We've realized that a lot of the logic within an
`EditPredictionProvider` is not specific to a particular edit prediction
model / service. Rather, it is just the generic state management
required to perform edit predictions at all in Zed. We want to move to a
setup where there's one "built-in" edit prediction provider in Zed,
which can be pointed at different edit prediction models. The only logic
that is different for different models is how we construct the prompt,
send the request, and parse the output.
This PR also changes the behavior of the staff-only `zeta2` feature flag
so that in only gates your *ability* to use Zeta2, but you can still use
your local settings file to choose between different edit prediction
models/services: zeta1, zeta2, and sweep.
This PR also makes zeta1's outcome reporting and prediction-rating
features work with all prediction models, not just zeta1.
To do:
* [x] remove duplicated logic around sending cloud requests between
zeta1 and zeta2
* [x] port the outcome reporting logic from zeta to zeta2.
* [x] get the "rate completions" modal working with all EP models
* [x] display edit prediction diff
* [x] show edit history events
* [x] remove the original `zeta` crate.
Release Notes:
- N/A
---------
Co-authored-by: Agus Zubiaga <agus@zed.dev>
Co-authored-by: Ben Kunkle <ben@zed.dev>
We've been considering removing workspace-hack for a couple reasons:
- Lukas ran into a situation where its build script seemed to be causing
spurious rebuilds. This seems more likely to be a cargo bug than an
issue with workspace-hack itself (given that it has an empty build
script), but we don't necessarily want to take the time to hunt that
down right now.
- Marshall mentioned hakari interacts poorly with automated crate
updates (in our case provided by rennovate) because you'd need to have
`cargo hakari generate && cargo hakari manage-deps` after their changes
and we prefer to not have actions that make commits.
Currently removing workspace-hack causes our workspace to grow from
~1700 to ~2000 crates being built (depending on platform), which is
mainly a problem when you're building the whole workspace or running
tests across the the normal and remote binaries (which is where
feature-unification nets us the most sharing). It doesn't impact
incremental times noticeably when you're just iterating on `-p zed`, and
we'll hopefully get these savings back in the future when
rust-lang/cargo#14774 (which re-implements the functionality of hakari)
is finished.
Release Notes:
- N/A
Also:
* Adds tests for can_collect_data.
* Temporarily removes collection of diagnostics.
Release Notes:
- Edit Prediction: Fixed a bug where requests were marked eligible for
data collection despite the recent edit history in the request involving
files that may not be open source. The requests affected by this bug
will not be used in training data.
See discussion on #36564
Adds a simple ad-hoc substring matching pattern language which allows
skipping a bounded number of chars between matched substrings. Before
this change compiling the regex was taking ~120ms on a fast machine and
~8mb of memory. This new version is way faster and uses minimal memory.
Checked the behavior of this vs by running it against 10k licenses that
happened to be in my home dir. There were only 4 differences of behavior
with the regex implementation, and these were false negatives for the
regex implementation that are true positives with the new one.
Of the ~10k licenses in my home dir, ~1k do not match one of these
licenses, usually because it's GPL/MPL/etc.
Release Notes:
- N/A
See discussion on #36564. Makes the license regexes a less fragile by
not matching on symbols, while also excluding cases where a long file
ends with a valid license. Also adds Zlib license, a commented out test
to check all license-like files discovered in the homedir, and more
testcases.
Not too happy with the efficiency here, on my quite good computer it
takes ~120ms to compile the regex and allocates ~8mb for it. This is
just not a great use of regexes, I think something using eager substring
matching would be much more efficient - hoping to followup with that.
Release Notes:
- Edit Prediction: Added Zlib license to open-source licenses eligible
for data collection.
This PR updates the Zed Edit Prediction provider to acquire the LLM
token from Cloud instead of Collab to allow using Edit Predictions even
when disconnected from or unable to connect to the Collab server.
Release Notes:
- N/A
---------
Co-authored-by: Richard Feldman <oss@rtfeldman.com>
This includes making sure that both the agent panel and Zed's edit
prediction have a consistent narrative when it comes to onboarding users
into the AI features, considering the possible different plans and
conditions (such as being signed in/out, account age, etc.)
Release Notes:
- N/A
---------
Co-authored-by: Bennet Bo Fenner <53836821+bennetbo@users.noreply.github.com>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
This adds a "workspace-hack" crate, see
[mozilla's](https://hg.mozilla.org/mozilla-central/file/3a265fdc9f33e5946f0ca0a04af73acd7e6d1a39/build/workspace-hack/Cargo.toml#l7)
for a concise explanation of why this is useful. For us in practice this
means that if I were to run all the tests (`cargo nextest r
--workspace`) and then `cargo r`, all the deps from the previous cargo
command will be reused. Before this PR it would rebuild many deps due to
resolving different sets of features for them. For me this frequently
caused long rebuilds when things "should" already be cached.
To avoid manually maintaining our workspace-hack crate, we will use
[cargo hakari](https://docs.rs/cargo-hakari) to update the build files
when there's a necessary change. I've added a step to CI that checks
whether the workspace-hack crate is up to date, and instructs you to
re-run `script/update-workspace-hack` when it fails.
Finally, to make sure that people can still depend on crates in our
workspace without pulling in all the workspace deps, we use a `[patch]`
section following [hakari's
instructions](https://docs.rs/cargo-hakari/0.9.36/cargo_hakari/patch_directive/index.html)
One possible followup task would be making guppy use our
`rust-toolchain.toml` instead of having to duplicate that list in its
config, I opened an issue for that upstream: guppy-rs/guppy#481.
TODO:
- [x] Fix the extension test failure
- [x] Ensure the dev dependencies aren't being unified by Hakari into
the main dependencies
- [x] Ensure that the remote-server binary continues to not depend on
LibSSL
Release Notes:
- N/A
---------
Co-authored-by: Mikayla <mikayla@zed.dev>
Co-authored-by: Mikayla Maki <mikayla.c.maki@gmail.com>
This PR removes the dependents of the `language_models` crate.
The following types have been moved from `language_models` to
`language_model` to facilitate this:
- `LlmApiToken`
- `RefreshLlmTokenListener`
- `MaxMonthlySpendReachedError`
- `PaymentRequiredError`
With this change only `zed` now depends on `language_models`.
Release Notes:
- N/A
Closes https://github.com/zed-industries/zed/issues/10122
Closes https://github.com/zed-industries/zed/issues/25034
When formatting buffers or reloading them after they change on disk, we
performed a diff between the buffer's current contents and the new
content. We need this diff in order preserve the positions of cursors
and other decorations when updating the buffer's text.
In order to handle changes within lines, we would previously compute a
*character-wise* diff. This was extremely expensive for large files.
This PR gets rid of the character-wise diff, and instead performs a
normal line-wise diff. Then, for certain replace hunks, we compute a
secondary word-based diff. Also, I've switched to the
[`imara-diff`](https://github.com/pascalkuthe/imara-diff) crate, instead
of `similar`.
Release Notes:
- Fixed a hang that could occur when large files were changed on disk or
formatted.
This fixes a bug where we'd update your settings to an invalid state if
you were using the old `inline_completion_provider` setting, then
onboarded to Zeta, then migrated your settings.
Release Notes:
- N/A
Co-authored-by: Michael Sloan <mgsloan@gmail.com>
Co-authored-by: Agus Zubiaga <hi@aguz.me>
This PR makes it so Zeta respects the `x-zed-minimum-required-version`
header sent back from the server.
If the current Zed version is strictly less than the indicated minimum
required version, we show an error indicating that an update is required
in order to continue using Zeta:
<img width="472" alt="Screenshot 2025-02-12 at 6 15 44 PM"
src="https://github.com/user-attachments/assets/51b85dff-23a0-464c-ae4b-5b8f46b5915c"
/>
Release Notes:
- N/A
This PR makes it so we send up the diagnostic groups as additional data
with the edit prediction request.
We're not yet making use of them, but we are recording them so we can
use them later (e.g., to train the model).
Release Notes:
- N/A
---------
Co-authored-by: Nathan <nathan@zed.dev>
This PR updates the `zeta` crate to use the predictive edit DTOs defined
in the `zed_llm_client` crate.
This way we aren't duplicating their definitions (and risk them going
out of sync).
Release Notes:
- N/A
When a user types, chances are the model will anticipate what they are
about to do. Previously, we would continuously cancel the pending
completion until the user stopped typing. With this commit, we allow at
most two completions to race with each other (the first and the last
one):
- If the completion that was requested first completes first, we will
show it (assuming we can interpolate it) but avoid canceling the last
one.
- When the completion that was requested last completes, we will cancel
the first one if it's pending.
In both cases, if a completion is already on-screen we have a special
case for when the completions are just insertions and the new completion
is a superset of the existing one. In this case, we will replace the
existing completion with the new one. Otherwise we will keep showing the
old one to avoid thrashing the UI.
This should make latency a lot better. Note that I also reduced the
debounce timeout to 8ms.
Release Notes:
- N/A
This PR adds keyboard shortcuts to common interactions you might want to
do in the Zeta rating panel.
This PR also adds a way to fake inline completion requests, as well as
the test data used to create this PR, to make it easier to adjust the UI
in the future.
It also changes the status bar from the text "Zeta" to "ζ", because I
thought it looked cool.
Release Notes:
- N/A
Piotr OsiewiczandConrad Irwin
Julia Ryan
Michael Sloan
Marshall BowersandRichard Feldman
Bennet Bo Fenner
Antonio ScandurraandThorsten
