Upload 2 files

schemas.py

@@ -14,6 +14,8 @@ class HealthResponse(BaseModel):
     ready: bool
     max_context_length: int
     max_horizon_step: int
+    patch_length: int
+    runtime_revision: str
 
 
 class PredictRequest(BaseModel):

service.py

@@ -22,8 +22,9 @@ class TimesFmService:
             "TIMESFM_MODEL_NAME",
             "google/timesfm-2.5-200m-transformers",
         )
-        self.backend = os.getenv("TIMESFM_BACKEND", "hf_cpu").strip() or "hf_cpu"
-        self.device = "cpu"
+        self.backend = os.getenv("TIMESFM_BACKEND", "hf_cpu").strip() or "hf_cpu"
+        self.device = "cpu"
+        self.runtime_revision = os.getenv("TIMESFM_RUNTIME_REVISION", "timesfm-hf-patch-align-v1")
         self.max_context_length = int(os.getenv("TIMESFM_MAX_CONTEXT_LENGTH", "512"))
         self.max_horizon_step = int(os.getenv("TIMESFM_MAX_HORIZON_STEP", "288"))
         self.patch_length = int(os.getenv("TIMESFM_PATCH_LENGTH", "32"))
@@ -46,10 +47,12 @@ class TimesFmService:
             model_id=self.model_id,
             backend=self.backend,
             device=self.device,
-            ready=self.ready,
-            max_context_length=self.max_context_length,
-            max_horizon_step=self.max_horizon_step,
-        )
+            ready=self.ready,
+            max_context_length=self.max_context_length,
+            max_horizon_step=self.max_horizon_step,
+            patch_length=self.patch_length,
+            runtime_revision=self.runtime_revision,
+        )
 
     def predict(self, payload: PredictRequest) -> PredictResponse:
         self._validate_request(payload)