From d582a8ed1bc6e6a933fd560b46b50964c79945bb Mon Sep 17 00:00:00 2001
From: Gahow Wang <gahow.wang@gmail.com>
Date: Thu, 9 Apr 2026 22:50:23 +0800
Subject: [PATCH] Validate served model name consistency

---
 ..._qwen235b_decode_thinking_run2_tpot40.json |  4 ++--
 src/aituner/spec.py                           | 10 +++++++++-
 tests/test_core_flow.py                       | 19 +++++++++++++++++++
 3 files changed, 30 insertions(+), 3 deletions(-)

diff --git a/configs/examples/dash0_qwen235b_decode_thinking_run2_tpot40.json b/configs/examples/dash0_qwen235b_decode_thinking_run2_tpot40.json
index 32619e2..7b92e04 100644
--- a/configs/examples/dash0_qwen235b_decode_thinking_run2_tpot40.json
+++ b/configs/examples/dash0_qwen235b_decode_thinking_run2_tpot40.json
@@ -9,7 +9,7 @@
   },
   "model": {
     "model_id": "qwen3-235b-a22b-256k-0717-internal",
-    "served_model_name": "qwen3-235b-decode-aituner"
+    "served_model_name": "qwen3-235b-decode"
   },
   "engine": {
     "engine_name": "vllm",
@@ -107,7 +107,7 @@
     "base_flags": {
       "host": "127.0.0.1",
       "port": 18120,
-      "served-model-name": "qwen3-235b-decode-aituner",
+      "served-model-name": "qwen3-235b-decode",
       "gpu-memory-utilization": 0.75,
       "max-model-len": 262144,
       "enable-chunked-prefill": true,
diff --git a/src/aituner/spec.py b/src/aituner/spec.py
index d099f01..7648a8c 100644
--- a/src/aituner/spec.py
+++ b/src/aituner/spec.py
@@ -576,7 +576,7 @@ class StudySpec:
 
     @classmethod
     def from_dict(cls, data: Mapping[str, Any]) -> "StudySpec":
-        return cls(
+        study = cls(
             study_id=_require_str(data.get("study_id"), context="study_id"),
             hardware=HardwareSpec.from_dict(
                 _require_mapping(data.get("hardware"), context="hardware")
@@ -595,6 +595,14 @@ class StudySpec:
             if data.get("capability_profile_path")
             else None,
         )
+        served_model_name = str(study.engine.base_flags.get("served-model-name") or "").strip()
+        if served_model_name and served_model_name != study.model.served_model_name:
+            raise SpecError(
+                "model.served_model_name must match engine.base_flags['served-model-name'] "
+                f"when both are set. Got model.served_model_name={study.model.served_model_name!r} "
+                f"and engine.base_flags['served-model-name']={served_model_name!r}."
+            )
+        return study
 
 
 @dataclass(frozen=True)
diff --git a/tests/test_core_flow.py b/tests/test_core_flow.py
index 532d581..d6fe7fe 100644
--- a/tests/test_core_flow.py
+++ b/tests/test_core_flow.py
@@ -256,6 +256,25 @@ class CoreFlowTests(unittest.TestCase):
             self.assertIn("There is no TTFT SLO for this study.", prompt)
             self.assertIn("decode-only", prompt)
 
+    def test_load_study_spec_rejects_mismatched_served_model_name(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            tmp_path = Path(tmp)
+            study_path = _write_study_assets(
+                tmp_path,
+                engine_overrides={
+                    "base_flags": {
+                        "host": "127.0.0.1",
+                        "port": 8000,
+                        "served-model-name": "engine-name",
+                    }
+                },
+            )
+            payload = json.loads(study_path.read_text(encoding="utf-8"))
+            payload["model"]["served_model_name"] = "trace-name"
+            study_path.write_text(json.dumps(payload), encoding="utf-8")
+            with self.assertRaisesRegex(SpecError, "must match engine.base_flags"):
+                load_study_spec(study_path)
+
     def test_bailian_endpoint_defaults(self) -> None:
         endpoint = LLMEndpointSpec.from_dict({"provider": "bailian", "model": "qwen-plus"})
         self.assertEqual(endpoint.provider, "bailian")