language_models: Pass up mode for completion requests through Zed (#29294)

This PR makes it so we pass up the `mode` for completion requests
through the Zed provider.

Release Notes:

- N/A

Cargo.lock

@@ -18400,9 +18400,9 @@ dependencies = [
 
 [[package]]
 name = "zed_llm_client"
-version = "0.6.1"
+version = "0.7.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "ad17428120f5ca776dc5195e2411a282f5150a26d5536671f8943c622c31274f"
+checksum = "3c1666cd923c5eb4635f3743e69c6920d0ed71f29b26920616a5d220607df7c4"
 dependencies = [
  "anyhow",
  "serde",

Cargo.toml

@@ -605,7 +605,7 @@ wasmtime-wasi = "29"
 which = "6.0.0"
 wit-component = "0.221"
 workspace-hack = "0.1.0"
-zed_llm_client = "0.6.1"
+zed_llm_client = "0.7.0"
 zstd = "0.11"
 metal = "0.29"
 

crates/language_models/src/provider/cloud.rs

@@ -35,7 +35,7 @@ use strum::IntoEnumIterator;
 use thiserror::Error;
 use ui::{TintColor, prelude::*};
 use zed_llm_client::{
-    CURRENT_PLAN_HEADER_NAME, CompletionBody, EXPIRED_LLM_TOKEN_HEADER_NAME,
+    CURRENT_PLAN_HEADER_NAME, CompletionBody, CompletionMode, EXPIRED_LLM_TOKEN_HEADER_NAME,
     MAX_LLM_MONTHLY_SPEND_REACHED_HEADER_NAME, MODEL_REQUESTS_RESOURCE_HEADER_VALUE,
     SUBSCRIPTION_LIMIT_RESOURCE_HEADER_NAME,
 };
@@ -748,6 +748,7 @@ impl LanguageModel for CloudLanguageModel {
                         CompletionBody {
                             thread_id,
                             prompt_id,
+                            mode: Some(CompletionMode::Max),
                             provider: zed_llm_client::LanguageModelProvider::Anthropic,
                             model: request.model.clone(),
                             provider_request: serde_json::to_value(&request)?,
@@ -794,6 +795,7 @@ impl LanguageModel for CloudLanguageModel {
                         CompletionBody {
                             thread_id,
                             prompt_id,
+                            mode: Some(CompletionMode::Max),
                             provider: zed_llm_client::LanguageModelProvider::OpenAi,
                             model: request.model.clone(),
                             provider_request: serde_json::to_value(&request)?,
@@ -824,6 +826,7 @@ impl LanguageModel for CloudLanguageModel {
                         CompletionBody {
                             thread_id,
                             prompt_id,
+                            mode: Some(CompletionMode::Max),
                             provider: zed_llm_client::LanguageModelProvider::Google,
                             model: request.model.clone(),
                             provider_request: serde_json::to_value(&request)?,