This PR decouples the scripting tool from the `Tool` trait while still
allowing it to be used as a tool from the model's perspective.
This will allow us to evolve the scripting tool as more of a first-class
citizen while still retaining the ability to have the model call it as a
regular tool.
Release Notes:
- N/A
We decided to expose scripting as tools again. We are aware of the UX
downsides of doing so, but we want to focus on getting it working well
first, and the model seems to make better use of it as an actual tool.
In the future, the tools API might support streaming. If it doesn't and
we need to ship, we can consider reverting this.
Release Notes:
- N/A
- Truncate branch names based on the width of the picker
- Use a footer for "Create branch" instead of a picker entry
Still to do:
- [x] Select the footer button when no matches and run the create logic
on `enter`
- [x] Make it possible to quickly select the footer button from the
keyboard when there are matches
Release Notes:
- Git Beta: Removed limitation that made it impossible to create a
branch from the branch picker when it too closely resembled an existing
branch name
Per a conversation with @nathansobo, have the Lua scripts run
unsandboxed for now (while this feature is behind the staff feature
flag).
Release Notes:
- N/A
Closes#20658
Now, when the "Go to Line" palette is open:
- Clicking on the editor will dismiss the palette without changing the
scroll position. (PR change)
- Pressing Enter will jump to the line number entered in the palette.
(Unchanged)
- Pressing Escape will jump back to the previous cursor location.
(Unchanged)
Release Notes:
- Fixed an issue where clicking the editor with the mouse while the "Go
to Line" palette is open would cause it to jump to the previous scroll
position.
Was dabbling on the shaders these past few days and felt like we could
have the WGSL logo. This is based on the logo found on the GPU Web
repository: https://github.com/gpuweb/gpuweb/tree/main/logo
Release Notes:
- N/A
Closes#25421
This PR adds support for external file managers to show Zed as an option
in the "Open With" context menu for directories on macOS.
<img width="350" alt="image"
src="https://github.com/user-attachments/assets/c52acd48-73c4-47be-8683-6950e0371b73"
/>
Release Notes:
- Added support for opening folders in Zed from third-party macOS file
managers like Path Finder and Super Charge through their "Open With"
menu.
This PR includes lots of small fixes to get our `build.nix` and
`shell.nix` back to a working state.
I've tested this by running `cargo run` (inside the devshell) and `nix
run` on x86 nixos and arm64 darwin machines. I'd appreciate it if others
could test building inside the devshell to double-check that it's not
just working because I happen to have some system-level packages
installed, as well as seeing if it works on other platforms (non-nixos
linux, arm linux, x86 darwin).
I couldn't get the full test suite (`cargo nextest run --workspace`)
passing in the devshell on darwin, but they _are_ all passing on nixos.
nixpkgs [disables some of our
tests](92d11f06d5/pkgs/by-name/ze/zed-editor/package.nix (L226-L234))
that apparently fail or are flakey on hydra, but they don't know why.
I'm going to punt on debugging those for now, especially given that they
seem to be working for me. I'm also unsure of whether we actually want
the nix checkPhase to run the full test suite (it's currently not
passing `--workspace`) given that we have separate CI that should
enforce that those pass on all PRs.
Here's an overview of the changes made:
- Fix our `generate-licenses` script
- Relaxes the `cargo-about` version requirement slightly so it doesn't
try to install an older binary when the nixpkgs one is newer than our
requirement
- Add a workaround for [this cargo-about
issue](https://github.com/zed-industries/zed/issues/19971) obviating the
need for the patching done in the nixpkgs package
- Set the new `--frozen` flag to avoid network access/mutating the
lockfile
- Use dynamic webrtc lib from nixpkgs, and fixes up the build script in
webrtc-sys that hardcodes it to be statically linked.
- Use `inputsFrom` in `shell.nix` and avoid duplicating everything from
`build.nix`
- Add a temporary workaround for an [upstream crane
bug](https://github.com/ipetkov/crane/issues/808).
- Fix shebangs in our `script` dir to not hard-code `/bin/bash`
There are still a bunch of issues that aren't resolved here, I'll make a
tracking issue for those and try to land this first just to get back to
an unbroken state. Eventually among other things I'd like to use a
`libgit2` from `staticPkgs` and musl cross compilation to build the
remote server under nix, and then add that as a separate flake output
and include it in the shell's `inputsFrom` list.
Thanks @niklaskorz, @GaetanLepage, @bbigras and all the other nixpkgs
maintainers that have kept the `zed-editor` package working and up to
date! I seriously considered just making our flake `overrideAttrs` the
package in nixpkgs given how well maintained it is.
Thanks @WeetHet for your volunteer maintinance of this flake. I
referenced #24953 while working on these fixes, and I'd love to
collaborate on adding some of those pieces like treefmt and a github
action. If you're interested I'd really appreciate some help debugging
why crane's `buildDepsOnly` isn't working for us. I'm assuming it'd make
our `nix build` times go way down from the improved dep caching if we
could get it working.
Thanks @rrbutani for all the help on this PR 💙.
Release Notes:
- N/A
---------
Co-authored-by: Rahul Butani <rrbutani@users.noreply.github.com>
Co-authored-by: Rahul Butani <rr.butani@gmail.com>
Let's play around with it. This should not be added to tomorrow's
preview.
Release Notes:
- Git Beta: Added a panel header with an open diff and stage/unstage all
buttons.
Closes https://github.com/zed-industries/zed/issues/21277
To the left is current Zed, right is the improved version.
3rd message, from Zed, to resolve the item, does not have `textEdit` on
the right side, and has one on the left.
Seems to not influence the end result though, but at least Zed behaves
more appropriate now.
<img width="1727" alt="image"
src="https://github.com/user-attachments/assets/ca1236fd-9ce2-41ba-88fe-1f3178cdcbde"
/>
Instead of modifying the original LSP completion item, store completion
list defaults and apply them when the item is requested (except `data`
defaults, needed for resolve).
Now, the only place that can modify the completion items is this method,
and Python impl seems to be the one doing it:
ca9c3af56f/crates/languages/src/python.rs (L182-L204)
Seems ok to leave untouched for now.
Release Notes:
- Fixed LSP completion items modified before resolve request
Reverts zed-industries/zed#25135
This approach was not the best as explained in the response to the
original PR. Likely, the better approach is to create a newer specific
scope for these kinds of variables under the `@variable` prefix so that
themes can control these pseudo-keywords specifically
This PR makes refactors the scripting functionality to be a first-class
concept of the assistant instead of a generic tool, which will allow us
to build a more customized experience.
- The tool prompt has been slightly tweaked and is now included as a
system message in all conversations. I'm getting decent results, but now
that it isn't in the tools framework, it will probably require more
refining.
- The model will now include an `<eval ...>` tag at the end of the
message with the script. We parse this tag incrementally as it streams
in so that we can indicate that we are generating a script before we see
the closing `</eval>` tag. Later, this will help us interpret the script
as it arrives also.
- Threads now hold a `ScriptSession` entity which manages the state of
all scripts (from parsing to exited) in a centralized way, and will
later collect all script operations so they can be displayed in the UI.
- `script_tool` has been renamed to `assistant_scripting`
- Script source now opens in a regular read-only buffer
Note: We still need to handle persistence properly
Release Notes:
- N/A
---------
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
Closes#25509
Changes:
- If active item is already diagnostics, don't try to focus it again.
Instead of not focusing, should it just not activate instead? Something
like:
if !workspace
.active_item(cx)
.map(|item| item.item_id() == existing.item_id())
.unwrap_or(false)
{
workspace.activate_item(&existing, true, true, window, cx);
}
Release Notes:
- N/A
Closes#25594
This PR fixes an issue where signing into Copilot required restarting
Zed.
Copilot depends on an OAuth token that comes from either `hosts.json` or
`apps.json`. Initially, both files don't exist. If neither file is
found, we fallback to watching `hosts.json` for updates. However, if the
auth process creates `apps.json`, we won't receive updates from it,
causing the UI to remain outdated.
This PR fixes that by watching the parent `github-copilot` directory
instead, which will always contain one of those files along with an
additional version file.
I have tested this on macOS and Linux Wayland.
Release Notes:
- Fixed an issue where signing into Copilot required restarting Zed.
This is motivated by trying to make the Project Diff view usable with
huge Git change sets.
Release Notes:
- Improved performance of rendering multibuffers with very large numbers
of buffers
Release Notes:
- Git Beta: added `git:Add` as an alias for the existing `git::Diff`
- Git Beta: Fixed a bug where the 'generate commit message' keybinding
wasn't working.
- Git Beta: Made the empty project diff state a little more helpful with
a button to push, and a button to close the item.
Closes#26002
Release Notes:
- Added `multibuffer` key context.
- `cmd-down` and `cmd-shift-down` on Mac now moves to the end of the
last line of a singleton buffer instead of the beginning. In
multibuffers, these now move to the start of the next excerpt.
- Fixed `vim::PreviousSectionEnd` (bound to `[ ]`) to move to the
beginning of the line, matching the behavior of `vim::NextSectionEnd`.
- Added `editor::MoveToStartOfNextExcerpt` and
`editor::MoveToEndOfPreviousExcerpt`.
This gives us _very_ rudimentary support for `git switch` and `git
checkout` now, by making them aliases for our existing `git::branch`
call.
Release Notes:
- Git Beta: Added `git::Switch` and `git::CheckoutBranch` as aliases for
the existing `git::Branch`
This PR factors out a new `Thread::all_tools_finished` method to
encapsulate some of the boilerplate in the `ThreadEvent::ToolFinished`
event handler.
This should make this event handler easier to replicate for the eval
use-case.
Release Notes:
- N/A
This PR adds two new helper methods to the `Thread` for dealing with
tool use:
- `use_pending_tools` - This uses all of the tools that are pending
- The reason we aren't calling this directly in `stream_completion` is
that we still might need to have a way for users to confirm that they
want tools to be run, which would need to happen at the UI layer in the
`ActiveThread`.
- `send_tool_results_to_model` - This encapsulates inserting a new user
message that contains the tool results and sending them up to the model.
Release Notes:
- N/A
This PR decouples the `Tool` trait from the `Workspace` (and from the
UI, in general).
`Tool::run` now takes a `WeakEntity<Project>` instead of a
`WeakEntity<Workspace>` and a `Window`.
Release Notes:
- N/A
A preparation for https://github.com/zed-industries/zed/issues/4957 that
pushes all LSP-related data out from the basic completion item, so that
it's possible to create completion items without any trace of LSP
clearly.
Release Notes:
- N/A
This PR updates the `ExtensionBuilder` to require that grammar names are
written in snake_case.
The grammar names are used to construct identifiers, so we need them to
be valid C identifiers.
Release Notes:
- N/A
This PR adds pages to component preview when clicking on a given
component in the sidebar.
This will let us create richer previews & better docs for using
components in the future.
Release Notes:
- N/A
This reduces the number of multibuffer syncs from 100,000 to 20,000.
Before this change each editor individually observed the project, so
literally any project change was amplified by the number of editors you
had open.
Now editors listen to their buffers instead of the project, and other
users of `cx.observe` on the project have been updated to use specific
events to reduce churn.
Follow up to #26237
Release Notes:
- Improved performance of Zed in large repos with lots of file system
events.
---------
Co-authored-by: Max Brunsfeld <maxbrunsfeld@gmail.com>
We're still recreating a session for every tool call, but the idea is to
have a long-lived `Session` per assistant thread.
Release Notes:
- N/A
---------
Co-authored-by: Agus Zubiaga <hi@aguz.me>
## Problem
Running `cargo run .` twice in Zed repository required a rebuild two
times in a row. The second rebuild was triggered around libz-sys, which
in practice caused a rebuild of the ~entire project.
Some concrete examples:
```
cargo test -p project # Requires a rebuild (warranted)
cargo run .
cargo test -p project # Requires a rebuild (unwarranted)
```
or
```
cargo run . # Requires a rebuild (warranted)
cargo run . # Requires a rebuild (unwarranted)
```
## What's going on
Zed build script on MacOS sets MACOSX_DEPLOYMENT_TARGET to 10.15. This
is fine. However, **cargo propagates all environment variables to child
processes during `cargo run`**. This then affects Rust Analyzer spawned
by dev Zed - it clobbers build cache of whatever package it touches,
because it's behavior is not same between running it with `cargo run`
(where MACOS_DEPLOYMENT_TARGET gets propagated to child Zed) and running
it directly via `target/debug/zed` or whatever (where the env variable
is not set, so that build behaves roughly like Zed Dev.app).
## Solution
~We'll unset that env variable from user environment when we're
reasonably confident that we're running under `cargo run` by exploiting
other env variables set by cargo:
https://doc.rust-lang.org/cargo/reference/environment-variables.html
CARGO_PKG_NAME is always set to `zed` when running it via `cargo run`,
as it's the value propagated from the build.~
~The alternative I've considered is running [via a custom
runner](https://doc.rust-lang.org/cargo/reference/config.html#targetcfgrunner),
though the problem here is that we'd have to use a shell script to unset
the env variable - that could be problematic with e.g. fish. I just
didn't want to deal with that, though admittedly it would've been
cleaner in other aspects.~
Redact all above. We'll just set MACOSX_DEPLOYMENT_TARGET regardless of
whether you have it in your OG shell environment or not.
Release Notes:
- N/A
205f9a9f03/crates/editor/src/element.rs (L1643)
Due to the snippet above, Zed is supposed to have `row` larger or equal
to `start_row` here:
205f9a9f03/crates/editor/src/element.rs (L1694)
yet the panic were reported when clicking in the project diff.
That project diff has a lot of highlighting happening already, so the PR
disables inline diagnostics within a git diff view.
Release Notes:
- N/A
Access to `cx` will be needed for anything that queries entities. In
this commit this is use of `WorktreeStore::find_search_candidates`. In
the future it will be things like access to LSP / tree-sitter outlines /
etc.
Changes to support access to `cx` from functions provided to the Lua
script:
* Adds a channel of requests that require a `cx`. Work enqueued to this
channel is run on the foreground thread.
* Adds `async` and `send` features to `mlua` crate so that async rust
functions can be used from Lua.
* Changes uses of `Rc<RefCell<...>>` to `Arc<Mutex<...>>` so that the
futures are `Send`.
One benefit of reusing project search logic for search candidates is
that it properly ignores paths.
Release Notes:
- N/A
