This PR unships tool use from Assistant1.
This was only ever partially implemented, and was never released to end
users.
Assistant2 will support tool use.
Release Notes:
- N/A
Fix a bug where a GPUI macro still used `ModelContext`
Rename `AsyncAppContext` -> `AsyncApp`
Rename update_model, read_model, insert_model, and reserve_model to
update_entity, read_entity, insert_entity, and reserve_entity
Release Notes:
- N/A
There's still a bit more work to do on this, but this PR is compiling
(with warnings) after eliminating the key types. When the tasks below
are complete, this will be the new narrative for GPUI:
- `Entity<T>` - This replaces `View<T>`/`Model<T>`. It represents a unit
of state, and if `T` implements `Render`, then `Entity<T>` implements
`Element`.
- `&mut App` This replaces `AppContext` and represents the app.
- `&mut Context<T>` This replaces `ModelContext` and derefs to `App`. It
is provided by the framework when updating an entity.
- `&mut Window` Broken out of `&mut WindowContext` which no longer
exists. Every method that once took `&mut WindowContext` now takes `&mut
Window, &mut App` and every method that took `&mut ViewContext<T>` now
takes `&mut Window, &mut Context<T>`
Not pictured here are the two other failed attempts. It's been quite a
month!
Tasks:
- [x] Remove `View`, `ViewContext`, `WindowContext` and thread through
`Window`
- [x] [@cole-miller @mikayla-maki] Redraw window when entities change
- [x] [@cole-miller @mikayla-maki] Get examples and Zed running
- [x] [@cole-miller @mikayla-maki] Fix Zed rendering
- [x] [@mikayla-maki] Fix todo! macros and comments
- [x] Fix a bug where the editor would not be redrawn because of view
caching
- [x] remove publicness window.notify() and replace with
`AppContext::notify`
- [x] remove `observe_new_window_models`, replace with
`observe_new_models` with an optional window
- [x] Fix a bug where the project panel would not be redrawn because of
the wrong refresh() call being used
- [x] Fix the tests
- [x] Fix warnings by eliminating `Window` params or using `_`
- [x] Fix conflicts
- [x] Simplify generic code where possible
- [x] Rename types
- [ ] Update docs
### issues post merge
- [x] Issues switching between normal and insert mode
- [x] Assistant re-rendering failure
- [x] Vim test failures
- [x] Mac build issue
Release Notes:
- N/A
---------
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Cole Miller <cole@zed.dev>
Co-authored-by: Mikayla <mikayla@zed.dev>
Co-authored-by: Joseph <joseph@zed.dev>
Co-authored-by: max <max@zed.dev>
Co-authored-by: Michael Sloan <michael@zed.dev>
Co-authored-by: Mikayla Maki <mikaylamaki@Mikaylas-MacBook-Pro.local>
Co-authored-by: Mikayla <mikayla.c.maki@gmail.com>
Co-authored-by: joão <joao@zed.dev>
Closes#12553
* [x] Fix `diff_hunk_before`
* [x] Fix failure to show deleted text when expanding hunk w/ cursor on
second line of the hunk
* [x] Failure to expand diff hunk below the cursor.
* [x] Delete the whole file, and expand the diff. Backspace over the
deleted hunk, panic!
* [x] Go-to-line now counts the diff hunks, but it should not
* [x] backspace at the beginning of a deleted hunk deletes too much text
* [x] Indent guides are rendered incorrectly
* [ ] Fix randomized multi buffer tests
Maybe:
* [ ] Buffer search should include deleted text (in vim mode it turns
out I use `/x` all the time to jump to the next x I can see).
* [ ] vim: should refuse to switch into insert mode if selection is
fully within a diff.
* [ ] vim `o` command when cursor is on last line of deleted hunk.
* [ ] vim `shift-o` on first line of deleted hunk moves cursor but
doesn't insert line
* [x] `enter` at end of diff hunk inserts a new line but doesn't move
cursor
* [x] (`shift-enter` at start of diff hunk does nothing)
* [ ] Inserting a line just before an expanded hunk collapses it
Release Notes:
- Improved diff rendering, allowing you to navigate with your cursor
inside of deleted text in diff hunks.
---------
Co-authored-by: Conrad <conrad@zed.dev>
Co-authored-by: Cole <cole@zed.dev>
Co-authored-by: Mikayla <mikayla@zed.dev>
Co-authored-by: Conrad Irwin <conrad.irwin@gmail.com>
Co-authored-by: Michael <michael@zed.dev>
Co-authored-by: Agus <agus@zed.dev>
Co-authored-by: João <joao@zed.dev>
This PR adds a new `assistant_context_editor` crate.
This will ultimately house the `ContextEditor` so that it can be
consumed by both `assistant` and `assistant2`.
For the purposes of this PR, we just introduce the crate and move some
supporting constructs to it, such as the `ContextStore`.
Release Notes:
- N/A
* Collects and reports all parse errors
* Shares parsed `KeyBindingContextPredicate` among the actions.
* Updates gpui keybinding and action parsing to return structured
errors.
* Renames "block" to "section" to match the docs, as types like
`KeymapSection` are shown in `json-language-server` hovers.
* Removes wrapping of `context` and `use_key_equivalents` fields so that
`json-language-server` auto-inserts `""` and `false` instead of `null`.
* Updates `add_to_cx` to take `&self`, so that the user keymap doesn't
get unnecessarily cloned.
In retrospect I wish I'd just switched to using TreeSitter to do the
parsing and provide proper diagnostics. This is tracked in #23333
Release Notes:
- Improved handling of errors within the user keymap file. Parse errors
within context, keystrokes, or actions no longer prevent loading the key
bindings that do parse.
https://github.com/user-attachments/assets/78db908e-cfe5-4803-b0dc-4f33bc457840
* starts to extract usernames out of `users/` GitHub API responses, and
pass those along with e-mails in the collab sessions as part of the
`User` data
* adjusts various prefill and seed test methods so that the new data can
be retrieved from GitHub properly
* if there's an active call, where guests have write permissions and
e-mails, allow to trigger `FillCoAuthors` action in the context of the
git panel, that will fill in `co-authored-by:` lines, using e-mail and
names (or GitHub handle names if name is absent)
* the action tries to not duplicate such entries, if any are present
already, and adds those below the rest of the commit input's text
Concerns:
* users with write permissions and no e-mails will be silently omitted
— adding odd entries that try to indicate this or raising pop-ups is
very intrusive (maybe, we can add `#`-prefixed comments?), logging seems
pointless
* it's not clear whether the data prefill will run properly on the
existing users — seems tolerable now, as it seems that we get e-mails
properly already, so we'll see GitHub handles instead of names in the
worst case. This can be prefilled better later.
* e-mails and names for a particular project may be not what the user
wants.
E.g. my `.gitconfig` has
```
[user]
email = mail4score@gmail.com
# .....snip
[includeif "gitdir:**/work/zed/**/.git"]
path = ~/.gitconfig.work
```
and that one has
```
[user]
email = kirill@zed.dev
```
while my GitHub profile is configured so, that `mail4score@gmail.com` is
the public, commit e-mail.
So, when I'm a participant in a Zed session, wrong e-mail will be
picked.
The problem is, it's impossible for a host to get remote's collaborator
git metadata for a particular project, as that might not even exist on
disk for the client.
Seems that we might want to add some "project git URL <-> user name and
email" mapping in the settings(?).
The design of this is not very clear, so the PR concentrates on the
basics for now.
When https://github.com/zed-industries/zed/pull/23308 lands, most of the
issues can be solved by collaborators manually, before committing.
Release Notes:
- N/A
This reverts commit 1b3b825c7f.
When debugging git diffs we found that this introduced a re-ordering of
messages sent to the LSP:
* User hits "format"
* Zed adjusts spacing, and sends "spaces changed" to the LSP
* Zed sends "format" to LSP
With the async approach here, the format request can now arrive before
the space changed request.
You can reproduce this with `test_strip_whitespace_and_format_via_lsp`
under some conditions.
Release Notes:
- N/A
First, parse the output of `git status --porcelain=v1` into a
representation that can handle the full "grammar" and doesn't lose
information.
Second, as part of pushing this throughout the codebase, expand the use
of the existing `GitSummary` type to all the places where status
propagation is in play (i.e., anywhere we're dealing with a mix of files
and directories), and get rid of the previous `GitSummary ->
GitFileStatus` conversion.
- [x] Synchronize new representation over collab
- [x] Update zed.proto
- [x] Update DB models
- [x] Update `GitSummary` and summarization for the new `FileStatus`
- [x] Fix all tests
- [x] worktree
- [x] collab
- [x] Clean up `FILE_*` constants
- [x] New collab tests to exercise syncing of complex statuses
- [x] Run it locally and make sure it looks good
Release Notes:
- N/A
---------
Co-authored-by: Mikayla <mikayla@zed.dev>
Co-authored-by: Conrad <conrad@zed.dev>
Closes https://github.com/zed-industries/zed/issues/7928
* uncomments and fixes all inlay hint cache tests
* fixes a bug, where invalidated range did not store the new queried
ranges in the cache: this resulted in extra requests in editor that do
not fit into the screen
* comments a peculiarity with the `RefreshInlayHints` event: all editors
react to that when a new language server is inserted, even though
certain editors are not related to the new language server
* fixes handling of inlay hints for the same position: now the same
order is kept, as in the language server's response
(https://github.com/zed-industries/zed/issues/7928)
* queries for hints when on excerpt(s) expansion
Release Notes:
- Fixed inlay hints handling for the same position
This is a follow-up to #12640.
While profiling latency of working with a project with 8192 diagnostics
I've noticed that while we're parsing the LSP messages into a generic
message struct on a background thread, we can still block the main
thread as the conversion between that generic message struct and the
actual LSP message (for use by callback) is still happening on the main
thread.
This PR significantly constrains what a message callback can use, so
that it can be executed on any thread; we also send off message
conversion to the background thread. In practice new callback
constraints were already satisfied by all call sites, so no code outside
of the lsp crate had to be adjusted.
This has improved throughput of my 8192-benchmark from 40s to send out
all diagnostics after saving to ~20s. Now main thread is spending most
of the time updating our diagnostics sets, which can probably be
improved too.
Closes #ISSUE
Release Notes:
- Improved app responsiveness with huge # of diagnostics.
In current code this doesn't have benefit. In preparation for avoiding a
clone of workspace configuration. Having the interface this way may make
opportunities for efficiency clearer in the future
Release Notes:
- N/A
TODO:
- [ ] Investigate incorrect hit target for `stage all` button
- [ ] Add top level context menu
- [ ] Add entry context menus
- [x] Show paths in list view
- [ ] For now, `enter` can just open the file
- [ ] 🐞: Hover deadzone in list caused by scrollbar
- [x] 🐞: Incorrect status/nothing shown when multiple worktrees are
added
---
This PR continues work on the feature flagged git panel.
Changes:
- Defines and wires up git panel actions & keybindings
- Re-scopes some actions from `git_ui` -> `git`.
- General git actions (StageAll, CommitChanges, ...) are scoped to
`git`.
- Git panel specific actions (Close, FocusCommitEditor, ...) are scoped
to `git_panel.
- Staging actions & UI are now connected to git!
- Unify more reusable git status into the GitState global over being
tied to the panel directly.
- Uses the new git status codepaths instead of filtering all workspace
entries
Release Notes:
- N/A
---------
Co-authored-by: Cole Miller <53574922+cole-miller@users.noreply.github.com>
Co-authored-by: Cole Miller <cole@zed.dev>
Before this change it was using the same multibuffer point ranges in
every buffer, which only worked correctly for singleton buffers.
Release Notes:
- Fixed handling of selection ranges when formatting selections within a
multibuffer.
---------
Co-authored-by: Thorsten Ball <mrnugget@gmail.com>
This reverts commit 3ae6aa0e4d.
If "group_id = 0" really did mean a diagnostic not from a language
server then various methods related to diagnostic set would need to be
updated. Something like [this
diff](https://gist.github.com/mgsloan/e902153bcaec207b39260a8f40d3134d).
Plan instead is to use InfoPopover instead of DiagnosticPopover for
these.
Release Notes:
- N/A
In particular, `DiagnosticPopover` both:
* Supports moving the selection to a diagnostic when clicked, based on
`group_id`
* Provides Diagnostic values with `group_id: 0` providing informztion on
hover about invisible characters.
So, clicking such a popover would navigate to the very first error
produced by a language server. Really not a big deal of course, but
seems good to fix as it might result in surprising behavior in other
future circumstances
Release Notes:
- N/A
- [x] Rewrite worktree git handling
- [x] Fix tests
- [x] Fix `test_propagate_statuses_for_repos_under_project`
- [x] Replace `WorkDirectoryEntry` with `WorkDirectory` in
`RepositoryEntry`
- [x] Add a worktree event for capturing git status changes
- [x] Confirm that the local repositories are correctly updating the new
WorkDirectory field
- [x] Implement the git statuses query as a join when pulling entries
out of worktree
- [x] Use this new join to implement the project panel and outline
panel.
- [x] Synchronize git statuses over the wire for collab and remote dev
(use the existing `worktree_repository_statuses` table, adjust as
needed)
- [x] Only send changed statuses to collab
Release Notes:
- N/A
---------
Co-authored-by: Cole Miller <cole@zed.dev>
Co-authored-by: Mikayla <mikayla@zed.com>
Co-authored-by: Nathan <nathan@zed.dev>
Still TODO:
* [x] Factor out `start_language_server` so we can call it on register
(instead of on detect language)
* [x] Only call register in singleton editors (or when
editing/go-to-definition etc. in a multibuffer?)
* [x] Refcount on register so we can unregister when no buffer remain
* [ ] (maybe) Stop language servers that are no longer needed after some
time
Release Notes:
- Fixed language servers starting when doing project search
- Fixed high CPU usage when ignoring warnings in the diagnostics view
---------
Co-authored-by: Piotr Osiewicz <24362066+osiewicz@users.noreply.github.com>
Co-authored-by: Cole <cole@zed.dev>
Similar to #20826 but keeps the Swift implementation. There were quite a
few changes in the `call` crate, and so that code now has two variants.
Closes#13714
Release Notes:
- Added preliminary Linux support for voice chat and viewing
screenshares.
---------
Co-authored-by: Kirill Bulatov <mail4score@gmail.com>
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
Co-authored-by: Mikayla <mikayla@zed.dev>
This is a pure refactor of our Git diff state management. Buffers are no
longer are associated with one single diff (the unstaged changes).
Instead, there is an explicit project API for retrieving a buffer's
unstaged changes, and the `Editor` view layer is responsible for
choosing what diff to associate with a buffer.
The reason for this change is that we'll soon want to add multiple "git
diff views" to Zed, one of which will show the *uncommitted* changes for
a buffer. But that view will need to co-exist with other views of the
same buffer, which may want to show the unstaged changes.
### Todo
* [x] Get git gutter and git hunks working with new structure
* [x] Update editor tests to use new APIs
* [x] Update buffer tests
* [x] Restructure remoting/collab protocol
* [x] Update assertions about staged text in
`random_project_collaboration_tests`
* [x] Move buffer tests for git diff management to a new spot, using the
new APIs
Release Notes:
- N/A
---------
Co-authored-by: Richard <richard@zed.dev>
Co-authored-by: Cole <cole@zed.dev>
Co-authored-by: Conrad <conrad@zed.dev>
This PR factors the tool definitions out of the `assistant` crate so
that they can be shared between `assistant` and `assistant2`.
`ToolWorkingSet` now lives in `assistant_tool`. The tool definitions
themselves live in `assistant_tools`, with the exception of the
`ContextServerTool`, which has been moved to the `context_server` crate.
As part of this refactoring I needed to extract the
`ContextServerSettings` to a separate `context_server_settings` crate so
that the `extension_host`—which is referenced by the `remote_server`—can
name the `ContextServerSettings` type without pulling in some undesired
dependencies.
Release Notes:
- N/A
This PR overhauls extension registration in order to make it more
modular.
The `extension` crate now contains an `ExtensionHostProxy` that can be
used to register various proxies that the extension host can use to
interact with the rest of the system.
There are now a number of different proxy traits representing the
various pieces of functionality that can be provided by an extension.
The respective crates that provide this functionality can implement
their corresponding proxy trait in order to register a proxy that the
extension host will use to register the bits of functionality provided
by the extension.
Release Notes:
- N/A
I've observed that Zed's implementation of [Code Actions On
Format](https://zed.dev/docs/configuring-zed#code-actions-on-format)
uses the
[CodeActionContext.only](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#codeActionContext)
parameter to request specific code action kinds from the server. The
issue is that it does not filter out code actions from the response,
believing that the server will do it.
The [LSP
specification](https://microsoft.github.io/language-server-protocol/specifications/lsp/3.17/specification/#codeActionContext)
says that the client is responsible for filtering out unwanted code
actions:
```js
/**
* Requested kind of actions to return.
*
* Actions not of this kind are filtered out by the client before being
* shown. So servers can omit computing them.
*/
only?: CodeActionKind[];
```
This PR will filter out unwanted code action on the client side.
I have initially encountered this issue because the [ZLS language
server](https://github.com/zigtools/zls) (until
https://github.com/zigtools/zls/pull/2087) does not filter code action
based on `CodeActionContext.only` so Zed runs all received code actions
even if they are explicitly disabled in the `code_actions_on_format`
setting.
Release Notes:
- Fix the `code_actions_on_format` setting when used with a language
server like ZLS
---------
Co-authored-by: Piotr Osiewicz <24362066+osiewicz@users.noreply.github.com>
We occasionally see dates in the future appearing in our telemetry. One
hypothesis is that this is caused by a clock change while Zed is running
causing date math based on chrono to be incorrect.
Instant *should* be a more stable source of relative timestamps.
Release Notes:
- N/A
Motivation for this is to make things more understandable while figuring
out #20775.
This is intended to be a refactoring that does not affect behavior, but
there are a few tricky spots:
* Previously `File.mtime()` (now `File.disk_state().mtime()`) would
return last known modification time for deleted files. Looking at uses,
I believe this will not affect anything. If there are behavior changes
here I believe they would be improvements.
* `BufferEvent::DirtyChanged` is now only emitted if dirtiness actually
changed, rather than if it may have changed. This should only be an
efficiency improvement.
Release Notes:
- N/A
Co-authored-by: Mikayla Maki <mikayla@zed.dev>
Issues found:
* audio does not work well with various set-ups using USB
* switching audio during initial join may leave the client with no audio
at all
* audio streaming is done on the main thread, beachballing certain
set-ups
* worse screenshare quality (seems that there's no dynamic scaling
anymore, compared to the Swift SDK)
This reverts commit 1235d0808e.
Release Notes:
- N/A
See https://github.com/livekit/rust-sdks/pull/355
Todo:
* [x] make `call` / `live_kit_client` crates use the livekit rust sdk
* [x] create a fake version of livekit rust API for integration tests
* [x] capture local audio
* [x] play remote audio
* [x] capture local video tracks
* [x] play remote video tracks
* [x] tests passing
* bugs
* [x] deafening does not work
(https://github.com/livekit/rust-sdks/issues/359)
* [x] mute and speaking status are not replicated properly:
(https://github.com/livekit/rust-sdks/issues/358)
* [x] **linux** - crash due to symbol conflict between WebRTC's
BoringSSL and libcurl's openssl
(https://github.com/livekit/rust-sdks/issues/89)
* [x] **linux** - libwebrtc-sys adds undesired dependencies on `libGL`
and `libXext`
* [x] **windows** - linker error, maybe related to the C++ stdlib
(https://github.com/livekit/rust-sdks/issues/364)
```
libwebrtc_sys-54978c6ad5066a35.rlib(video_frame.obj) : error LNK2038:
mismatch detected for 'RuntimeLibrary': value 'MT_StaticRelease' doesn't
match value 'MD_DynamicRelease' in
libtree_sitter_yaml-df6b0adf8f009e8f.rlib(2e40c9e35e9506f4-scanner.o)
```
* [x] audio problems
Release Notes:
- Switch from Swift to Rust LiveKit SDK 🦀
---------
Co-authored-by: Mikayla Maki <mikayla@zed.dev>
Co-authored-by: Conrad Irwin <conrad@zed.dev>
Co-authored-by: Kirill Bulatov <kirill@zed.dev>
Co-authored-by: Michael Sloan <michael@zed.dev>
This optimizes and fixes bugs in our logic for maintaining a set of
running context servers, based on the combination of the user's
`context_servers` settings and their installed extensions.
Release Notes:
- N/A
---------
Co-authored-by: Marshall <marshall@zed.dev>
Co-authored-by: Marshall Bowers <elliott.codes@gmail.com>
This fixes the issue of multiple language servers showing up as `node`
in the language server logs dropdown.
It does this by changing `language_server.name()` to return the
adapter's name, not the binary name, and changing types to make sure
that we always use this.
Release Notes:
- Fixed language server names showing up only as `"node"`
---------
Co-authored-by: Sam Rose <hello@samwho.dev>
Co-authored-by: Bennet <bennet@zed.dev>
This PR reworks how the Assistant Panel references slash commands,
context servers, and tools.
Previously we were always reading them from the global registries, but
now we store individual collections on each Assistant Panel instance so
that there can be different ones registered for each project.
Release Notes:
- N/A
---------
Co-authored-by: Max <max@zed.dev>
Co-authored-by: Antonio <antonio@zed.dev>
Co-authored-by: Joseph <joseph@zed.dev>
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
This contains the main changes to the extensions crate from #20049. The
primary goal here is removing dependencies that we can't include on the
remote.
Release Notes:
- N/A
---------
Co-authored-by: Mikayla <mikayla@zed.dev>
Co-authored-by: Marshall Bowers <elliott.codes@gmail.com>
Co-authored-by: Marshall <marshall@zed.dev>
This PR is only updating UI strings and pieces of the documentation—it
doesn't touch the actual code, where it's still using things such as
`NewContext` and similar terminology for variables, actions, etc.
Release Notes:
- N/A
This is a behavior-preserving change, but lays the groundwork for
expanding selections when the cursor lands inside of a "replace" block.
Release Notes:
- N/A
Fixed the bug when shared ssh project did not account for client
changing things in their buffers.
Also ensures Prettier formatting workflow works for both ssh project
owner and ssh project clients.
Release Notes:
- N/A
---------
Co-authored-by: Conrad Irwin <conrad@zed.dev>
Closes#19758
Release Notes:
- Added feature to show commit summary as part of the inline Git blame
---------
Co-authored-by: Thorsten Ball <mrnugget@gmail.com>
Use `Fs` instead of `std::fs` and do entry existence checks better:
* first, check the worktree entry existence without any FS checks
* then, only for local cases, use `Fs` to check for abs_path existence
of items, in case those came from single-filed worktrees that got closed
and removed.
Remote entries do not get file existence checks, so might try opening
previously removed buffers for now.
Release Notes:
- N/A
Closes https://github.com/zed-industries/zed/issues/8534
Supersedes https://github.com/zed-industries/zed/pull/16349
Potential concerns:
* we do not follow up to the `/` when looking for `.editorconfig`, only
up to the worktree root.
Seems fine for most of the cases, and the rest should be solved
generically later, as the same issue exists for settings.json
* `fn language` in `AllLanguageSettings` is very hot, called very
frequently during rendering. We accumulate and parse all `.editorconfig`
file contents beforehand, but have to go over globs and match these
against the path given + merge the properties still.
This does not seem to be very bad, but needs more testing and
potentially some extra caching.
Release Notes:
- Added .editorconfig support
---------
Co-authored-by: Ulysse Buonomo <buonomo.ulysse@gmail.com>
Before this change messages could be lost on reconnect, now they will
not be.
Release Notes:
- SSH Remoting: make reconnects smoother
---------
Co-authored-by: Nathan <nathan@zed.dev>
