Also adds reuse of the markdown documentation cache even when
completions are re-queried, so that markdown documentation doesn't
flicker when `is_incomplete: true` (completions provided by rust
analyzer always set this)
Release Notes:
- Added support for filtering language server completions instead of
re-querying.
Previously, LMDB was used for storing threads, but it consumed excessive
disk space and was capped at 1GB.
This change migrates thread storage to an SQLite database. Thread JSON
objects are now compressed using zstd.
I considered training a custom zstd dictionary and storing it in a
separate table. However, the additional complexity outweighed the modest
space savings (up to 20%). I ended up using the default dictionary
stored with data.
Threads can be exported relatively easily from outside the application:
```
$ sqlite3 threads.db "SELECT hex(data) FROM threads LIMIT 5;" |
xxd -r -p |
zstd -d |
fx
```
Benchmarks:
- Original heed database: 200MB
- Sqlite uncompressed: 51MB
- sqlite compressed (this PR): 4.0MB
- sqlite compressed with a trained dictionary: 3.8MB
Release Notes:
- Migrated thread storage to SQLite with compression
This PR adds a Danger check to remind engineers that any changes to our
various prompts need to be verified against the LLM Worker.
When changes to the prompt files are detected, we will fail the PR with
a message:
<img width="929" alt="Screenshot 2025-05-30 at 8 40 58 AM"
src="https://github.com/user-attachments/assets/79afab4e-e799-45f1-a90e-0fd7c9a73706"
/>
Once the corresponding changes have been made (or no changes to the LLM
Worker have been determined to be necessary), including the indicated
attestation message will convert the errors into informational messages:
<img width="926" alt="Screenshot 2025-05-30 at 8 41 52 AM"
src="https://github.com/user-attachments/assets/ff51c17a-7a76-46a7-b468-a7d864d480c3"
/>
Release Notes:
- N/A
This method was added in #20649 to be an alternative of `occlude` which
allows scroll events. It seems a bit arbitrary to only stop left mouse
downs, so this seems like it's probably an improvement.
Release Notes:
- N/A
tl;dr: This adds `.block_mouse_except_scroll()` which should typically
be used instead of `.occlude()` for cases when the mouse shouldn't
interact with elements drawn below an element. The rationale for
treating scroll events differently:
* Mouse move / click / styles / tooltips are for elements the user is
interacting with directly.
* Mouse scroll events are about finding the current outer scroll
container.
Most use of `occlude` should probably be switched to this, but I figured
I'd derisk this change by minimizing behavior changes to just the 3 uses
of `block_mouse_except_scroll`.
GPUI changes:
* Added `InteractiveElement::block_mouse_except_scroll()`, and removes
`stop_mouse_events_except_scroll()`
* Added `Hitbox::should_handle_scroll()` to be used when handling scroll
wheel events.
* `Window::insert_hitbox` now takes `HitboxBehavior` instead of
`occlude: bool`.
- `false` for that bool is now `HitboxBehavior::Normal`.
- `true` for that bool is now `HitboxBehavior::BlockMouse`.
- The new mode is `HitboxBehavior::BlockMouseExceptScroll`.
* Removes `Default` impl for `HitboxId` since applications should not
manually create `HitboxId(0)`.
Release Notes:
- N/A
This PR adds a new `intent` field to completion requests to assist in
categorizing them correctly.
Release Notes:
- N/A
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Was touching this part of the Agent Panel and thought it could be a
quick name consistency win here, so it is aligned with the terminology
we're currently actively using in the product/docs.
Release Notes:
- N/A
Follow up to https://github.com/zed-industries/zed/pull/31470.
I started looking at config and changed preferred_completion_mode to
burn to only find its max so made changes to align it better with
rebrand. As this is in preview build now.
This doesn't touch zed_llm_client. Only the Zed changes the code and doc
to match the new UI of burn mode. There are still more things to be
renamed, though.
Release Notes:
- N/A
---------
Signed-off-by: Umesh Yadav <git@umesh.dev>
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
One caveat with this PR is that the keybinding still doesn't work for text threads. Will do that in a follow-up.
Release Notes:
- agent: Added a keybinding to toggle Burn Mode on and off.
Previously, the click on the backdrop to dismiss the message editing was
unreliable. You would click on it and sometimes it would work and others
it wouldn't. This PR fixes that now.
Release Notes:
- agent: Fixes the previous message dismissal by clicking on the
backdrop
Previously, when you clicked on a previous message to edit it and then
dismissed it, your focus would jump to the buffer. This caught me
several times as the most obvious place to return to for me was the
agent panel main message editor, so I can continue prompting something
else. And this is what this PR changes.
Release Notes:
- agent: Improved previous message editing UX by returning focus to the
main panel's text area after dismissing it.
This expands our deserialization of JSON from models to be more tolerant
of different variations that the model may send, including
capitalization, wrapping things in objects vs. being plain strings, etc.
Also when deserialization fails, it reports the entire error in the JSON
so we can see what failed to deserialize. (Previously these errors were
very unhelpful at diagnosing the problem.)
Finally, also removes the `WrappedText` variant since the custom
deserializer just turns that style of JSON into a normal `Text` variant.
Release Notes:
- N/A
This PR improves the consecutive tool call UX by allowing users to
quickly continue an interrupted with one-click. What we do here is
insert a hidden "Continue" message that will just nudge the LLM to keep
going. We're also using the opportunity to upsell the previously called
"Max Mode", now rebranded as "Burn Mode", which allows users to don't be
interrupted anymore if they ever have 25 consecutive tool calls again.
Release Notes:
- agent: Improve consecutive tool call UX by allowing users to quickly
continue an interrupted thread with one click.
---------
Co-authored-by: Ben Brandt <benjamin.j.brandt@gmail.com>
Co-authored-by: Agus Zubiaga <hi@aguz.me>
Co-authored-by: Agus Zubiaga <agus@zed.dev>
This fixes an issue where requests were failing when MCP servers were
registering tools with the same name.
We now prefix the tool names with the context server name, in the UI we
still show the name that the MCP server gives us
Release Notes:
- agent: Fix an error were requests would fail if two MCP servers were
using an identical tool name
Thanks @mgsloan for introducing `stop_mouse_events_except_scroll` which
is exactly what we want here!
Release Notes:
- Fixed being unable to scroll editors when the cursor is positioned on
diff hunk controls.
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.
Follow up to https://github.com/zed-industries/zed/pull/31353. Just
ensuring we're walking toward a more consistent use of the multiple
terms we have floating around in the AI realm. In this case, `thread` is
the term for the now default view, the one that has agentic features;
`text_thread` is the term for the original view, the one where it's just
text. The settings now reflect this. Also took advantage of the
opportunity to add some docs, too.
Release Notes:
- N/A
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.
Related discussions #30240#30596
Release Notes:
- Added an option on the settings file to choose either the `agent`
panel or the `thread` panel as the default assistant panel when you
first open it. On `settings.json`:
```json
{
"agent": {
"default_view": "thread", // default is agent
}
}
```
Release Notes:
- agent: Added a space after @-mentioning something in the message
editor to allow for continuous typing.
---------
Co-authored-by: Peter Tripp <peter@zed.dev>
Co-authored-by: Danilo Leal <daniloleal09@gmail.com>
Addresses an issue where the agent context token meter in the panel
toolbar (showing usage like "X / Y tokens") failed to update its count
after the user cleared the current context via the context editor UI.
While the meter updated correctly when adding items, clearing them left
the display showing the old count.
The root cause was traced to the `ContextStore::clear` method in
`crates/agent/src/context_store.rs`. This method correctly cleared the
internal data structures holding the context items but neglected to call
`cx.notify()` to inform listeners of the state change. Consequently, the
UI components responsible for displaying the token count were not
triggered to re-render with the new (presumably lower) count.
This PR fixes the issue by adding the missing `cx.notify()` call to the
`ContextStore::clear` method. This ensures listeners are notified when
the context set is cleared, allowing the token meter UI to update
correctly.
Release Notes:
- Fixed an issue where the agent context token meter did not update when
the context was cleared.
This PR introduces a new `agent: chat with follow` action that
automatically enables "Follow Agent" when submitting a chat message with
`cmd-enter` or `ctrl-enter`. This is experimental. I'm not super
thrilled with the name, but the root action to submit a chat is called
`agent: chat`, so I'm following that wording. I'm also unsure if the
binding feels right or not.
Release Notes:
- Added an `agent: chat with follow` action via `cmd-enter` on macOS and
`ctrl-enter` on Linux
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>
- Rename current_user_account_too_young to account_too_young for
consistency
- Add has_overdue_invoices field to track billing status
- Block edit predictions when user has overdue invoices
- Add overdue invoice warning to inline completion menu
Release Notes:
- N/A
---------
Co-authored-by: Marshall Bowers <git@maxdeviant.com>
This is a partial reversion of
https://github.com/zed-industries/zed/pull/30878 - having it always
scroll to bottom whenever a new message is added makes it so that when
you're scrolled up, you don't have time to read what you're trying to
read before it autoscrolls to the end.
@danilo-leal when you're back, we can pair on addressing that in a
different way!
Release Notes:
- Fixed bug where scrolling up in the agent panel didn't prevent
automatic scroll-to-end whenever a new message arrived.
By consuming the event during processing we save a few clones during
event processing.
Overall in this PR we save one Clone each during:
- Paste to the terminal
- Writing to the terminal
- Setting the title
- On every terminal transaction
- On every ViMotion when not using shift
Release Notes:
- N/A
This PR replaces some `update()` calls with either `read()` or
`read_with()` when the `update()` call performed read-only operations on
the entity.
Many more likely exist, will follow-up with more PRs.
Release Notes:
- N/A
Right now you find this out the first time you try and submit a
completion.
These changes communicate much earlier to the user what the issue is
with their account and what they can do about it.
Release Notes:
- N/A
---------
Co-authored-by: Antonio Scandurra <me@as-cii.com>
This PR fixes an issue where the eval was incorrectly pulling the
provider/model from the user settings, which could cause problems when
running certain evals.
Was introduced in #30168 due to the restructuring after the removal of
the `assistant` crate.
Release Notes:
- N/A
This is a follow-up to https://github.com/zed-industries/zed/pull/31217
that removes the last turn after we get a `refusal` stop reason, as
advised by the Anthropic docs.
Meant to include it in that PR, but accidentally merged it before
pushing these changes 🤦🏻♂️.
Release Notes:
- N/A
https://github.com/zed-industries/zed/issues/30972 brought up another
case where our context is not enough to track the actual source of the
issue: we get a general top-level error without inner error.
The reason for this was `.ok_or_else(|| anyhow!("failed to read HEAD
SHA"))?; ` on the top level.
The PR finally reworks the way we use anyhow to reduce such issues (or
at least make it simpler to bubble them up later in a fix).
On top of that, uses a few more anyhow methods for better readability.
* `.ok_or_else(|| anyhow!("..."))`, `map_err` and other similar error
conversion/option reporting cases are replaced with `context` and
`with_context` calls
* in addition to that, various `anyhow!("failed to do ...")` are
stripped with `.context("Doing ...")` messages instead to remove the
parasitic `failed to` text
* `anyhow::ensure!` is used instead of `if ... { return Err(...); }`
calls
* `anyhow::bail!` is used instead of `return Err(anyhow!(...));`
Release Notes:
- N/A
Some providers sometimes send `{ "type": "text", "text": ... }` instead
of just the text as a string. Now we accept those instead of erroring.
Release Notes:
- N/A
Michael Sloan