because otherwise the padding for the traffic control buttons on macOS
wasn't being taken into consideration.
## How to Review
- Move a project to a new window from the sidebar
- See that the title bar for that new window looks normal, as it should
---
- [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 some issues I ran into when running this on a remote machine.
## 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:
- N/A
We can't add a current_dir when in remote scenarios in the same place,
it gets added elsewhere.
Reused the same worktree selection logic, and make sure this gets passed
to the RPC call instead.
This fixes something that hasn't made it to preview yet.
Release Notes:
- N/A
Context
The git panel's context menu caused visual jitter (flickering/jumping)
when opened via right-click on a tracked file. The root cause was in
`dispatch_context()`: it used `self.focus_handle == focused` to check if
the panel itself was directly focused, but when a context menu opened,
focus moved to the menu (a child element), causing the `"menu"` and
`"ChangesList"` key contexts to be dropped. This triggered a re-render
with different keybindings, which re-added them, creating a loop of
jitter.
The fix replaces the direct focus equality check with
`self.focus_handle.contains_focused(window, cx)`, which returns `true`
when any child element (including the context menu) holds focus within
the panel's focus tree. This is consistent with how other panels
(project panel, outline panel, collab panel) handle focus-based dispatch
contexts.
Closes#51813
## Demo
**Before fix:**
https://github.com/user-attachments/assets/e18d49b2-72a6-4411-8ec5-519e36628f29
**After fix:**
https://github.com/user-attachments/assets/94c936d2-1e81-4d28-a86a-8b1ed76ddde1
## How to Review
This is a small, focused change in a single file:
`crates/git_ui/src/git_panel.rs`.
1. **The fix** (~line 974): `dispatch_context()` method — the old code
checked direct focus equality (`self.focus_handle == focused`), the new
code uses `self.focus_handle.contains_focused(window, cx)` and
restructures the conditionals so `CommitEditor` is checked first via
`if/else if`.
2. **The test** (~line 7871): `test_dispatch_context_with_focus_states`
— verifies 4 focus state transitions: commit editor focused, changes
list focused, back to commit editor, and back to changes list. Each case
asserts the correct key contexts are present/absent.
## 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)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed git panel context menu jitter caused by keybinding dispatch
context flickering when right-clicking on tracked files (#51813)
## Context
When `git_panel.status_style` is set to `label_color`, the git panel
should indicate file status through text color only, without showing
status icons. However, the icons were still being rendered regardless
of this setting.
The root cause: `git_status_icon(status)` was called unconditionally
in `render_status_entry`, with no check against the current
`status_style` value.
Closes#51714
## How to Review
Single change in `crates/git_ui/src/git_panel.rs` —
`render_status_entry` function. The icon render call is now wrapped
in `.when(status_style != StatusStyle::LabelColor, ...)`.
Release Notes:
- Fixed git panel status icons still showing when
`git_panel.status_style`
is set to `label_color`
Bug:
<img width="2559" height="1357" alt="image"
src="https://github.com/user-attachments/assets/3cad9f96-8b5c-4554-8b9f-f931026bdfe1"
/>
Fix:
<img width="1512" height="801" alt="image"
src="https://github.com/user-attachments/assets/99f36d11-a8b2-412a-ab73-18200f6594d2"
/>
## Context
I was getting some leak detection failures in evals and tracked it down
to these entities getting passed into observe/subscribe callbacks and
causing cycles.
Release Notes:
- N/A
Co-authored-by: Lukas Wirth <me@lukaswirth.dev>
## Context
This PR builds on https://github.com/zed-industries/zed/pull/52047,
adding support for the new behavior to the keymap editor. The primary
change is replacing usages of `NoAction` with `Unbind` when updating the
users keymap, i.e. deleting/editing default bindings.
This PR does not completely solve the UI challenge of `Unbind`. For now,
we just don't show Unbind entries in the UI, and mark unbound
keybindings as unbound (shown in screenshot).
<img width="3854" height="2230" alt="Screenshot 2026-03-23 at 11 49
42 AM"
src="https://github.com/user-attachments/assets/843856c6-2c94-47c1-be44-21becfdf467e"
/>
## How to Review
- Check behavior changes in keymap updates
- Check UI changes and filters in keymap editor
<!-- Help reviewers focus their attention:
- For small PRs: note what to focus on (e.g., "error handling in
foo.rs")
- For large PRs (>400 LOC): provide a guided tour — numbered list of
files/commits to read in order. (The `large-pr` label is applied
automatically.)
- See the review process guidelines for comment conventions -->
## 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:
- N/A or Added/Fixed/Improved ...
## Context
Recently, we put bsn in the list of "macros to not reparse as rust"
(#51353).
We have since updated tree-sitter-rust to better handle the `@"string"`,
`:"string"`, etc cases, and can un-list bsn from the special casing.
on main, the highlighting currently looks like this for `bsn` and for
`not` (represents "other macros"). Note the `:"player.bsn"` string is
re-parsed as rust tokens.
<img width="1728" height="1084" alt="Screenshot 2026-03-23 at 9 33
35 AM"
src="https://github.com/user-attachments/assets/94ea5a6d-228b-44dd-857b-c41edcbd9af0"
/>
This pr, plus the tree-sitter-rust updates results in better handling by
default.
<img width="1728" height="1084" alt="Screenshot 2026-03-23 at 11 38
47 AM"
src="https://github.com/user-attachments/assets/fb6f71e3-7bcf-4c06-8e84-ce0823010b2e"
/>
Release Notes:
- N/A
This PR fixes a bug where archiving a thread could cause the *next*
thread to be opened in the current workspace, even though it belonged to
a different worktree.
Release Notes:
- N/A
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
We've decided to remove rodio backend for audio as we didn't have time
to get it to a nice spot. I personally believe we should eventually
re-land it (the composable pipelines are quite nice!), but given that we
need audio to work, this seems like the best way forward. We won't have
to worry about all the ways in which the legacy pipeline interoped with
rodio's.
## 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:
- audio: Experimental rodio audio backend has been removed.
---------
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
Co-authored-by: Yara <git@yara.blue>
## Context
When typing on the read-only queued message editor, we push the content
to the main, editable message editor. However, the cursor position was
being reset when that happened. This PR fixes that by tracking the
cursor offset when doing that, also accounting for the case where there
could be pre-existing content in the message editor.
## How to Review
- Queue a message in the agent panel
- Place your cursor somewhere in the middle of your queued message and
type something
- See how the cursor position is preserved once it goes down to the main
message editor
## 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:
- Agent: Fix cursor position being reset when editing queued messages.
## Context
This PR removes the behavior in which we turn the agent panel full
screen on worktree creation, even if there aren't any files open.
Changing the width behavior of the panel will ultimately solve this, and
auto-full screening it turned out to be surprising. Also, in reaction to
this, I've added an icon button in the agent panel's toolbar when in
full screen so it's easier to get out of it. Eventually though, we can
probably remove it (also because of the flexible width work).
## How to Review
- Create a new worktree from a thread and see the panel with the same
width
- See the "disable full screen" icon button in the toolbar when you full
screen it
---
- [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
This PR:
- Moves the "remove workspace" option into the context menu, so that
it's not swapping places with the + button in some circumstances. This
matches codexs behavior. Another solution to the problem could be having
the add button remain even when the "new thread" option exists.
- Adds a "go to workspace" button to the header
## 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:
- N/A
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
This PR adjusts the default panel layout for anyone on the agent v2
feature flag.
Note that this changes the right status bar items to show in reverse
priority order, and then adjusts priorities so that the "project
management" buttons appear to the right, and the outline panel appears
to the far left. The reversal "cancels out" most of the priority
changes, except the outline panel and the collab panel.
## 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:
- Swapped the order of the collab and outline status bar buttons
### Context
This fixes a bug where the sidebar would show a newly created git
worktree thread as its own project while the recently created workspace
is loading its git state. The fix is adding project APIs to await the
initial worktree store scan and then each git repo initial snapshot;
then awaiting on them before adding the new workspace to the
multi-workspace.
### Architecture:
I added the `Worktree::Remote::wait_for_snapshot` API to
`Worktree::Local` to enable `WorktreeStore` to await for both remote and
local projects until there's an initial scan. The `WorktreeStore` uses
the watcher pattern so it can update the initial scan state whenever
visible worktrees are added or removed from the store.
Before you mark this PR as ready for review, make sure that you have:
- [x] Added solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
Co-authored-by: cameron <cameron.studdstreet@gmail.com>
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
## Context
This will make it easier to diagnose issues if any come up when
migrating thread metadata for the sidebar
## How to Review
<!-- Help reviewers focus their attention:
- For small PRs: note what to focus on (e.g., "error handling in
foo.rs")
- For large PRs (>400 LOC): provide a guided tour — numbered list of
files/commits to read in order. (The `large-pr` label is applied
automatically.)
- See the review process guidelines for comment conventions -->
## 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:
- N/A
Currently, Zed always places three fixed window buttons (Minimize,
Maximize and Close) on the right side of the window in a fixed order
ignoring any user configuration or desktop environment preference (like
elementary).
This PR adds support for GNOME-style layout strings (`minimize:close`).
By default, we pull it from the gsettings portal, but we also allow
manual configuration via `title_bar.button_layout` config key.
<img width="1538" height="797" alt="image"
src="https://github.com/user-attachments/assets/5db6bfa2-3052-4640-9228-95c37f318929"
/>
Closes#46512
I know it's a relatively large PR for my first one and I'm new to Rust.
So, sorry if I've made any huge mistakes. I had just made it for
personal use and then decided to try to clean it up and submit it.
I've tested with different configs on Linux. Untested on other
platforms, but should have no impact.
If it's not up to par, it's okay, feel free to close :)
Release Notes:
- Added support for GNOME's window buttons configuration on Linux.
---------
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
MacWindow methods like activate(), zoom(), toggle_fullscreen(),
resize(), prompt(), and titlebar_double_click() capture the raw
native_window pointer and spawn detached async tasks on the foreground
executor. If the window is closed between the spawn and execution (e.g.
the user clicks the close button), the task sends AppKit messages to a
closed NSWindow, which throws an ObjC exception in
_changeKeyAndMainLimitedOK: and aborts the process.
Add an isVisible check inside each deferred task before sending messages
to the native window. After [NSWindow close], isVisible returns NO, so
the guard prevents the crash. Both the check and the subsequent AppKit
call execute synchronously within the same main-thread task, so there is
no TOCTOU race.
For prompt(), the else branch releases the alert object to avoid leaking
it; the oneshot sender inside the completion block is dropped, which
cancels the channel and propagates as an error to the caller.
Closes ZED-5TN
Release Notes:
- N/A or Added/Fixed/Improved ...
## Context
Before this change the branch diff would compare with local main instead
of upstream main. This causes confusion because it's a different diff
than the branch diff action would show users, and cause agents to
potentially review a much larger diff then it has too.
## 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:
- N/A
Follow-up to the terminal permission params refactor, addressing
@benbrandt's [review
feedback](https://github.com/zed-industries/zed/pull/51782#pullrequestreview-2926899804)
about string formatting leaking into the UI layer.
## Changes
### Outcome construction moved to acp_thread
- Added `PermissionOptionChoice::build_outcome(is_allow)` — builds a
`SelectedPermissionOutcome` from a choice, attaching
`SelectedPermissionParams::Terminal` when the choice carries
`sub_patterns`.
- Added
`PermissionOptions::build_outcome_for_checked_patterns(checked_indices,
is_allow)` — handles the `DropdownWithPatterns` per-command checklist
flow, returning `None` when zero patterns are checked (degrading to
once).
The UI's `authorize_with_granularity` no longer does any
`format!("always_allow:{}",...)` string formatting or
`SelectedPermissionParams` construction.
### `option_kind` folded into `SelectedPermissionOutcome`
`SelectedPermissionOutcome` now carries `option_kind:
acp::PermissionOptionKind`, eliminating the separate parameter that was
threaded through the entire `authorize_tool_call` chain:
```
ThreadView::authorize_tool_call
→ Conversation::authorize_tool_call
→ AcpThread::authorize_tool_call
```
Every signature in this chain dropped one parameter.
### `SelectedPermissionParams` removed from UI imports
The type is now only referenced by `acp_thread` (construction) and
`agent` (consumption). The UI passes `SelectedPermissionOutcome`
opaquely.
Release Notes:
- N/A
## Context
Zed's semantic token highlighting does not cover all token types
returned by language servers, so the highlighting looks fairly primitive
compared with tree-sitter highlighting, especially for Python language
servers. This PR adds some global and Python-specific rules for better
highlighting.
I need to admit that the built-in Python language servers currently have
weak semantic highlighting implementations. Pylance, the closed-source
Python language server from Microsoft, provides the best highlighting
for now, but I think ty will do better, even though it still has a long
way to go.
## How to Review
Basically, this is a rule-adding change. Some rules are made global, and
some are made Python-specific.
## 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:
- Improved semantic token highlighting for parameters and Python
`close_session` removed the session from `self.sessions` before calling
`save_thread`, but `save_thread` immediately looks up the session in
`self.sessions` and returns early if it's not found. This meant the
explicit final save was always a no-op.
Additionally, when the local `session` variable was dropped at the end
of the closure, it cancelled any in-flight save from a prior observe
callback by dropping `session.pending_save`.
So on session close:
- The explicit final save was a guaranteed no-op
- Any in-flight save from an earlier observation was cancelled
**Fix:** Call `save_thread` while the session is still in the map, then
extract the `pending_save` task and return it so the caller can actually
await the save completing.
**Test:** Added `test_close_session_saves_thread` which sets a draft
prompt without a `run_until_parked()` before `close_session`, so the
only way the data gets persisted is if `close_session` itself performs
the save. Also strengthened the existing `test_save_load_thread`
similarly.
Closes AI-95
Release Notes:
- Fixed a bug where closing an agent thread could lose unsaved changes
(e.g. draft prompts) made in the same frame.
In `remove_workspace`, the removed `Entity<Workspace>` could still have
a pending `serialize_workspace` throttle timer (200ms). When that timer
fired, `serialize_workspace_internal` would write the old `session_id`
back to the DB — undoing the removal. On next restart, the workspace
would reappear.
The race window opens whenever any state change (worktree change,
breakpoint change, etc.) triggers `serialize_workspace` within 200ms
before `remove_workspace` is called.
**Fix**: Before the DB cleanup task, `update` the removed workspace
entity to:
1. `session_id.take()` — so any in-flight serialization writes
`session_id: None`
2. `_schedule_serialize_workspace.take()` — cancel the pending throttle
timer
3. `_serialize_workspace_task.take()` — cancel any actively running
serialization task
This mirrors what `remove_from_session` already does (clearing
`session_id`), but `remove_workspace` was missing it.
Release Notes:
- Fixed a bug where a removed workspace could reappear on next launch
due to a serialization race.
Replace the typewriter-brand + random-hash naming scheme for
auto-generated git worktree branches with an adjective-noun approach
(e.g. `swift-falcon`, `calm-meadow`).
- Removed the ~650-entry typewriter brand name list and hash generation
- Added 235 adjectives and 228 nouns (~53,500 unique combinations)
- Branch names are now checked for exact-match collisions against
existing branches
- Updated error messaging
Closes AI-101
Release Notes:
- Improved auto-generated git worktree branch names to use friendlier
adjective-noun pairs (e.g. `swift-falcon`) instead of typewriter brand
names with random hashes.
## Context
`TerminalBounds::num_lines()` uses `floor(height / line_height)` to
compute the terminal grid row count. When the height is derived from `N
* line_height` (as it is for inline/embedded terminals in the Agent
Panel), IEEE 754 float32 arithmetic can produce `N - epsilon` instead of
`N`, causing `floor()` to return `N - 1`. This makes the terminal grid
one row too small, leaving the first line of output in invisible
scrollback (since `display_offset = 0`). The same issue applies to
`num_columns()`.
The fix adds a small tolerance (`0.01`) before flooring, which absorbs
float precision errors without affecting genuine fractional results.
Closes#51609
## How to Review
Small PR — focus on the tolerance value (`0.01`) in `num_lines()` and
`num_columns()` in `crates/terminal/src/terminal.rs`. The two new tests
(`test_num_lines_float_precision`, `test_num_columns_float_precision`)
verify the fix across 1,000+ float combinations that previously
triggered the bug.
## 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
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Fixed the first line of terminal output sometimes missing in Agent
Panel terminal blocks.
Fixes#51490
## Problem
When `code_actions_on_format` is configured (e.g.,
`source.fixAll.phpcs`) and `formatter` is not explicitly set, code
action edits are applied but then reverted during format-on-save.
The root cause is in the formatting pipeline's error handling. Code
actions and formatters share a single transaction via `.chain()`. When
`Formatter::Auto` resolves to `Formatter::Prettier` (PHP defaults to
`prettier.allowed: true`), and prettier is not installed or fails, the
error propagates via `?` out of the formatter loop, aborting the entire
`format_buffer_locally` function. This causes the formatting transaction
- which already contains the successfully applied code action edits - to
be lost.
The workaround of setting `"formatter": []` works because it removes all
formatters from the chain, so no formatter can fail and trigger the
abort.
## Fix
Change error handling for individual formatters (Prettier, External
command, Language Server) from propagating errors via `?` to logging the
error and continuing to the next formatter. This treats "formatter
failed" as a no-op for that specific formatter rather than aborting the
entire pipeline.
This matches the existing pattern used when `LanguageServer(Current)`
finds no server that supports formatting - it already logs and
`continue`s rather than erroring.
## Testing
- `cargo check -p project` passes
- `cargo fmt -p project -- --check` passes
Release Notes:
- Fixed `code_actions_on_format` edits being reverted when `formatter:
"auto"` resolves to an unavailable formatter.
---------
Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
## Context
We seemed to have disabled it in #28056, because our agent did not
support parallel tool calls at the time.
## 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:
- N/A or Added/Fixed/Improved ...
## Summary
Fixes#48289
Closes https://github.com/zed-industries/zed/issues/52021
Diagnostic error markers (red/yellow dots) in the project panel file
tree persisted after errors were resolved. Three root causes:
- **WorktreeUpdatedEntries ignored** — when files changed on disk (e.g.
`yarn install`), stale diagnostic summaries were never cleared. Added
`invalidate_diagnostic_summaries_for_updated_entries()` to clear
summaries for `Removed`/`Updated`/`AddedOrUpdated` paths.
- **Missing DiagnosticsUpdated emission on server stop** —
`stop_local_language_server()` cleared summaries and sent proto messages
but never emitted `LspStoreEvent::DiagnosticsUpdated`, so the project
panel never refreshed.
- **Buffer reload not handled** — reloading a buffer from disk did not
clear stale summaries. Added `BufferEvent::Reloaded` handler.
All three paths also send zeroed `UpdateDiagnosticSummary` proto
messages to downstream collab clients.
## Test plan
- [x] `./script/clippy` passes
- [x] `cargo test -p project -p project_panel -p worktree` passes (319
tests, 0 failures)
- [x] 4 new tests added:
- `test_diagnostic_summaries_cleared_on_worktree_entry_removal`
- `test_diagnostic_summaries_cleared_on_worktree_entry_update`
- `test_diagnostic_summaries_cleared_on_server_restart`
- `test_diagnostic_summaries_cleared_on_buffer_reload`
- [x] Manual testing: error markers clear when files change on disk
- [x] Manual testing: error markers clear on LSP restart
Release Notes:
- Fixed stale diagnostic data persisting after file reloads, server
restarts and FS entry removals
## Context
Deepseek seems to have started to return empty text deltas, which causes
issues cause their upstream API does not like empty text blocks to be
included in their request.
In practice this caused issues like `An assistant message with
'tool_calls' must be followed by tool messages responding to each
'tool_call_id'..., because the payload looked like this:
```
User {
content: "Call the now tool a bunch of times",
},
Assistant {
content: Some(
"",
),
},
Tool {
content: "The current datetime is 2026-03-23T13:21:24.084848+00:00.",
tool_call_id: "call_00_iozCfHXJAPR13XwHiAwJ9OEw",
},
```
Now we filter out that empty text, which seems to fix the issue. This
matches our implementation for OpenAI
Closes#51854
<!-- 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. -->
## How to Review
<!-- Help reviewers focus their attention:
- For small PRs: note what to focus on (e.g., "error handling in
foo.rs")
- For large PRs (>400 LOC): provide a guided tour — numbered list of
files/commits to read in order. (The `large-pr` label is applied
automatically.)
- See the review process guidelines for comment conventions -->
## 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:
- deepseek: Fixed issue with tool calling (`An assistant message with
'tool_calls' must be followed by tool messages responding to each
'tool_call_id'...`)
## Context
Fixes https://github.com/zed-industries/zed/issues/51921
`await` was grouped under the generic `@keyword` capture in the
tree-sitter highlights queries for JavaScript, TypeScript, and TSX. It
should be `@keyword.control`, since it's a control flow keyword — same
as `return`, `yield`, `throw`, `break`, etc., which were already
correctly classified.
This aligns Zed's highlighting with VSCode's behavior, where `await` is
tokenized as `keyword.control`.
## How to Review
Three single-line moves across highlights files — `await` removed from
the `@keyword` list, added to the `@keyword.control` list:
1. `crates/languages/src/javascript/highlights.scm`
2. `crates/languages/src/typescript/highlights.scm`
3. `crates/languages/src/tsx/highlights.scm`
## 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
## Image :
<img width="3072" height="1728" alt="Screenshot from 2026-03-20
22-15-40"
src="https://github.com/user-attachments/assets/9e849d4d-dad4-4fa0-b3b8-8e633f96c585"
/>
Release Notes:
- Fixed `await` keyword not being highlighted as `keyword.control` in
JavaScript, TypeScript, and TSX files
---------
Co-authored-by: Kunall Banerjee <hey@kimchiii.space>
## Context
This should allow the UI to update faster with stateful git operations
like branch changes, commits, etc. This will also allow Zed to
recognized when a new project belongs to a branch sooner than before.
## 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:
- git: Speed up UI responsiveness to branch/worktree changes
Before you mark this PR as ready for review, make sure that you have:
- [x] Added a solid test coverage and/or screenshots from doing manual
testing
- [x] Done a self-review taking into account security and performance
aspects
- [x] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Per Opencode's website:
> Zen gives you access to a curated set of AI models that OpenCode has
tested and benchmarked specifically for coding agents. No need to worry
about inconsistent performance and quality, use validated models that
work.
> - [x] Testing select models and consulting their teams
> - [x] Working with providers to ensure they're delivered properly
> - [x] Benchmarking all model-provider combinations we recommend
There are so many models available, but only a few work well with coding
agents. Most providers configure them differently with varying results.
The models under the Zen umbrella typically have a more reliable
token(s) per second speed with minimal outages. The opencode ecosystem
has improved my workflow if not many others' !
Release Notes:
- Added [Opencode Zen](https://opencode.ai/zen) to list of providers
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
## Context
We were seeing some off by one errors because we weren't counting in
progress as either pending or completed.
Release Notes:
- acp: Fix plan stats showing all tasks as done when items were still in
progress.
## Context
If you went back and edited the first message, we were resetting the
provisional title, even though the title had previously already been set
from a summarization pass.
In order to achieve this, we moved the default thread title logic up the
UI layer, and allowed AcpThreads to have an optional title.
## 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:
- N/A
---------
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Closes#43162
Implements the OAuth 2.0 Authorization Code + PKCE authentication flow
for remote MCP servers using Streamable HTTP transport, as specified by
the [MCP auth
specification](https://modelcontextprotocol.io/specification/2025-03-26/basic/authorization).
Previously, connecting to a remote MCP server that required OAuth would
silently fail with a timeout — the server's 401 response was never
handled. Now, Zed detects the 401, performs OAuth discovery, and guides
the user through browser-based authentication.
Step-up authentication and pre-registered clients are not in scope for
this PR, but will be done as follow-ups.
## Overview
- **401 detection** — When the HTTP transport receives a 401 during
server startup, it surfaces a typed `TransportError::AuthRequired` with
parsed `WWW-Authenticate` header info.
- **OAuth discovery** — Protected Resource Metadata (RFC 9728) and
Authorization Server Metadata (RFC 8414) are fetched to locate the
authorization and token endpoints.
- **Client registration** — Zed first tries CIMD (Client ID Metadata
Document) hosted at `zed.dev`. If the server doesn't support CIMD, falls
back to Dynamic Client Registration (DCR).
- **Browser flow** — A loopback HTTP callback server starts on a
preferred fixed port (27523, listed in the CIMD), the user's browser
opens to the authorization URL, and Zed waits for the callback with the
authorization code.
- **Token exchange & persistence** — The code is exchanged for
access/refresh tokens using PKCE. The session is persisted in the system
keychain so subsequent startups restore it without another browser flow.
- **Automatic refresh** — The HTTP transport transparently refreshes
expired tokens using the refresh token, and persists the updated session
to the keychain.
## UI changes
- Servers requiring auth show a warning indicator with an
**"Authenticate"** button
- During auth, a spinner and **"Waiting for authorization..."** message
are shown
- A **"Log Out"** option is available in the server settings menu for
OAuth-authenticated servers
- The configure server modal handles the auth flow inline when
configuring a new server that needs authentication.
Release Notes:
- Added OAuth authentication support for remote MCP servers. Servers
requiring OAuth now show an "Authenticate" button when they need you to
log in. You will be redirected in your browser to the authorization
server of the MCP server to go through the authorization flow.
---------
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
* Add `editor::RewrapOptions::line_length` to, optionally, override the
line length used when rewrapping text.
* Update `editor::Editor::rewrap_impl` to prefer
`editor::RewrapOptions::line_length`, when set.
* Add a `line_length` field to the `vim::rewrap::Rewrap` action.
* Update the `:reflow` vim command with `vim::command::VimCommand::args`
so as to be able to parse the provided argument as `usize`, ensuring
that no effect is taken if the argument can't be parsed as such.
Release Notes:
- N/A
## What
Fix image context mentions always showing the generic label `Image`
instead of the actual filename when pasting from Finder or picking via
the `+` → Image button in the Agent Panel.
## Why
`insert_images_as_context` hardcoded the crease label to
`MentionUri::PastedImage.name()` (`"Image"`) for every image, regardless
of whether it originated from a named file. Both code paths that load
images from file paths — `paste_images_as_context` and
`add_images_from_picker` — discarded the filename before passing images
to the shared insert function.
## Fix
- `agent_ui/src/mention_set.rs`: Changed `insert_images_as_context` to
accept `Vec<(gpui::Image, SharedString)>` instead of `Vec<gpui::Image>`,
using the provided name as the crease label. In
`paste_images_as_context`, extract `file_name()` from each path and pair
it with the loaded image. Raw clipboard images (screenshots, copy from
image editors) continue to use `"Image"` as there is no filename.
- `agent_ui/src/message_editor.rs`: Same fix for
`add_images_from_picker` — extract `file_name()` from each selected path
and pass it alongside the image.
Closes#52079
## Test Plan
- [x] `cargo build -p agent_ui` compiles clean
- [x] `cargo fmt --all -- --check` format check
- [x] Manual verification of:
- [x] Copy an image file in Finder (`Cmd+C`), paste into Agent Panel —
mention shows actual filename
- [x] `+` → Image → pick a file — mention shows actual filename
- [x] Screenshot paste (`Cmd+Shift+4`) still shows `Image`
- [x] Regular text paste still works
## Screenshots
<img width="638" height="569" alt="image"
src="https://github.com/user-attachments/assets/859d852c-66f6-4faa-a5fe-59bd34cd3d85"
/>
---
Release Notes:
- Fixed image context mentions always showing `Image` instead of the
actual filename when pasting from Finder or using the image picker in
the Agent Panel
For now, the only source of logprobs is `./run.py infer`, which writes
them to under the `.predictions` struct. `ep score` copies these values
to `.score`. There is some duplication (same value stored in two
places), which is unfortunate, but can't be fixed without reworking how
scores are stored.
Release Notes:
- N/A
Danilo Leal