Route all LLM traffic through cloud.zed.dev (#34404)

This PR makes it so all LLM traffic is routed through `cloud.zed.dev`.

We're already routing `llm.zed.dev` to `cloud.zed.dev` on the server,
but we want to standardize on `cloud.zed.dev` moving forward.

Release Notes:

- N/A

Cargo.lock

@@ -9036,7 +9036,6 @@ dependencies = [
  "credentials_provider",
  "deepseek",
  "editor",
- "feature_flags",
  "fs",
  "futures 0.3.31",
  "google_ai",
@@ -18390,7 +18389,6 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "client",
- "feature_flags",
  "futures 0.3.31",
  "gpui",
  "http_client",

crates/feature_flags/src/feature_flags.rs

@@ -98,17 +98,6 @@ impl FeatureFlag for AcpFeatureFlag {
     const NAME: &'static str = "acp";
 }
 
-pub struct ZedCloudFeatureFlag {}
-
-impl FeatureFlag for ZedCloudFeatureFlag {
-    const NAME: &'static str = "zed-cloud";
-
-    fn enabled_for_staff() -> bool {
-        // Require individual opt-in, for now.
-        false
-    }
-}
-
 pub trait FeatureFlagViewExt<V: 'static> {
     fn observe_flag<T: FeatureFlag, F>(&mut self, window: &Window, callback: F) -> Subscription
     where

crates/http_client/src/http_client.rs

@@ -226,21 +226,10 @@ impl HttpClientWithUrl {
     }
 
     /// Builds a Zed LLM URL using the given path.
-    pub fn build_zed_llm_url(
+    pub fn build_zed_llm_url(&self, path: &str, query: &[(&str, &str)]) -> Result<Url> {
-        &self,
-        path: &str,
-        query: &[(&str, &str)],
-        use_cloud: bool,
-    ) -> Result<Url> {
         let base_url = self.base_url();
         let base_api_url = match base_url.as_ref() {
-            "https://zed.dev" => {
+            "https://zed.dev" => "https://cloud.zed.dev",
-                if use_cloud {
-                    "https://cloud.zed.dev"
-                } else {
-                    "https://llm.zed.dev"
-                }
-            }
             "https://staging.zed.dev" => "https://llm-staging.zed.dev",
             "http://localhost:3000" => "http://localhost:8787",
             other => other,

crates/language_models/Cargo.toml

@@ -28,7 +28,6 @@ credentials_provider.workspace = true
 copilot.workspace = true
 deepseek = { workspace = true, features = ["schemars"] }
 editor.workspace = true
-feature_flags.workspace = true
 fs.workspace = true
 futures.workspace = true
 google_ai = { workspace = true, features = ["schemars"] }

crates/language_models/src/provider/cloud.rs

@@ -2,7 +2,6 @@ use anthropic::AnthropicModelMode;
 use anyhow::{Context as _, Result, anyhow};
 use chrono::{DateTime, Utc};
 use client::{Client, ModelRequestUsage, UserStore, zed_urls};
-use feature_flags::{FeatureFlagAppExt as _, ZedCloudFeatureFlag};
 use futures::{
     AsyncBufReadExt, FutureExt, Stream, StreamExt, future::BoxFuture, stream::BoxStream,
 };
@@ -137,7 +136,6 @@ impl State {
         cx: &mut Context<Self>,
     ) -> Self {
         let refresh_llm_token_listener = RefreshLlmTokenListener::global(cx);
-        let use_cloud = cx.has_flag::<ZedCloudFeatureFlag>();
 
         Self {
             client: client.clone(),
@@ -165,7 +163,7 @@ impl State {
                             .await;
                     }
 
-                    let response = Self::fetch_models(client, llm_api_token, use_cloud).await?;
+                    let response = Self::fetch_models(client, llm_api_token).await?;
                     this.update(cx, |this, cx| {
                         this.update_models(response, cx);
                     })
@@ -184,7 +182,7 @@ impl State {
                     let llm_api_token = this.llm_api_token.clone();
                     cx.spawn(async move |this, cx| {
                         llm_api_token.refresh(&client).await?;
-                        let response = Self::fetch_models(client, llm_api_token, use_cloud).await?;
+                        let response = Self::fetch_models(client, llm_api_token).await?;
                         this.update(cx, |this, cx| {
                             this.update_models(response, cx);
                         })
@@ -268,18 +266,13 @@ impl State {
     async fn fetch_models(
         client: Arc<Client>,
         llm_api_token: LlmApiToken,
-        use_cloud: bool,
     ) -> Result<ListModelsResponse> {
         let http_client = &client.http_client();
         let token = llm_api_token.acquire(&client).await?;
 
         let request = http_client::Request::builder()
             .method(Method::GET)
-            .uri(
+            .uri(http_client.build_zed_llm_url("/models", &[])?.as_ref())
-                http_client
-                    .build_zed_llm_url("/models", &[], use_cloud)?
-                    .as_ref(),
-            )
             .header("Authorization", format!("Bearer {token}"))
             .body(AsyncBody::empty())?;
         let mut response = http_client
@@ -543,7 +536,6 @@ impl CloudLanguageModel {
         llm_api_token: LlmApiToken,
         app_version: Option<SemanticVersion>,
         body: CompletionBody,
-        use_cloud: bool,
     ) -> Result<PerformLlmCompletionResponse> {
         let http_client = &client.http_client();
 
@@ -551,11 +543,9 @@ impl CloudLanguageModel {
         let mut refreshed_token = false;
 
         loop {
-            let request_builder = http_client::Request::builder().method(Method::POST).uri(
+            let request_builder = http_client::Request::builder()
-                http_client
+                .method(Method::POST)
-                    .build_zed_llm_url("/completions", &[], use_cloud)?
+                .uri(http_client.build_zed_llm_url("/completions", &[])?.as_ref());
-                    .as_ref(),
-            );
             let request_builder = if let Some(app_version) = app_version {
                 request_builder.header(ZED_VERSION_HEADER_NAME, app_version.to_string())
             } else {
@@ -782,7 +772,6 @@ impl LanguageModel for CloudLanguageModel {
                 let model_id = self.model.id.to_string();
                 let generate_content_request =
                     into_google(request, model_id.clone(), GoogleModelMode::Default);
-                let use_cloud = cx.has_flag::<ZedCloudFeatureFlag>();
                 async move {
                     let http_client = &client.http_client();
                     let token = llm_api_token.acquire(&client).await?;
@@ -798,7 +787,7 @@ impl LanguageModel for CloudLanguageModel {
                         .method(Method::POST)
                         .uri(
                             http_client
-                                .build_zed_llm_url("/count_tokens", &[], use_cloud)?
+                                .build_zed_llm_url("/count_tokens", &[])?
                                 .as_ref(),
                         )
                         .header("Content-Type", "application/json")
@@ -847,9 +836,6 @@ impl LanguageModel for CloudLanguageModel {
         let intent = request.intent;
         let mode = request.mode;
         let app_version = cx.update(|cx| AppVersion::global(cx)).ok();
-        let use_cloud = cx
-            .update(|cx| cx.has_flag::<ZedCloudFeatureFlag>())
-            .unwrap_or(false);
         let thinking_allowed = request.thinking_allowed;
         match self.model.provider {
             zed_llm_client::LanguageModelProvider::Anthropic => {
@@ -888,7 +874,6 @@ impl LanguageModel for CloudLanguageModel {
                             provider_request: serde_json::to_value(&request)
                                 .map_err(|e| anyhow!(e))?,
                         },
-                        use_cloud,
                     )
                     .await
                     .map_err(|err| match err.downcast::<ApiError>() {
@@ -941,7 +926,6 @@ impl LanguageModel for CloudLanguageModel {
                             provider_request: serde_json::to_value(&request)
                                 .map_err(|e| anyhow!(e))?,
                         },
-                        use_cloud,
                     )
                     .await?;
 
@@ -982,7 +966,6 @@ impl LanguageModel for CloudLanguageModel {
                             provider_request: serde_json::to_value(&request)
                                 .map_err(|e| anyhow!(e))?,
                         },
-                        use_cloud,
                     )
                     .await?;
 

crates/web_search_providers/Cargo.toml

@@ -14,7 +14,6 @@ path = "src/web_search_providers.rs"
 [dependencies]
 anyhow.workspace = true
 client.workspace = true
-feature_flags.workspace = true
 futures.workspace = true
 gpui.workspace = true
 http_client.workspace = true

crates/web_search_providers/src/cloud.rs

@@ -2,7 +2,6 @@ use std::sync::Arc;
 
 use anyhow::{Context as _, Result};
 use client::Client;
-use feature_flags::{FeatureFlagAppExt as _, ZedCloudFeatureFlag};
 use futures::AsyncReadExt as _;
 use gpui::{App, AppContext, Context, Entity, Subscription, Task};
 use http_client::{HttpClient, Method};
@@ -63,10 +62,7 @@ impl WebSearchProvider for CloudWebSearchProvider {
         let client = state.client.clone();
         let llm_api_token = state.llm_api_token.clone();
         let body = WebSearchBody { query };
-        let use_cloud = cx.has_flag::<ZedCloudFeatureFlag>();
+        cx.background_spawn(async move { perform_web_search(client, llm_api_token, body).await })
-        cx.background_spawn(async move {
-            perform_web_search(client, llm_api_token, body, use_cloud).await
-        })
     }
 }
 
@@ -74,7 +70,6 @@ async fn perform_web_search(
     client: Arc<Client>,
     llm_api_token: LlmApiToken,
     body: WebSearchBody,
-    use_cloud: bool,
 ) -> Result<WebSearchResponse> {
     const MAX_RETRIES: usize = 3;
 
@@ -91,11 +86,7 @@ async fn perform_web_search(
 
         let request = http_client::Request::builder()
             .method(Method::POST)
-            .uri(
+            .uri(http_client.build_zed_llm_url("/web_search", &[])?.as_ref())
-                http_client
-                    .build_zed_llm_url("/web_search", &[], use_cloud)?
-                    .as_ref(),
-            )
             .header("Content-Type", "application/json")
             .header("Authorization", format!("Bearer {token}"))
             .body(serde_json::to_string(&body)?.into())?;

crates/zeta/src/zeta.rs

@@ -8,7 +8,6 @@ mod rate_completion_modal;
 
 pub(crate) use completion_diff_element::*;
 use db::kvp::KEY_VALUE_STORE;
-use feature_flags::{FeatureFlagAppExt as _, ZedCloudFeatureFlag};
 pub use init::*;
 use inline_completion::DataCollectionState;
 use license_detection::LICENSE_FILES_TO_CHECK;
@@ -391,7 +390,6 @@ impl Zeta {
         let client = self.client.clone();
         let llm_token = self.llm_token.clone();
         let app_version = AppVersion::global(cx);
-        let use_cloud = cx.has_flag::<ZedCloudFeatureFlag>();
 
         let buffer = buffer.clone();
 
@@ -482,7 +480,6 @@ impl Zeta {
                 llm_token,
                 app_version,
                 body,
-                use_cloud,
             })
             .await;
             let (response, usage) = match response {
@@ -748,7 +745,6 @@ and then another
                 llm_token,
                 app_version,
                 body,
-                use_cloud,
                 ..
             } = params;
 
@@ -764,7 +760,7 @@ and then another
                     } else {
                         request_builder.uri(
                             http_client
-                                .build_zed_llm_url("/predict_edits/v2", &[], use_cloud)?
+                                .build_zed_llm_url("/predict_edits/v2", &[])?
                                 .as_ref(),
                         )
                     };
@@ -824,7 +820,6 @@ and then another
         let client = self.client.clone();
         let llm_token = self.llm_token.clone();
         let app_version = AppVersion::global(cx);
-        let use_cloud = cx.has_flag::<ZedCloudFeatureFlag>();
         cx.spawn(async move |this, cx| {
             let http_client = client.http_client();
             let mut response = llm_token_retry(&llm_token, &client, |token| {
@@ -835,7 +830,7 @@ and then another
                     } else {
                         request_builder.uri(
                             http_client
-                                .build_zed_llm_url("/predict_edits/accept", &[], use_cloud)?
+                                .build_zed_llm_url("/predict_edits/accept", &[])?
                                 .as_ref(),
                         )
                     };
@@ -1131,7 +1126,6 @@ struct PerformPredictEditsParams {
     pub llm_token: LlmApiToken,
     pub app_version: SemanticVersion,
     pub body: PredictEditsBody,
-    pub use_cloud: bool,
 }
 
 #[derive(Error, Debug)]