[Rust Frontend] Make unhandled vllm serve frontend flags explicit

rust/src/cmd/src/cli/tests.rs

@@ -1507,6 +1507,79 @@ fn frontend_args_json_aggregates_multiple_unsupported_fields() {
     "#]].assert_eq(&actual);
 }
 
+/// Python args that have no meaningful Rust equivalent are accepted on both the
+/// `vllm-rs serve` and the `--args-json` paths.
+#[test]
+fn noop_args_are_accepted() {
+    Cli::try_parse_from([
+        "vllm-rs",
+        "serve",
+        "Qwen/Qwen3-0.6B",
+        "--disable-fastapi-docs",
+    ])
+    .expect("--disable-fastapi-docs is accepted");
+
+    Cli::try_parse_from([
+        "vllm-rs",
+        "frontend",
+        "--listen-fd",
+        "3",
+        "--input-listener-fd",
+        "999999",
+        "--output-listener-fd",
+        "999998",
+        "--args-json",
+        r#"{"model_tag":"Qwen/Qwen3-0.6B","disable_fastapi_docs":true}"#,
+    ])
+    .expect("disable_fastapi_docs is accepted");
+}
+
+/// Frontend-read Python args that Rust recognizes but does not support yet must
+/// fail startup on both the `vllm-rs serve` and the `--args-json` paths.
+#[test]
+fn unsupported_args_fail_startup() {
+    let cases = [
+        (
+            "--custom-histogram-buckets",
+            r#"{"request_latency":[0.1,1.0]}"#,
+            "custom_histogram_buckets",
+        ),
+        (
+            "--enable-tower-connector-lora",
+            "true",
+            "enable_tower_connector_lora",
+        ),
+        ("--enable-elastic-ep", "true", "enable_elastic_ep"),
+    ];
+    for (flag, value, key) in cases {
+        let error = Cli::try_parse_from(["vllm-rs", "serve", "Qwen/Qwen3-0.6B", flag, value])
+            .unwrap_err()
+            .to_string();
+        assert!(
+            error.contains("argument is not implemented in Rust frontend yet"),
+            "{flag}: {error}"
+        );
+
+        let args_json =
+            serde_json::json!({ "model_tag": "Qwen/Qwen3-0.6B", key: value }).to_string();
+        let error = Cli::try_parse_from([
+            "vllm-rs",
+            "frontend",
+            "--listen-fd",
+            "3",
+            "--input-listener-fd",
+            "999999",
+            "--output-listener-fd",
+            "999998",
+            "--args-json",
+            &args_json,
+        ])
+        .unwrap_err()
+        .to_string();
+        assert!(error.contains(&format!("- {key}\n")), "{key}: {error}");
+    }
+}
+
 #[test]
 fn frontend_args_json_rejects_malformed_json() {
     let error = Cli::try_parse_from([

rust/src/cmd/src/cli/unsupported.rs

@@ -298,6 +298,15 @@ pub struct EngineUnsupportedArgs {
     )]
     pub enable_fault_tolerance: Option<Unsupported>,
 
+    /// Enable elastic expert parallelism with stateless NCCL groups for DP/EP.
+    #[arg(
+        long,
+        visible_alias = "no-enable-elastic-ep",
+        default_missing_value = "true",
+        num_args = 0..=1
+    )]
+    pub enable_elastic_ep: Option<Unsupported>,
+
     /// This feature is work in progress and no prefill optimization takes place
     /// with this flag enabled currently.
     #[arg(
@@ -308,6 +317,17 @@ pub struct EngineUnsupportedArgs {
     )]
     pub kv_sharing_fast_prefill: Option<Unsupported>,
 
+    /// If `True`, LoRA support for the tower (vision encoder) and connector
+    /// of multimodal models will be enabled. This is an experimental feature
+    /// and currently only supports some MM models such as the Qwen VL series.
+    #[arg(
+        long,
+        visible_alias = "no-enable-tower-connector-lora",
+        default_missing_value = "true",
+        num_args = 0..=1
+    )]
+    pub enable_tower_connector_lora: Option<Unsupported>,
+
     /// Additional args passed to process media inputs, keyed by modalities.
     #[arg(long)]
     pub media_io_kwargs: Option<Unsupported>,
@@ -359,6 +379,14 @@ pub struct EngineUnsupportedArgs {
     #[arg(long)]
     pub collect_detailed_traces: Option<Unsupported>,
 
+    /// Custom Prometheus histogram bucket boundaries, as a JSON mapping from
+    /// bucket-family key to a list of strictly increasing, positive, finite
+    /// upper bounds. When a family key is present, its list replaces the
+    /// default buckets for every histogram in that family; families not listed
+    /// keep their defaults.
+    #[arg(long)]
+    pub custom_histogram_buckets: Option<Unsupported>,
+
     /// Maximum number of requests that can be in-flight (waiting or running)
     /// at the same time. When the limit is reached, new requests are rejected
     /// with HTTP 503 so the client can retry on another instance.
@@ -601,13 +629,17 @@ pub struct ServerUnsupportedArgs {
     pub middleware: Option<Unsupported>,
 
     /// Disable FastAPI's OpenAPI schema, Swagger UI, and ReDoc endpoint.
+    ///
+    /// Accepted as a no-op: the Rust frontend does not serve these docs
+    /// endpoints at all.
     #[arg(
         long,
         visible_alias = "no-disable-fastapi-docs",
         default_missing_value = "true",
-        num_args = 0..=1
+        num_args = 0..=1,
+        hide = true
     )]
-    pub disable_fastapi_docs: Option<Unsupported>,
+    pub disable_fastapi_docs: Option<Noop>,
 
     /// Maximum size (bytes) of an incomplete HTTP event (header or body) for
     /// h11 parser. Helps mitigate header abuse. Default: 4194304 (4 MB).