[Rust Frontend] Report vllm-proto schema version in ServerInfo (#61533)

Signed-off-by: Alec Flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Bugen Zhao <i@bugenzhao.com>

rust/Cargo.lock

@@ -5947,7 +5947,7 @@ dependencies = [
 
 [[package]]
 name = "vllm-proto"
-version = "0.5.0"
+version = "0.6.0"
 dependencies = [
  "prost",
  "prost-types",

rust/proto/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "vllm-proto"
-version = "0.5.0"
+version = "0.6.0"
 edition.workspace = true
 license.workspace = true
 description = "Protobuf messages and gRPC bindings for the vLLM Rust frontend"

rust/proto/README.md

@@ -16,7 +16,7 @@ the schemas on pull requests, including on forks.
 
 ```toml
 [dependencies]
-vllm-proto = "0.5"
+vllm-proto = "0.6"
 ```
 
 For example, import `vllm_proto::inference_client::InferenceClient` or

rust/proto/control.proto

@@ -54,6 +54,7 @@ message ServerInfo {
   uint64 max_batched_tokens = 9;
   uint32 max_loras = 10;
   RlCapabilities rl_capabilities = 11;
+  string proto_version = 12;
   // Full-attention block size in tokens after initialization; absent if unavailable.
   optional uint64 effective_attention_block_size = 14;
 }

rust/proto/src/lib.rs

@@ -4,3 +4,6 @@
 //! Protobuf messages and gRPC clients and servers for vLLM's Inference and Control APIs.
 
 tonic::include_proto!("vllm");
+
+/// Schema version of this crate, reported by servers as `ServerInfo.proto_version`.
+pub const VERSION: &str = env!("CARGO_PKG_VERSION");

rust/src/server/src/grpc/control.rs

@@ -216,6 +216,7 @@ impl pb::control_server::Control for ControlServiceImpl {
             max_batched_tokens: ready.max_num_batched_tokens,
             max_loras: ready.max_loras,
             rl_capabilities: Some(self.rl_capabilities()),
+            proto_version: pb::VERSION.to_string(),
             effective_attention_block_size: self.client().effective_attention_block_size(),
         }))
     }

rust/src/server/src/grpc/tests.rs

@@ -2013,6 +2013,7 @@ async fn control_reports_server_and_model_info() {
     assert_eq!(server.engine_version, "test-vllm-version");
     assert_eq!(server.effective_attention_block_size, Some(64));
     assert_eq!(server.api_version, "vllm");
+    assert_eq!(server.proto_version, pb::VERSION);
     assert_eq!(server.instance_id, "test-instance");
     assert_eq!(server.max_model_len, DEFAULT_MOCK_MAX_MODEL_LEN as u32);
     assert_eq!(server.kv_block_size, DEFAULT_MOCK_BLOCK_SIZE as u32);