This PR adds the ability to hear a sound notification when the agent is
done generating and/or needs user input. This setting is turned off by
default and can be used together with the visual notification. The
specific sound I'm using here comes from the [Material Design 2 Sound
Library](https://m2.material.io/design/sound/sound-resources.html#).
Release Notes:
- agent: Added the ability to have a sound notification when the agent
is done generating and/or needs user input.
Editing JSON styles is not very helpful for bringing style changes back
to the actual code. This PR adds a buffer that pretends to be Rust,
applying any style attribute identifiers it finds. Also supports
completions with display of documentation. The effect of the currently
selected completion is previewed. Warning diagnostics appear on any
unrecognized identifier.
https://github.com/user-attachments/assets/af39ff0a-26a5-4835-a052-d8f642b2080c
Adds a `#[derive_inspector_reflection]` macro which allows these methods
to be enumerated and called by their name. The macro code changes were
95% generated by Zed Agent + Opus 4.
Release Notes:
* Added an element inspector for development. On debug builds,
`dev::ToggleInspector` will open a pane allowing inspecting of element
info and modifying styles.
This change instructs models to wrap new file content in Markdown fences
and introduces a parser for this format. The reasons are:
1. This is the format we put a lot of effort into explaining in the
system prompt.
2. Gemini really prefers to do it.
3. It adds an option for a model to think before writing the content
The `eval_zode` pass rate for GEmini models goes from 0% to 100%. Other
models were already at 100%, this hasn't changed.
Release Notes:
- N/A
Cleans things up now that wasm32-wasip2 is a supported target.
Before we merge, I will need to test against the current extensions to
make sure this is fine.
However, since our wit world isn't using any wasi package imports, this
shouldn't be a breaking change.
Release Notes:
- N/A
Don't hard wrap interactively; instead, soft wrap in `Bounded` mode
(editor width or 72 chars, whichever is smaller), and then hard wrap
before sending the commit message to git.
This also makes the soft wrap mode and width for commit messages
configurable in language settings.
Previously we didn't support soft wrap modes other than `EditorWidth` in
auto-height editors; I tried to add support for this by analogy with
code that was already there, and it seems to work pretty well.
Closes#27508
Release Notes:
- Fixed confusing wrapping behavior in commit message editors.
An unintended consequence of format on save is that we start (correctly)
informing the model that the file changed on disk every time the
formatter changes anything, which in turn can lead the model to things
like extra reads.
Until we have a solution in place to prevent this downside, we're going
back to not formatting on save by reverting
cb112a4012.
Release Notes:
- N/A
Open inspector with `dev: toggle inspector` from command palette or
`cmd-alt-i` on mac or `ctrl-alt-i` on linux.
https://github.com/user-attachments/assets/54c43034-d40b-414e-ba9b-190bed2e6d2f
* Picking of elements via the mouse, with scroll wheel to inspect
occluded elements.
* Temporary manipulation of the selected element.
* Layout info and JSON-based style manipulation for `Div`.
* Navigation to code that constructed the element.
Big thanks to @as-cii and @maxdeviant for sorting out how to implement
the core of an inspector.
Release Notes:
- N/A
---------
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
Co-authored-by: Federico Dionisi <code@fdionisi.me>
This PR updates the Zed LLM provider to fetch the available models from
the server instead of hard-coding them in the binary.
Release Notes:
- Updated the Zed provider to fetch the list of available language
models from the server.
Nearly all generated by Zed Agent + Claude Opus 4. I just wrote the test
`Args` struct and pointed it at the [2.0 release
notes](https://github.com/dtolnay/syn/releases/tag/2.0.0).
Release Notes:
- N/A
If gdb doesn't send a thread name we display the thread's process id in
the thread drop down menu instead now.
Co-authored-by: Remco Smits \<djsmits12@gmail.com\>
Release Notes:
- debugger beta: Handle bug where DAPs don't send thread names
Closes #ISSUE
Release Notes:
- debugger: Use integrated terminal for Python, allowing one to interact
with standard input/output when debugging Python projects.
This includes the following data:
- Where we spawned the session from (gutter, scenario list, custom form
filled by the user)
- Which debug adapter was used
- Which dock the debugger is in
Closes #ISSUE
Release Notes:
- debugger: Added telemetry for new session experience that includes
data about:
- How a session was spawned (gutter, scenario list or custom form)
- Which debug adapter was used
- Which dock the debugger is in
---------
Co-authored-by: Joseph T. Lyons <JosephTLyons@gmail.com>
This PR allows DAPs to define their own schema so users can see
completion items when editing their debug.json files.
Users facing this aren’t the biggest chance, but behind the scenes, this
affected a lot of code because we manually translated common fields from
Zed's config format to be adapter-specific. Now we store the raw JSON
from a user's configuration file and just send that.
I'm ignoring the Protobuf CICD error because the DebugTaskDefinition
message is not yet user facing and we need to deprecate some fields in
it.
Release Notes:
- debugger beta: Show completion items when editing debug.json
- debugger beta: Breaking change, debug.json schema now relays on what
DAP you have selected instead of always having the same based values.
---------
Co-authored-by: Remco Smits <djsmits12@gmail.com>
Co-authored-by: Cole Miller <m@cole-miller.net>
Co-authored-by: Cole Miller <cole@zed.dev>
This change improves `eval_extract_handle_command_output` results for
all models:
Model | Pass rate before | Pass rate after
----------------------------|------------------|----------------
claude-3.7-sonnet | 0.96 | 0.98
gemini-2.5-pro | 0.35 | 0.86
gpt-4.1 | 0.81 | 1.00
Part of this improvement comes from more robust evaluation, which now
accepts multiple possible outcomes. Another part is from the prompt
adaptation: addressing common Gemini failure modes, adding a few-shot
example, and, in the final commit, auto-rewriting instructions for
clarity and conciseness.
This change still needs validation from larger end-to-end evals.
Release Notes:
- N/A
TODO:
- [x] Release a new version of `zed_llm_client`
Release Notes:
- N/A
---------
Co-authored-by: Mikayla Maki <mikayla.c.maki@gmail.com>
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
1. Add system prompt: this is how it's called from threads. Previously,
we were sending
2. Fix an issue with writing agent thought into a newly created empty
file.
Release Notes:
- N/A
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Antonio Scandurra <me@as-cii.com>
Builds on top of https://github.com/zed-industries/zed/pull/30942
This turns on incremental compilation and decreases extension
compilation times by up to another 41%
Putting us at roughly 92% improved extension load times from what is in
the app today.
Because we only have a static engine, I can't reset the cache between
every run. So technically the benchmarks are always running with a
warmed cache. So the first extension we load will take the 8.8ms, and
then any subsequent extensions will be closer to the measured time in
this benchmark.
This is also measuring the entire load process, not just the
compilation. However, since this is the loading we likely think of when
thinking about extensions, I felt it was likely more helpful to see the
impact on the overall time.
This works because our extensions are largely the same Wasm bytecode
(SDK code + std lib functions etc) with minor changes in the trait impl.
The more different that extensions implementation is, there will be less
benefit, however, there will always be a large part of every extension
that is always the same across extensions, so this should be a speedup
regardless.
I used `moka` to provide a bound to the cache. We could use a bare
`DashMap`, however if there was some issue this could lead to a memory
leak. `moka` has some slight overhead, but makes sure that we don't go
over 32mb while using an LRU-style mechanism for deciding which
compilation artifacts to keep.
I measured our current extensions to take roughly 512kb in the cache.
Which means with a cap of 32mb, we can keep roughly 64 *completely
novel* extensions with no overlap. Since our extensions will have more
overlap than this though, we can actually keep much more in the cache
without having to worry about it.
#### Before:
```
load/1 time: [8.8301 ms 8.8616 ms 8.8931 ms]
change: [-0.1880% +0.3221% +0.8679%] (p = 0.23 > 0.05)
No change in performance detected.
```
#### After:
```
load/1 time: [5.1575 ms 5.1726 ms 5.1876 ms]
change: [-41.894% -41.628% -41.350%] (p = 0.00 < 0.05)
Performance has improved.
```
Release Notes:
- N/A
When building for the `x86_64-unknown-linux-musl` target, the default
`openssl-dev` is compiled for the GNU toolchain, which causes a build
error due to missing OpenSSL. This PR fixes the issue by avoiding the
use of OpenSSL on non-macOS and non-Windows platforms.
Release Notes:
- N/A
`ImageItem`'s `file` is returning `""` as its `path` for single-filed
worktrees like the ones are created for the images dropped from the OS.
`ImageItem::load_image_metadata` had used that `path` in FS operations
and the other method tried to use for icon resolving.
Rework the code to use a more specific, `worktree::File` instead and
always use the `abs_path` when dealing with paths from this `file`.
Release Notes:
- Fixed images not opening on drag and drop into the editor
Reverts zed-industries/zed#30812
This PR broke nightly builds on linux by adding an OpenSSL dependency to
the `remote_server` binary, which failed to link when building against
musl.
Release Notes:
- N/A
Closes https://github.com/zed-industries/zed/issues/29855
Implement tool use handling in Mistral provider, including mapping tool
call events and updating request construction. Add support for
tool_choice and parallel_tool_calls in Mistral API requests.
This works fine with all the existing models. Didn't touched anything
else but for future. Fetching models using their models api, deducting
tool call support, parallel tool calls etc should be done from model
data from api response.
<img width="547" alt="Screenshot 2025-05-06 at 4 52 37 PM"
src="https://github.com/user-attachments/assets/4c08b544-1174-40cc-a40d-522989953448"
/>
Tasks:
- [x] Add tool call support
- [x] Auto Fetch models using mistral api
- [x] Add tests for mistral crates.
- [x] Fix mistral configurations for llm providers.
Release Notes:
- agent: Add tool call support for existing mistral models
---------
Co-authored-by: Peter Tripp <peter@zed.dev>
Co-authored-by: Bennet Bo Fenner <bennet@zed.dev>
Precursor to other optimizations, but this already gets us a big
improvement.
Wasm compilation can easily be parallelized, and with all of the cores
on my M4 Max this already gets us an 86% improvement, bringing loading
an extension down to <9ms.
Not all setups will see this much improvement, but it will use the cores
available (it just uses rayon under the hood like we do elsewhere).
Since we load extensions in sequence, this should have a nice impact for
users with a lot of extensions.
#### Before
```
Benchmarking load: Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.5s, or reduce sample count to 70.
load time: [64.859 ms 64.935 ms 65.027 ms]
Found 8 outliers among 100 measurements (8.00%)
2 (2.00%) low mild
3 (3.00%) high mild
3 (3.00%) high severe
```
#### After
```
load time: [8.8685 ms 8.9012 ms 8.9344 ms]
change: [-86.347% -86.292% -86.237%] (p = 0.00 < 0.05)
Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
2 (2.00%) high mild
```
Release Notes:
- N/A
I was able to get this fix in upstream, so now we can have simpler code
paths for our model selection.
I also added a test to catch if this would cause a bug again in the
future.
Release Notes:
- N/A
This PR adds a new picker for viewing a list of jj bookmarks, like you
would with `jj bookmark list`.
This is an exploration around what it would look like to begin adding
some dedicated jj features to Zed.
This is behind the `jj-ui` feature flag.
Release Notes:
- N/A
This PR removes the `DerivePathStr` macro, as it is no longer used.
Also removes the `PathStaticStr` macro from `gpui_macros`, which was
also unused.
Release Notes:
- N/A
Replace dynamic downloading of WASI adapter with the provided crate.
More importantly, this makes sure we are using the same adapter version
as our version of wasmtime, which includes several fixes.
Arguably we could also at this point update to wasm32-wasip2 target and
remove this dependency as well if we want, but that might need further
testing.
Release Notes:
- N/A
I noticed the discussion in #28881, and had thought of exactly the same
a few days prior.
This implementation should preserve existing functionality fairly well.
I've added a dependency (serde_with) to allow the deserializer to skip
models which cannot be deserialized, which could occur if a future
provider, for instance, is added. Without this modification, such a
change could break all models. If extra dependencies aren't desired, a
manual implementation could be used instead.
- Closes#29369
Release Notes:
- Dynamically detect available Copilot Chat models, including all models
with tool support
---------
Co-authored-by: AidanV <aidanvanduyne@gmail.com>
Co-authored-by: imumesh18 <umesh4257@gmail.com>
Co-authored-by: Bennet Bo Fenner <bennet@zed.dev>
Co-authored-by: Agus Zubiaga <hi@aguz.me>
