Currently we always compute breadcrumbs and sticky headers on every
editor paint which is not cheap to do especially in bigger files, moving
this off to be computed on event handling where they change and then
caching them can save serveral milliseconds per render in bigger files.
This also puts matching brackets refreshing and document highlights on a
background task, as this tends to block the main task for prolonged time
as well.
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Thanks to @jackTabsCode for the original commit that I based my work off
of.
See: #27703
Release Notes:
- Fixed clicking "Collapse All" from context menu on a project root
collapsing all project roots instead of that one.
- Added new `project panel: collapse selected entry and children` action
that collapses the selected directory and all its subdirectories.
- Added "Collapse All" option to the context menu for all directories,
not just project roots.
---------
Co-authored-by: Jack T <jack@jackt.space>
Co-authored-by: Smit Barmase <heysmitbarmase@gmail.com>
For: https://github.com/zed-industries/zed/pull/47328
Release Notes:
- Fixed an issue where selecting "Collapse All" on the root directory or
triggering the collapse all action would sometimes collapse the entire
root instead of keeping it expanded when there's a single worktree.
## Problem
Zed's terminal permission system validates commands using regex patterns
against string representations of parsed commands. The
`SimpleCommand::to_string()` method in brush-parser preserves the raw
lexical form including shell quotes, causing pattern matching to fail on
obfuscated commands.
For example, all of these are semantically equivalent to `rm -rf /` in
POSIX shells, but produced different strings from `to_string()` that
bypassed the hardcoded deny pattern `rm\s+(-[rf]+\s+)*/\s*$`:
| Input | `extract_commands()` returned | Matched deny pattern? |
|---|---|---|
| `rm -rf /` | `"rm -rf /"` | ✅ Yes |
| `rm -rf '/'` | `"rm -rf '/'"` | ❌ No |
| `'rm' -rf /` | `"'rm' -rf /"` | ❌ No |
| `r'm' -rf /` | `"r'm' -rf /"` | ❌ No |
| `rm -r'f' /` | `"rm -r'f' /"` | ❌ No |
| `rm -rf \/` | `"rm -rf \\/"` | ❌ No |
Both hardcoded deny patterns (Phase 2, which operates on extracted
sub-commands) and user-configured `always_deny` patterns were affected.
## Fix
Replace the `simple_command.to_string()` call in
`extract_commands_from_simple_command` with manual construction that:
1. Iterates the command name (`word_or_name`) and word arguments from
the suffix
2. Normalizes each word by parsing it into `WordPiece`s via
`brush_parser::word::parse`
3. Reconstructs the semantic (unquoted) value from each piece:
- `Text` — used as-is
- `SingleQuotedText` — used as-is (brush-parser already strips the
wrapping quotes)
- `EscapeSequence` — leading `\` stripped to produce the unescaped
character
- `AnsiCQuotedText` — used as-is (brush-parser already evaluates
`$'...'`)
- `DoubleQuotedSequence` / `GettextDoubleQuotedSequence` — inner pieces
recursively normalized
- `TildePrefix` — preserved as `~` + suffix
- `ParameterExpansion` / `CommandSubstitution` /
`BackquotedCommandSubstitution` / `ArithmeticExpression` — original
source text preserved so patterns like `\$HOME` and `\${HOME}` continue
to match
4. Joins normalized words with spaces
Falls back to the raw `word.value` if word parsing fails.
## Cross-shell safety
The normalization operates on POSIX-parsed ASTs (brush-parser uses POSIX
grammar), so it applies POSIX quoting rules. For non-POSIX shells
(Nushell, PowerShell, Xonsh, Elvish, etc.) where quoting semantics
differ, any mismatch results in false positives (overly cautious
denials), never false negatives. This is the correct error direction for
a security feature. When brush-parser fails to parse a non-POSIX
command, `extract_commands` returns `None` and the existing fallback
(raw-string matching with `allow_enabled=false`) handles it.
(No release notes because granular tool permissions are still
feautre-flagged.)
Release Notes:
- N/A
`find_target()` failed to match numbers followed by a dot and a
non-digit (e.g. `1. item`), because the dot unconditionally reset
the scan state, discarding the number. Additionally, numbers at the
start of non-first lines were missed because the backward scan
stopped on the preceding newline and the forward scan immediately
broke on it.
Closes#47761
Release Notes:
- Fixed vim increment (`ctrl-a`) and decrement (`ctrl-x`) not working on Markdown ordered list markers like `1.`, `2.`, etc.
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: dino <dinojoaocosta@gmail.com>
- Update AI docs with missing features we shipped in December 2025
- Apply new brand voice to new docs
Release Notes:
- N/A *or* Added/Fixed/Improved ...
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
These changes update subword motions in order to also take `$` and `=`
into consideration as stopping punctuation, improving the subword
motions for certain languages where `$` and `=` are commonly used, like
PHP, for variables and assignments.
Closes#48267
Release Notes:
- Improved Vim's subword motions to stop at `$` and `=` characters
Before this change any agent diff caused us to recompute the entire
buffer's wraps, because we didn't handle the base text of diffs
changing.
Now, thanks to anchors, we can do that.
Release Notes:
- Fixed the editor rewrapping like crazy when agent edits were present
Adds `checkerboard` to `Background`, and use it for the side-by-side
diff.
Note that, since the blockmap can contain multiple `Spacer`s without any
gaps
in between, we ensure that the checkerboard pattern always shows an even
integer number of squares per line, so there are no obvious
discontinuities
in the pattern when this happens.
<img width="590" height="300" alt="image"
src="https://github.com/user-attachments/assets/e8f75f90-b230-4078-bce0-cb3c15613fe7"
/>
Release Notes:
- N/A *or* Added/Fixed/Improved ...
---------
Co-authored-by: Cole Miller <cole@zed.dev>
This extends the LSP settings schema system to also provide autocomplete
for the `settings` field (used for `workspace/configuration` responses),
in addition to the existing `initialization_options` support (#).
Changes:
- Add `settings_schema` method to `LspAdapter` trait and
`CachedLspAdapter`
- Update schema URL paths to be more explicit:
- `lsp/{adapter}/initialization_options` for init options schema
- `lsp/{adapter}/settings` for settings schema
- Add schema resolution logic for the new settings path
- Update tests to verify both schema references
Release Notes:
- Added autocomplete support for the `settings` field in LSP
configuration, complementing the existing `initialization_options`
autocomplete.
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Finn found out the hard way that the new cool tech does not quite work
at times. Back to the drawing board we go (at least partially).
Closes #ISSUE
Release Notes:
- N/A
## Summary
Reviewed and updated all 18 files in `docs/src/ai/` to meet Zed's brand
voice standards. Each file was scored against the 8-criterion rubric and
rewritten where needed to achieve 4+ on all criteria.
**Changes include:**
- Remove "Learn how to/about" patterns (8 instances)
- Replace vague phrases ("various types of tasks", "and more") with
specific examples
- Remove taboo words ("seamless", "naturally extends")
- Remove emotional language ("We're excited/thrilled/happy")
- Remove marketing patterns ("Stay tuned!", exclamation points)
- Remove unverifiable claims ("world's fastest")
- Replace hedging with direct statements
- Remove promotional emphasis (`_free_`, `**_for free_**`)
- Fix typo in external-agents.md ("servers" → "serves")
- Fix broken link syntax for Ollama Turbo
## Test plan
- [x] `mdbook build` succeeds
- [x] Grep for remaining taboo phrases (only one "seamless" in MCP
official quote, preserved intentionally)
- [x] All facts preserved (URLs, keybindings, settings keys, prices,
model names)
Fixes AI-8
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Some language servers include local symbols (e.g., local variables,
parameters) in workspace symbol results. Without the `containerName`
information, these symbols lack context information, making it difficult
to distinguish them from top-level definitions and hindering efficient
symbol lookup.
This change exposes the `container_name` field from LSP
[`SymbolInformation`](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#symbolInformation)
to the extension API, allowing language server extensions to access
`symbol.container_name` in `label_for_symbol` and provide meaningful
context when rendering symbol labels.
Note: The `container_name `field is added to all extension API versions
because they seem to share the same underlying Rust types via wasmtime
bindgen. The field is optional, so existing extensions would remain
compatible as far as I understand.
Closes #ISSUE
Release Notes:
- Added `container_name` field to `lsp::Symbol`, accessible via the
extension API's `label_for_symbol` function
---------
Co-authored-by: MrSubidubi <finn@zed.dev>
### Summary
Adds accept/reject tracking for Mercury edit predictions.
### Changes
Sends events to https://api-feedback.inceptionlabs.ai/feedback when:
Accept — user presses Tab
Reject — user presses Escape
Ignore — prediction dismissed implicitly (typing, cursor move, etc.)
Added `discard_explicit` method to the delegate trait to distinguish
explicit vs implicit dismissal. Updated `reject_prediction` and
`reject_current_prediction` methods with an `explicit` bool parameter to
thread this through to the Mercury feedback logic. Other providers are
unaffected—they use the default implementation.
Feedback is fire-and-forget in a background thread, only sent for
predictions that were shown.
### Data Collected
- Request ID (returned from Inception API)
- User action (either accept/reject/ignore)
- Client Zed version (to track updates made to Zed client which could
potentially affect nextedit implementation)
Release Notes:
- N/A
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
Expands the hardcoded security rules that block dangerous `rm` commands
in the agent terminal tool.
### New blocked patterns
- `rm -rf $HOME` / `rm -rf $HOME/` / `rm -rf ${HOME}` / `rm -rf
${HOME}/`
- `rm -rf .` / `rm -rf ./`
- `rm -rf ..` / `rm -rf ../`
- `rm -rf ~/` (previously only `rm -rf ~` was blocked)
### Flag handling improvements
- Simplified the flag character class from `[rRfF]` to `[rf]` since the
regex is already compiled with case-insensitive mode — less confusing,
same behavior.
- Added tests verifying that reversed flags (`-fr`), uppercase (`RM
-RF`), split flags (`-r -f`), and chained commands all get caught.
### Safe commands still allowed
Paths like `rm -rf ./build`, `rm -rf ~/Documents`, `rm -rf
$HOME/Documents`, `rm -rf ../some_dir`, and `rm -rf .hidden_dir` are
**not** blocked.
Release Notes:
- Auto-block a wider range of agent terminal commands, e.g. `rm -rf
$HOME` in addition to `rm -rf ~`
`brush-parser` handles `;` (sequential execution) and `|` (piping) which
all these shells use, so we can safely parse their commands for
`always_allow` pattern matching.
(No release notes because granular tool permissions haven't been
released yet.)
Release Notes:
- N/A
This regressed in #47619 -- previously we would throw away and reopen
the diffs used to construct the branch diff pretty eagerly, and as a
result the branch diff would update itself when you committed or
switched branches. We inadvertently lost that when we switched to
retaining those diffs in the git store.
Release Notes:
- N/A
Closes#46947
When installed with `npm install -g`, devcontainer gets created as
`devcontainer.cmd`. However, when invoked as a script from node, it's
still `devcontainer.js`. This addresses that difference so that we don't
accidentally invoke `devcontainer.cmd.js`.
Release Notes:
- Fixed devcontainer operation on Windows for when the devcontainer cli
is not installed globally
I primarily contribute to Doom source ports and related utilities. When
working with Doom WADs, the primary archive format for the game, Zed
will sometimes incorrectly assume that such files are plain text, due to
the nature of said container format not being compressed, some common
text files included in them will trigger a false positive in Zed's text
detection mode.
Considering some of these files can go from a couple dozen to a few
hundred megabytes, this usually makes Zed hang on my lower end machine,
until I manually terminate it :/ and has happened when I accidentally
clicked on one via the file explorer, or when viewing Git diffs.
Release Notes:
- Fixed Doom WAD files being erroneously treated as text
Part of #7450
Big thanks to @macmv for pushing this forwards so much!
Rebased version of https://github.com/zed-industries/zed/pull/39539 as
working on an in-org branch simplifies a lot of things for us)
Release Notes:
- Added LSP semantic tokens highlighting support
---------
Co-authored-by: Neil Macneale V <neil.macneale.v@gmail.com>
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
Co-authored-by: Zed Zippy <234243425+zed-zippy[bot]@users.noreply.github.com>
Currently, when overriding e.g a `binary` in dap settings, one have to
override all settings, including `args` as it otherwise set it to empty
string.
E.g this config will set `args` and `env` to their empty states, like
empty string and empty map:
```json
{
"dap": {
"Delve": {
"binary": "/path/to/binary"
}
}
}
```
This PR changes so that it is instead possible to only override a single
field. This is useful e.g on nixos where you don't want (or cant let)
zed download binaries for you, but you want to use the default args,
such as in Delve, you want zed to configure it with a random listening
port.
Release Notes:
- Improved dap settings so that it is possible to override individual
fields, such as `binary`
Otherwise block_with_timeout is effectively useless on windows as we
would block a minimum of 15ms which is our entire frame budget
Release Notes:
- N/A *or* Added/Fixed/Improved ...
There was a `when()` condition here looking at
`hovered_edited_file_buttons`, which is never turned to true when
hovering over the edited file button, and given we started to pass the
hover styles, tooltip, and on_click handler inside the condition, they
would never work. The solution here is to make these three things
unconditional, so edited files are always clickable.
Release Notes:
- N/A
Lukas Wirth