[PD] Introduce runtime role switching between prefill and decode (#28403)

Signed-off-by: huanglong <huanglong@linux.alibaba.com>
Signed-off-by: inkcherry <mingzhi.liu@amd.com>
Co-authored-by: huanglong <huanglong@linux.alibaba.com>
Co-authored-by: Shangming Cai <csmthu@gmail.com>
Co-authored-by: Huang Long <121648372+LLLL114@users.noreply.github.com>

rust/sglang-server/src/api_server/common.rs

@@ -94,6 +94,7 @@ async fn model_info(State(state): State<Arc<AppState>>) -> Response {
         // selected parser into `server_args` before the scheduler forks.
         "reasoning_parser": sa.reasoning_parser,
         "tool_call_parser": sa.tool_call_parser,
+        "disaggregation_mode": sa.disaggregation_mode,
     });
     (
         StatusCode::OK,

rust/sglang-server/src/message/config.rs

@@ -317,7 +317,9 @@ impl<'py> pyo3::FromPyObject<'_, 'py> for PreferredSamplingParams {
     from_py_object,
     module = "sglang.srt.rust_extensions._server"
 )]
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+// Lowercase to match the values Python reports for the same field.
+#[derive(Clone, Copy, Debug, PartialEq, Eq, Serialize)]
+#[serde(rename_all = "lowercase")]
 pub enum DisaggregationMode {
     /// Unified prefill + decode.
     Null,
@@ -619,6 +621,14 @@ mod tests {
         assert_eq!(ServerArgs::default().bind(), "127.0.0.1:30000");
     }
 
+    #[test]
+    fn disaggregation_mode_wire_values_match_python() {
+        let json = |m| serde_json::to_string(&m).unwrap();
+        assert_eq!(json(DisaggregationMode::Null), "\"null\"");
+        assert_eq!(json(DisaggregationMode::Prefill), "\"prefill\"");
+        assert_eq!(json(DisaggregationMode::Decode), "\"decode\"");
+    }
+
     #[test]
     fn pd_role_derivations() {
         let prefill = ServerArgs {