From f1f99331a4ffa99fb3408cd2ee66e02491a94d5a Mon Sep 17 00:00:00 2001
From: Oleksiy Syvokon <oleksiy@zed.dev>
Date: Mon, 11 May 2026 12:10:05 +0300
Subject: [PATCH] ep: Generate request IDs on the client (#56386)

This change lays the groundwork for canceling in-flight requests.
Without it, we always wait for a request to complete, even when we
already know its result won't be used.

Release Notes:

- N/A
---
 crates/cloud_llm_client/src/predict_edits_v3.rs | 1 +
 crates/edit_prediction/src/edit_prediction.rs   | 8 +++++---
 2 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/crates/cloud_llm_client/src/predict_edits_v3.rs b/crates/cloud_llm_client/src/predict_edits_v3.rs
index f934962848..ef21d593b9 100644
--- a/crates/cloud_llm_client/src/predict_edits_v3.rs
+++ b/crates/cloud_llm_client/src/predict_edits_v3.rs
@@ -5,6 +5,7 @@ use std::ops::Range;
 use strum::{AsRefStr, EnumString};
 
 pub const PREDICT_EDITS_MODE_HEADER_NAME: &str = "X-Zed-Predict-Edits-Mode";
+pub const PREDICT_EDITS_REQUEST_ID_HEADER_NAME: &str = "X-Zed-Predict-Edits-Request-Id";
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Serialize, Deserialize, AsRefStr, EnumString)]
 #[serde(rename_all = "snake_case")]
diff --git a/crates/edit_prediction/src/edit_prediction.rs b/crates/edit_prediction/src/edit_prediction.rs
index 7f835dfbdf..cc17ea3b93 100644
--- a/crates/edit_prediction/src/edit_prediction.rs
+++ b/crates/edit_prediction/src/edit_prediction.rs
@@ -3,8 +3,8 @@ use client::{Client, EditPredictionUsage, NeedsLlmTokenRefresh, UserStore, globa
 use cloud_api_client::LlmApiToken;
 use cloud_api_types::{OrganizationId, SubmitEditPredictionFeedbackBody};
 use cloud_llm_client::predict_edits_v3::{
-    PREDICT_EDITS_MODE_HEADER_NAME, PredictEditsMode, PredictEditsV3Request,
-    PredictEditsV3Response, RawCompletionRequest, RawCompletionResponse,
+    PREDICT_EDITS_MODE_HEADER_NAME, PREDICT_EDITS_REQUEST_ID_HEADER_NAME, PredictEditsMode,
+    PredictEditsV3Request, PredictEditsV3Response, RawCompletionRequest, RawCompletionResponse,
 };
 use cloud_llm_client::{
     EditPredictionRejectReason, EditPredictionRejection,
@@ -2601,6 +2601,7 @@ impl EditPredictionStore {
             .build_zed_llm_url("/predict_edits/v3", &[])?;
 
         let request = PredictEditsV3Request { input, trigger };
+        let request_id = uuid::Uuid::new_v4().to_string();
 
         let json_bytes = serde_json::to_vec(&request)?;
         let compressed = zstd::encode_all(&json_bytes[..], 3)?;
@@ -2610,7 +2611,8 @@ impl EditPredictionStore {
                 let builder = builder
                     .uri(url.as_ref())
                     .header("Content-Encoding", "zstd")
-                    .header(PREDICT_EDITS_MODE_HEADER_NAME, mode.as_ref());
+                    .header(PREDICT_EDITS_MODE_HEADER_NAME, mode.as_ref())
+                    .header(PREDICT_EDITS_REQUEST_ID_HEADER_NAME, request_id.as_str());
                 let builder = if let Some(preferred_experiment) = preferred_experiment.as_deref() {
                     builder.header(PREFERRED_EXPERIMENT_HEADER_NAME, preferred_experiment)
                 } else {