Introduce a new StreamingEditFileTool (#29733)

This pull request introduces a new tool for streaming edits. The
short-term goal is for this tool to replace the existing `EditFileTool`,
but we want to get this out the door as soon as possible so that we can
start testing it.

`StreamingEditFileTool` is mutually exclusive with `EditFileTool`. It
will be enabled by default for anyone who has the `agent-stream-edits`
feature flag, as well as people that set `assistant.stream_edits` to
`true` in their settings.

### Implementation

Streaming is achieved by requesting a completion while the `edit_file`
tool gets called. We invoke the model by taking the existing
conversation with the agent and appending a prompt specifically tailored
for editing. In that prompt, we ask the model to produce a stream of
`<old_text>`/`<new_text>` tags. As the model streams text in, we
incrementally parse it and start editing as soon as we can.

### Evals

Note that, as part of this pull request, I also defined some new evals
that I used to drive the behavior of the recursive LLM call. To run
them, use this command:

```bash
cargo test --package=assistant_tools --features eval -- eval_extract_handle_command_output
```

Or comment out the `#[cfg_attr(not(feature = "eval"), ignore)]` macro.

I recommend running them one at a time, because right now we don't
really have a way of orchestrating of all these evals. I think we should
invest into that effort once the new agent panel goes live.

Release Notes:

- N/A

---------

Co-authored-by: Nathan Sobo <nathan@zed.dev>
Co-authored-by: Bennet Bo Fenner <bennetbo@gmx.de>
Co-authored-by: Oleksiy Syvokon <oleksiy.syvokon@gmail.com>
This commit is contained in:
Antonio Scandurra 2025-05-01 17:37:43 +02:00 committed by GitHub
parent e3a2d52472
commit f891dfb358
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
32 changed files with 49077 additions and 20 deletions

View file

@ -6,7 +6,7 @@ use ::open_ai::Model as OpenAiModel;
use anthropic::Model as AnthropicModel;
use anyhow::{Result, bail};
use deepseek::Model as DeepseekModel;
use feature_flags::{Assistant2FeatureFlag, FeatureFlagAppExt};
use feature_flags::{AgentStreamEditsFeatureFlag, Assistant2FeatureFlag, FeatureFlagAppExt};
use gpui::{App, Pixels};
use indexmap::IndexMap;
use language_model::{CloudModel, LanguageModel};
@ -87,9 +87,14 @@ pub struct AssistantSettings {
pub profiles: IndexMap<AgentProfileId, AgentProfile>,
pub always_allow_tool_actions: bool,
pub notify_when_agent_waiting: NotifyWhenAgentWaiting,
pub stream_edits: bool,
}
impl AssistantSettings {
pub fn stream_edits(&self, cx: &App) -> bool {
cx.has_flag::<AgentStreamEditsFeatureFlag>() || self.stream_edits
}
pub fn are_live_diffs_enabled(&self, cx: &App) -> bool {
if cx.has_flag::<Assistant2FeatureFlag>() {
return false;
@ -218,6 +223,7 @@ impl AssistantSettingsContent {
profiles: None,
always_allow_tool_actions: None,
notify_when_agent_waiting: None,
stream_edits: None,
},
VersionedAssistantSettingsContent::V2(ref settings) => settings.clone(),
},
@ -245,6 +251,7 @@ impl AssistantSettingsContent {
profiles: None,
always_allow_tool_actions: None,
notify_when_agent_waiting: None,
stream_edits: None,
},
None => AssistantSettingsContentV2::default(),
}
@ -495,6 +502,7 @@ impl Default for VersionedAssistantSettingsContent {
profiles: None,
always_allow_tool_actions: None,
notify_when_agent_waiting: None,
stream_edits: None,
})
}
}
@ -550,6 +558,10 @@ pub struct AssistantSettingsContentV2 {
///
/// Default: "primary_screen"
notify_when_agent_waiting: Option<NotifyWhenAgentWaiting>,
/// Whether to stream edits from the agent as they are received.
///
/// Default: false
stream_edits: Option<bool>,
}
#[derive(Clone, Debug, Serialize, Deserialize, JsonSchema, PartialEq)]
@ -712,6 +724,7 @@ impl Settings for AssistantSettings {
&mut settings.notify_when_agent_waiting,
value.notify_when_agent_waiting,
);
merge(&mut settings.stream_edits, value.stream_edits);
merge(&mut settings.default_profile, value.default_profile);
if let Some(profiles) = value.profiles {
@ -843,6 +856,7 @@ mod tests {
profiles: None,
always_allow_tool_actions: None,
notify_when_agent_waiting: None,
stream_edits: None,
},
)),
}