This PR significantly improves the quality of the initial file search
that occurs when the model doesn't yet know the full path to a file it
needs to read/edit.
Previously, the assertions in file_search often failed on main as the
model attempted to guess full file paths. On this branch, it reliably
calls `find_path` (previously `path_search`) before reading files.
After getting the model to find paths first, I noticed it would try
using `grep` instead of `path_search`. This motivated renaming
`path_search` to `find_path` (continuing the analogy to unix commands)
and adding system prompt instructions about proper tool selection.
Note: I know the command is just called `find`, but that seemed too
general.
In my eval runs, the `file_search` example improved from 40% ± 10% to
98% ± 2%. The only assertion I'm seeing occasionally fail is "glob
starts with `**` or project". We can probably add some instructions in
that regard.
Release Notes:
- N/A
Closes#27171
The `outline_panel::Open` action seems to open the outline panel, but
instead, it moves the editor's cursor to the position of the selected
entry in the outline panel. This PR renames it to
`outline_panel::OpenSelectedEntry` for better clarity.
Meanwhile, there is an existing action, `outline_panel::ToggleFocus`,
that should be used for opening the outline panel.
Todo:
- [x] Added migration
Release Notes:
- Renamed `outline_panel::Open` to `outline_panel::OpenSelectedEntry`
for better clarity.
Release Notes:
- agent: Replace `bash` tool with `terminal` tool which uses the current
shell
---------
Co-authored-by: Bennet <bennet@zed.dev>
Co-authored-by: Antonio <antonio@zed.dev>
This enables hiding mouse cursor even on cursor movements like up, down,
etc. or selections made using keyboard, etc.
Renamed existing boolean setting "hide_mouse_while_typing" to
"hide_mouse". It can have three values: `on_typing_and_movement`,
`on_typing`, `never`.
Release Notes:
- Now mouse cursor hides even when you navigate, or make selections
using keyboard in editor. This behavior can be changed by setting
`hide_mouse` to `on_typing_and_movement`, `on_typing` or `never`.
This PR updates two existing settings to use the settings migrator
instead of a manually implemented visitor. Both of these settings were
changed prior to the introduction of automatic migrations and the
visitor ensured that the settings were kept backwards compatible. See
https://github.com/zed-industries/zed/pull/22200 and
https://github.com/zed-industries/zed/pull/22364 respectively.
WIth this change, existing user configurations are updated accordingly
and the corresponding settings can derive `Deserialize` again.
I also added tests for the replacement of settings values, as there was
no test for this behaviour. Additionally, I added a seperate test for
the existing migration of `always_show_close_button`, since that
migration updated both the key and value.
Release Notes:
- N/A
There is a drawback to how we currently write our migrations:
For example:
1. Suppose we change one of our actions from a string to an array and
rename it, then roll out the preview build:
Before: `"ctrl-x": "editor::GoToPrevHunk"`
Latest: `"ctrl-x": ["editor::GoToPreviousHunk", { "center_cursor": true
}]`
To handle this, we wrote migration `A` to convert the string to an
array.
2. Now, suppose we decide to change it back to a string:
- User who hasn't migrated yet on Preview: `"ctrl-x":
"editor::GoToPrevHunk"`
- User who has migrated on Preview: `"ctrl-x":
["editor::GoToPreviousHunk", { "center_cursor": true }]`
- Latest: `"ctrl-x": "editor::GoToPreviousHunk"`
To handle this, we would need to remove migration `A` and add two more
migrations:
- **Migration B**: `"ctrl-x": "editor::GoToPrevHunk"` -> `"ctrl-x":
"editor::GoToPreviousHunk"`
- **Migration C**: `"ctrl-x": ["editor::GoToPreviousHunk", {
"center_cursor": true }]` -> `"ctrl-x": "editor::GoToPreviousHunk"`
Nice. But over time, this keeps increasing, making it impossible to
track outdated versions and handle all cases. Missing a case means users
stuck on `"ctrl-x": "editor::GoToPrevHunk"` will remain there and won't
be automatically migrated to the latest state.
---
To fix this, we introduce versioned migrations. Instead of removing
migration `A`, we simply write a new migration that takes the user to
the latest version—i.e., in this case, migration `C`.
- A user who hasn't migrated before will go through both migrations `A`
and `C` in order.
- A user who has already migrated will only go through `C`, since `A`
wouldn't change anything for them.
With incremental migrations, we only need to write migrations on top of
the latest state (big win!), as know internally they all would be on
latest state. You *must not* modify previous migrations. Always create
new ones instead.
This also serves as base for only prompting user to migrate, when
feature reaches stable. That way, preview and stable keymap and settings
are in sync.
cc: @mgsloan @ConradIrwin @probably-neb
Release Notes:
- N/A
