- limit diagnostics sent to teacher
- improve parallelism in `ep` commands when using `--max-parallelism` by
only grouping by repo when instructed too (repo grouping is only useful
for context collection)
- update rejected and rated queries to also fetch settled editable
region
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Switches the Anthropic provider from hand-stamping `cache_control` onto
the last message content block over to Anthropic's top-level automatic
prompt caching, paired with explicit long-TTL (1h) anchors on the last
tool definition and on the system prompt.
The prefix order `tools` → `system` → `messages` satisfies Anthropic's
requirement that longer TTLs appear earlier in the prefix, so the static
prefix is cached for 1h (surviving idle gaps longer than the 5-minute
default) while the rapidly-changing conversation tail uses the
free-to-refresh 5-minute TTL via the top-level automatic breakpoint.
Three of the four available cache breakpoints are used (last tool,
system, automatic conversation), leaving one in reserve.
As a side benefit, this fixes a latent issue where the previous stamping
loop could place `cache_control` on a `Thinking` content block, which
the Anthropic API does not allow. Automatic caching is documented to
walk past ineligible blocks (including thinking) when selecting its
breakpoint, so we now delegate that responsibility to the server.
The new shape we send (when caching is enabled):
```json
{
"tools": [{ "...": "...", "cache_control": {"type": "ephemeral", "ttl": "1h"} }],
"system": [
{"type": "text", "text": "...", "cache_control": {"type": "ephemeral", "ttl": "1h"}}
],
"messages": [ /* no per-block cache_control */ ],
"cache_control": {"type": "ephemeral"}
}
```
Release Notes:
- Improved Anthropic prompt cache utilization, reducing latency and cost
for ongoing conversations
---------
Co-authored-by: Martin Ye <martinye022@gmail.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A
`ep split --stratify=project` will use the first component of
`cursor_path`, which is a worktree name. This is an alternative to
`--stratify=repo` in absence of `repository_url`
Release Notes:
- N/A
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
Release Notes:
- Added interleaved_reasoning option to openai compatible models
---
This PR adds the interleaved_reasoning option for OpenAI-compatible
models, addressing the issue described in
https://github.com/ggml-org/llama.cpp/issues/20837.
In my testing, enabling interleaved_reasoning not only resolved the
tool-calling issues encountered by Qwen3.5 models in llama.cpp, but also
appeared to improve the model's coding capabilities. I have also
verified the outgoing requests using a proxy to ensure the parameter is
being sent correctly.It is also likely that this change will benefit
other models and providers as well.
Note: While I used AI to assist with the implementation, I have reviewed
and tested the changes. As I am relatively new to Rust and the Zed
codebase, I would appreciate any feedback or suggestions for
improvement. I am happy to make further adjustments if needed.
Thank you all for building such an amazing editor!
Co-authored-by: Oleksiy Syvokon <oleksiy@zed.dev>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
This PR adds the `correctly_deleted_chars` field and updates `kept_rate`
to account for it, not just inserted chars.
It also adds `recall_rate` to measure coverage of reference
insertions/deletions.
Finally, it renames "final" to "reference" and "prediction" to
"candidate".
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- N/A
Stacked on https://github.com/zed-industries/zed/pull/50566.
Begin collecting kept chars rate, as well as the count of tree-sitter
errors in the code before and after applying the prediction.
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
- `language_model` no longer depends on provider-specific crates such as
`anthropic` and `open_ai` (inverted dependency)
- `language_model_core` was extracted from `language_model` which
contains the types for the provider-specific crates to convert to/from.
- `gpui::SharedString` has been extracted into its own crate (still
exposed by `gpui`), so `language_model_core` and provider API crates
don't have to depend on `gpui`.
- Removes some unnecessary `&'static str` | `SharedString` -> `String`
-> `SharedString` conversions across the codebase.
- Extracts the core logic of the cloud `LanguageModelProvider` into its
own crate with simpler dependencies.
Release Notes:
- N/A
---------
Co-authored-by: John Tur <john-tur@outlook.com>
Self-Review Checklist:
- [x] I've reviewed my own diff for quality, security, and reliability
- [ ] Unsafe blocks (if any) have justifying comments
- [ ] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Closes #ISSUE
Release Notes:
- N/A or Added/Fixed/Improved ...
This PR decouples `language_model`'s dependence on Zed-specific
implementation details. In particular
* `credentials_provider` is split into a generic `credentials_provider`
crate that provides a trait, and `zed_credentials_provider` that
implements the said trait for Zed-specific providers and has functions
that can populate a global state with them
* `zed_env_vars` is split into a generic `env_var` crate that provides
generic tooling for managing env vars, and `zed_env_vars` that contains
Zed-specific statics
* `client` is now dependent on `language_model` and not vice versa
Release Notes:
- N/A
This PR mostly moves some code around. It also adds a high-level
`format_expected_output` function that routes patch formatting to
specific prompt formats. This was `zeta_prompt` can format
`expected_output` for training.
Keeping everything prompt-related in a "pure" module (with no heavy
dependencies) makes it easier to write bindings.
Release Notes:
- N/A
The watcher had been broken for some time, but became even more broken
after the recent move of the queries.
This PR restores the reloading behavior for debug builds so that
languages are reloaded once a scheme file is changed.
Release Notes:
- N/A
This extracts a `language_core` crate from the existing `language`
crate, and creates a `grammars` data crate. The goal is to separate
tree-sitter grammar infrastructure, language configuration, and LSP
adapter types from the heavier buffer/editor integration layer in
`language`.
## Motivation
The `language` crate pulls in `text`, `theme`, `settings`, `rpc`,
`task`, `fs`, `clock`, `sum_tree`, and `fuzzy` — all of which are needed
for buffer integration (`Buffer`, `SyntaxMap`, `Outline`,
`DiagnosticSet`) but not for grammar parsing or language configuration.
Extracting the core types lets downstream consumers depend on
`language_core` without pulling in the full integration stack.
## Dependency graph after extraction
```
language_core ← gpui, lsp, tree-sitter, util, collections
grammars ← language_core, rust_embed, tree-sitter-{rust,python,...}
language ← language_core, text, theme, settings, rpc, task, fs, ...
languages ← language, grammars
```
## What moved to `language_core`
- `Grammar`, `GrammarId`, and all query config/builder types
- `LanguageConfig`, `LanguageMatcher`, bracket/comment/indent config
types
- `HighlightMap`, `HighlightId` (theme-dependent free functions
`highlight_style` and `highlight_name` stay in `language`)
- `LanguageName`, `LanguageId`
- `LanguageQueries`, `QUERY_FILENAME_PREFIXES`
- `CodeLabel`, `CodeLabelBuilder`, `Symbol`
- `Diagnostic`, `DiagnosticSourceKind`
- `Toolchain`, `ToolchainScope`, `ToolchainList`, `ToolchainMetadata`
- `ManifestName`
- `SoftWrap`
- LSP data types: `BinaryStatus`, `ServerHealth`,
`LanguageServerStatusUpdate`, `PromptResponseContext`, `ToLspPosition`
## What stays in `language`
- `Buffer`, `BufferSnapshot`, `SyntaxMap`, `Outline`, `DiagnosticSet`,
`LanguageScope`
- `LspAdapter`, `CachedLspAdapter`, `LspAdapterDelegate` (reference
`Arc<Language>` and `WorktreeId`)
- `ToolchainLister`, `LanguageToolchainStore` (reference `task` and
`settings` types)
- `ManifestQuery`, `ManifestProvider`, `ManifestDelegate` (reference
`WorktreeId`)
- Parser/query cursor pools, `PLAIN_TEXT`, point conversion functions
## What the `grammars` crate provides
- Embedded `.scm` query files and `config.toml` files for all built-in
languages (via `rust_embed`)
- `load_queries(name)`, `load_config(name)`,
`load_config_for_feature(name, grammars_loaded)`, and `get_file(path)`
functions
- `native_grammars()` for tree-sitter grammar registration (behind
`load-grammars` feature)
## Pre-cleanup (also in this PR)
- Removed unused `Option<&Buffer>` from
`LspAdapter::process_diagnostics`
- Removed unused `&App` from `LspAdapter::retain_old_diagnostic`
- Removed `fs: &dyn Fs` from `ToolchainLister` trait methods
(`PythonToolchainProvider` captures `fs` at construction time instead)
- Moved `Diagnostic`/`DiagnosticSourceKind` out of `buffer.rs` into
their own module
## Backward compatibility
The `language` crate re-exports everything from `language_core`, so
existing `use language::Grammar` (etc.) continues to work unchanged. The
only downstream change required is importing `CodeLabelExt` where
`.fallback_for_completion()` is called on the now-foreign `CodeLabel`
type.
Release Notes:
- N/A
---------
Co-authored-by: Agus Zubiaga <agus@zed.dev>
Co-authored-by: Tom Houlé <tom@tomhoule.com>
Closes#52115
## Context
Removes the third party edit prediction integration for Sweep AI ahead
of their servers shutting down.
This PR will not affect those who are already use or plan to use their
open weighted models. The code removed by this PR was required for
integrating with their proprietary API.
## Self-Review Checklist
<!-- Check before requesting review: -->
- [x] I've reviewed my own diff for quality, security, and reliability
- [x] Unsafe blocks (if any) have justifying comments
- [x] The content is consistent with the [UI/UX
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
- [x] Tests cover the new/changed behavior
- [x] Performance impact has been considered and is acceptable
Release Notes:
- Removed support for using the Sweep AI edit prediction provider
through their proprietary API, as the servers are shutting down
https://discord.com/channels/1100625416022138902/1100625417272045639/1480644297903575142,
https://x.com/wwzeng1/status/2033302698360180949
Increases timeouts, simplifies queries, and makes other changes needed to
pull data more efficiently.
Release Notes:
- N/A
---------
Co-authored-by: Ben Kunkle <ben@zed.dev>
For now, the only source of logprobs is `./run.py infer`, which writes
them to under the `.predictions` struct. `ep score` copies these values
to `.score`. There is some duplication (same value stored in two
places), which is unfortunate, but can't be fixed without reworking how
scores are stored.
Release Notes:
- N/A
This PR adds `v0318`, which is just like `v0316` but with bigger blocks.
It seems to perform best so far.
It also adds heuristics to avoid placing the marker token on closing
braces, which results in much nicer blocks.
Finally, it fixes a bug with inserting `<|marker_N|>` mid-line in
`v0316` and `v0317`.
Release Notes:
- N/A
V0316 differences from V0306:
- Use `<|marker_N|><|marker_N|>` instead of `NO_EDITS`
- All intermediate markers are now emitted (should be more friendly to
speculative decoding)
- No newline padding around marker tags
- No git conflict markers; we now use SeedCoder's EOS tag instead
V0317 is like V0316, but marker numbers are relative to the cursor block
(`<|marker-2|> ... -1 ... -0 ... +1 ... <|marker+2|>`)
Release Notes:
- N/A
## Summary
This PR enables token usage reporting in streaming responses for
OpenAI-compatible APIs (OpenAI, xAI/Grok, OpenRouter, etc).
## Problem
Currently, the token counter UI in the Agent Panel doesn't display usage
for some OpenAI-compatible providers because they don't return usage
data during streaming by default. According to OpenAI's API
documentation, the `stream_options.include_usage` parameter must be set
to `true` to receive usage statistics in streaming responses.
## Solution
- Added StreamOptions struct with `include_usage` field to the open_ai
crate
- Added `stream_options` field to the Request struct
- Automatically set `stream_options: { include_usage: true }` when
`stream: true`
- Updated edit_prediction requests with `stream_options: None`
(non-streaming)
## Testing
Tested with xAI Grok models - token counter now correctly shows usage
after sending a message.
## References
- [OpenAI Chat Completions API -
stream_options](https://platform.openai.com/docs/api-reference/chat/create#chat-create-stream_options)
- [xAI API Documentation](https://docs.x.ai/api)
This format generates fewer token while maintaining the quality:
```
Model Generated tokens ↓ DeltaChrF ↑
0306-seed-multi-regions 46,239 80.62
0304-seed-no-edits 110,871 80.61
0303-seed 271,457 79.62
```
In addition to the student format, this change adds a new teacher
prompt. It seems to be worse than the original, but I haven't optimized
it at all. Keeping it for now as a base for potential improvements.
Release Notes:
- N/A
Closes #ISSUE
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [ ] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Previously we didn't distinguish between an empty `.related_files[]` and
a case where context collection hadn't run yet. As a result, context
retrieval was always attempted for examples with empty `related_files`.
Release Notes:
- N/A
Closes #ISSUE
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [ ] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Emit client-side organization changed events through
`RefreshLlmTokenListener` so it produces the same `RefreshLlmTokenEvent`
used for server-pushed `UserUpdated` messages.
This keeps token refresh fan-out in one place.
Closes CLO-383.
Release Notes:
- N/A
---------
Co-authored-by: Tom Houlé <tom@tomhoule.com>
Closes #ISSUE
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [ ] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A *or* Added/Fixed/Improved ...
Closes #ISSUE
Before you mark this PR as ready for review, make sure that you have:
- [ ] Added a solid test coverage and/or screenshots from doing manual
testing
- [ ] Done a self-review taking into account security and performance
aspects
- [ ] Aligned any UI changes with the [UI
checklist](https://github.com/zed-industries/zed/blob/main/CONTRIBUTING.md#uiux-checklist)
Release Notes:
- N/A *or* Added/Fixed/Improved ...
This can be used like `ep predict --provider
baseten:V0131GitMergeMarkersPrefix`. Since it doesn't require
load_project, it can be used with captured requests.
Release Notes:
- N/A
Ben Kunkle