feat: expected output tokens plumbing (#5181)

Signed-off-by: PeaBrane <yanrpei@gmail.com>

lib/llm/src/preprocessor.rs

@@ -246,6 +246,7 @@ impl OpenAIPreprocessor {
                 decode_worker_id: nvext.decode_worker_id,
                 dp_rank: None, // dp_rank is set later in the pipeline
                 enable_local_updates: nvext.enable_local_updates,
+                expected_output_tokens: nvext.expected_output_tokens,
             };
             builder.routing(Some(routing));
         }

lib/llm/src/protocols/common/preprocessor.rs

@@ -42,6 +42,11 @@ pub struct RoutingHints {
     /// - `Some(false)`: External caller (e.g., GAIE sidecar) handles bookkeeping via C FFI
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub enable_local_updates: Option<bool>,
+
+    /// Expected number of output tokens for this request.
+    /// Used as a hint for routing decisions to estimate resource requirements.
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub expected_output_tokens: Option<u32>,
 }
 
 #[derive(Serialize, Deserialize, Debug, Clone, Default)]

lib/llm/src/protocols/openai/nvext.rs

@@ -116,6 +116,12 @@ pub struct NvExt {
     #[builder(default, setter(strip_option))]
     #[serde(default, skip_serializing_if = "Option::is_none")]
     pub enable_local_updates: Option<bool>,
+
+    /// Expected number of output tokens for this request.
+    /// Used as a hint for routing decisions to estimate resource requirements.
+    #[builder(default, setter(strip_option))]
+    #[serde(default, skip_serializing_if = "Option::is_none")]
+    pub expected_output_tokens: Option<u32>,
 }
 
 impl Default for NvExt {
@@ -165,6 +171,7 @@ mod tests {
         assert_eq!(nv_ext.prefill_worker_id, None);
         assert_eq!(nv_ext.decode_worker_id, None);
         assert_eq!(nv_ext.enable_local_updates, None);
+        assert_eq!(nv_ext.expected_output_tokens, None);
     }
 
     // Test valid builder configurations