modal 1.5.5.dev12__tar.gz → 1.5.5.dev13__tar.gz

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.5.5.dev12
+Version: 1.5.5.dev13
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License-Expression: Apache-2.0

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/modal/client.pyi

@@ -38,7 +38,7 @@ class _Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.5.5.dev12",
+        version: str = "1.5.5.dev13",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.
@@ -209,7 +209,7 @@ class Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.5.5.dev12",
+        version: str = "1.5.5.dev13",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/modal/experimental/flash.py

@@ -25,6 +25,7 @@ from ..client import _Client
 from ..config import logger
 from ..exception import InvalidError
 
+_FLASH_UPSTREAM_HEADER = "modal-flash-upstream"
 _MAX_FAILURES = 10
 
 
@@ -337,9 +338,14 @@ class _FlashPrometheusAutoscaler:
         )
 
         self.autoscaler_thread: asyncio.Task | None = None
+        self._flash_service_url: str | None = None
 
     async def start(self):
         await self.fn.hydrate(client=self.client)
+        flash_service_urls = await self.fn._experimental_get_flash_urls()
+        if not flash_service_urls:
+            raise InvalidError(f"Could not find a Flash service URL for {self.app_name}.{self.cls_name}")
+        self._flash_service_url = flash_service_urls[0]
         self.autoscaler_thread = asyncio.create_task(self._run_autoscaler_loop())
 
     async def _run_autoscaler_loop(self):
@@ -496,13 +502,20 @@ class _FlashPrometheusAutoscaler:
         return desired_replicas
 
     async def _get_scaling_info(self, containers) -> tuple[float, int]:
-        """Get metrics using container exposed metrics endpoints."""
+        """Get metrics from each container through the Flash service."""
+        if self._flash_service_url is None:
+            raise RuntimeError("Flash Prometheus autoscaler has not been started")
+
         sum_metric = 0
         n_containers_with_metrics = 0
+        metrics_url = f"{self._flash_service_url.rstrip('/')}/{self.metrics_endpoint.lstrip('/')}"
 
         container_metrics_list = await asyncio.gather(
             *[
-                self._get_metrics(f"https://{container.host}:{container.port}/{self.metrics_endpoint}")
+                self._get_metrics(
+                    metrics_url,
+                    headers={_FLASH_UPSTREAM_HEADER: f"{container.host}:{container.port}"},
+                )
                 for container in containers
             ]
         )
@@ -519,14 +532,20 @@ class _FlashPrometheusAutoscaler:
 
         return sum_metric, n_containers_with_metrics
 
-    async def _get_metrics(self, url: str) -> dict[str, list[Any]] | None:  # technically any should be Sample
+    async def _get_metrics(
+        self, url: str, headers: dict[str, str] | None = None
+    ) -> dict[str, list[Any]] | None:  # technically any should be Sample
         import aiohttp
         from prometheus_client.parser import text_string_to_metric_families
         from prometheus_client.samples import Sample
 
         # Fetch the metrics from the endpoint
         try:
-            response = await self.http_client.get(url, timeout=aiohttp.ClientTimeout(total=3))
+            response = await self.http_client.get(
+                url,
+                headers=headers,
+                timeout=aiohttp.ClientTimeout(total=3),
+            )
             response.raise_for_status()
         except asyncio.TimeoutError:
             logger.warning(f"[Modal Flash] Timeout getting metrics from {url}")

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/modal/experimental/flash.pyi

@@ -196,10 +196,12 @@ class _FlashPrometheusAutoscaler:
         ...
 
     async def _get_scaling_info(self, containers) -> tuple[float, int]:
-        """Get metrics using container exposed metrics endpoints."""
+        """Get metrics from each container through the Flash service."""
         ...
 
-    async def _get_metrics(self, url: str) -> typing.Optional[dict[str, list[typing.Any]]]: ...
+    async def _get_metrics(
+        self, url: str, headers: typing.Optional[dict[str, str]] = None
+    ) -> typing.Optional[dict[str, list[typing.Any]]]: ...
     async def _get_all_containers(self): ...
     async def _set_target_slots(self, target_slots: int): ...
     def _make_scaling_decision(
@@ -287,18 +289,22 @@ class FlashPrometheusAutoscaler:
 
     class ___get_scaling_info_spec(typing_extensions.Protocol):
         def __call__(self, /, containers) -> tuple[float, int]:
-            """Get metrics using container exposed metrics endpoints."""
+            """Get metrics from each container through the Flash service."""
             ...
 
         async def aio(self, /, containers) -> tuple[float, int]:
-            """Get metrics using container exposed metrics endpoints."""
+            """Get metrics from each container through the Flash service."""
             ...
 
     _get_scaling_info: ___get_scaling_info_spec
 
     class ___get_metrics_spec(typing_extensions.Protocol):
-        def __call__(self, /, url: str) -> typing.Optional[dict[str, list[typing.Any]]]: ...
-        async def aio(self, /, url: str) -> typing.Optional[dict[str, list[typing.Any]]]: ...
+        def __call__(
+            self, /, url: str, headers: typing.Optional[dict[str, str]] = None
+        ) -> typing.Optional[dict[str, list[typing.Any]]]: ...
+        async def aio(
+            self, /, url: str, headers: typing.Optional[dict[str, str]] = None
+        ) -> typing.Optional[dict[str, list[typing.Any]]]: ...
 
     _get_metrics: ___get_metrics_spec
 

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/modal.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.5.5.dev12
+Version: 1.5.5.dev13
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License-Expression: Apache-2.0

{modal-1.5.5.dev12 → modal-1.5.5.dev13}/modal_version/__init__.py

@@ -1,4 +1,4 @@
 # Copyright Modal Labs 2026
 """Supplies the current version of the modal client library."""
 
-__version__ = "1.5.5.dev12"
+__version__ = "1.5.5.dev13"