modal 1.1.5.dev49__py3-none-any.whl → 1.1.5.dev52__py3-none-any.whl

modal/_output.py

@@ -35,7 +35,7 @@ from modal._utils.time_utils import timestamp_to_localized_str
 from modal_proto import api_pb2
 
 from ._utils.grpc_utils import RETRYABLE_GRPC_STATUS_CODES, retry_transient_errors
-from ._utils.shell_utils import stream_from_stdin
+from ._utils.shell_utils import stream_from_stdin, write_to_fd
 from .client import _Client
 from .config import logger
 
@@ -507,17 +507,32 @@ async def put_pty_content(log: api_pb2.TaskLogs, stdout):
         # because the progress spinner can't interfere with output.
 
         data = log.data.encode("utf-8")
-        written = 0
-        n_retries = 0
-        while written < len(data):
-            try:
-                written += stdout.buffer.write(data[written:])
-                stdout.flush()
-            except BlockingIOError:
-                if n_retries >= 5:
-                    raise
-                n_retries += 1
-                await asyncio.sleep(0.1)
+        # Non-blocking terminals can fill the kernel buffer on output bursts, making flush() raise
+        # BlockingIOError (EAGAIN) and appear frozen until a key is pressed (this happened e.g. when
+        # printing large data from a pdb breakpoint). If stdout has a real fd, we await a
+        # non-blocking fd write (write_to_fd) instead.
+        fd = None
+        try:
+            if hasattr(stdout, "fileno"):
+                fd = stdout.fileno()
+        except Exception:
+            fd = None
+
+        if fd is not None:
+            await write_to_fd(fd, data)
+        else:
+            # For streams without fileno(), use the normal write/flush path.
+            written = 0
+            n_retries = 0
+            while written < len(data):
+                try:
+                    written += stdout.buffer.write(data[written:])
+                    stdout.flush()
+                except BlockingIOError:
+                    if n_retries >= 5:
+                        raise
+                    n_retries += 1
+                    await asyncio.sleep(0.1)
     else:
         # `stdout` isn't always buffered (e.g. %%capture in Jupyter notebooks redirects it to
         # io.StringIO).

modal/_runtime/gpu_memory_snapshot.py

@@ -18,6 +18,12 @@ from modal.config import config, logger
 
 CUDA_CHECKPOINT_PATH: str = config.get("cuda_checkpoint_path")
 
+# Maximum total duration for an entire toggle operation.
+CUDA_CHECKPOINT_TOGGLE_TIMEOUT: float = 5 * 60.0
+
+# Maximum total duration for each individual `cuda-checkpoint` invocation.
+CUDA_CHECKPOINT_TIMEOUT: float = 90
+
 
 class CudaCheckpointState(Enum):
     """State representation from the CUDA API [1].
@@ -44,7 +50,7 @@ class CudaCheckpointProcess:
     pid: int
     state: CudaCheckpointState
 
-    def toggle(self, target_state: CudaCheckpointState, timeout_secs: float = 5 * 60.0) -> None:
+    def toggle(self, target_state: CudaCheckpointState, skip_first_refresh: bool = False) -> None:
         """Toggle CUDA checkpoint state for current process, moving GPU memory to the
         CPU and back depending on the current process state when called.
         """
@@ -54,7 +60,11 @@ class CudaCheckpointProcess:
         retry_count = 0
         max_retries = 3
 
-        while self._should_continue_toggle(target_state, start_time, timeout_secs):
+        attempts = 0
+        while self._should_continue_toggle(
+            target_state, start_time, refresh=not (skip_first_refresh and attempts == 0)
+        ):
+            attempts += 1
             try:
                 self._execute_toggle_command()
                 # Use exponential backoff for retries
@@ -73,10 +83,11 @@ class CudaCheckpointProcess:
         logger.debug(f"PID: {self.pid} Target state {target_state.value} reached")
 
     def _should_continue_toggle(
-        self, target_state: CudaCheckpointState, start_time: float, timeout_secs: float
+        self, target_state: CudaCheckpointState, start_time: float, refresh: bool = True
     ) -> bool:
         """Check if toggle operation should continue based on current state and timeout."""
-        self.refresh_state()
+        if refresh:
+            self.refresh_state()
 
         if self.state == target_state:
             return False
@@ -85,7 +96,7 @@ class CudaCheckpointProcess:
             raise CudaCheckpointException(f"PID: {self.pid} CUDA process state is {self.state}")
 
         elapsed = time.monotonic() - start_time
-        if elapsed >= timeout_secs:
+        if elapsed >= CUDA_CHECKPOINT_TOGGLE_TIMEOUT:
             raise CudaCheckpointException(
                 f"PID: {self.pid} Timeout after {elapsed:.2f}s waiting for state {target_state.value}. "
                 f"Current state: {self.state}"
@@ -101,7 +112,7 @@ class CudaCheckpointProcess:
                 check=True,
                 capture_output=True,
                 text=True,
-                timeout=30,
+                timeout=CUDA_CHECKPOINT_TIMEOUT,
             )
             logger.debug(f"PID: {self.pid} Successfully toggled CUDA checkpoint state")
         except subprocess.CalledProcessError as e:
@@ -121,7 +132,7 @@ class CudaCheckpointProcess:
                 check=True,
                 capture_output=True,
                 text=True,
-                timeout=10,
+                timeout=CUDA_CHECKPOINT_TIMEOUT,
             )
 
             state_str = result.stdout.strip().lower()
@@ -190,6 +201,7 @@ class CudaCheckpointSession:
                 [CUDA_CHECKPOINT_PATH, "--get-state", "--pid", str(pid)],
                 capture_output=True,
                 text=True,
+                # This should be quick since no checkpoint has taken place yet
                 timeout=5,
             )
 
@@ -256,20 +268,11 @@ class CudaCheckpointSession:
             logger.debug("No CUDA sessions to restore.")
             return
 
-        # Validate all states first
-        for proc in self.cuda_processes:
-            proc.refresh_state()  # Refresh state before validation
-            if proc.state != CudaCheckpointState.CHECKPOINTED:
-                raise CudaCheckpointException(
-                    f"PID {proc.pid}: CUDA session not in {CudaCheckpointState.CHECKPOINTED.value} state. "
-                    f"Current state: {proc.state.value}"
-                )
-
         # See checkpoint() for rationale about parallelism.
         start = time.perf_counter()
 
         def restore_process(proc: CudaCheckpointProcess) -> None:
-            proc.toggle(CudaCheckpointState.RUNNING)
+            proc.toggle(CudaCheckpointState.RUNNING, skip_first_refresh=True)
 
         with ThreadPoolExecutor() as executor:
             futures = [executor.submit(restore_process, proc) for proc in self.cuda_processes]

modal/client.pyi

@@ -33,7 +33,7 @@ class _Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.5.dev49",
+        version: str = "1.1.5.dev52",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.
@@ -164,7 +164,7 @@ class Client:
         server_url: str,
         client_type: int,
         credentials: typing.Optional[tuple[str, str]],
-        version: str = "1.1.5.dev49",
+        version: str = "1.1.5.dev52",
     ):
         """mdmd:hidden
         The Modal client object is not intended to be instantiated directly by users.

modal/functions.pyi

@@ -450,7 +450,7 @@ class Function(
 
     _call_generator: ___call_generator_spec[typing_extensions.Self]
 
-    class __remote_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
+    class __remote_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> ReturnType_INNER:
             """Calls the function remotely, executing it with the given arguments and returning the execution's result."""
             ...
@@ -459,7 +459,7 @@ class Function(
             """Calls the function remotely, executing it with the given arguments and returning the execution's result."""
             ...
 
-    remote: __remote_spec[modal._functions.P, modal._functions.ReturnType, typing_extensions.Self]
+    remote: __remote_spec[modal._functions.ReturnType, modal._functions.P, typing_extensions.Self]
 
     class __remote_gen_spec(typing_extensions.Protocol[SUPERSELF]):
         def __call__(self, /, *args, **kwargs) -> typing.Generator[typing.Any, None, None]:
@@ -486,7 +486,7 @@ class Function(
         """
         ...
 
-    class ___experimental_spawn_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
+    class ___experimental_spawn_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> FunctionCall[ReturnType_INNER]:
             """[Experimental] Calls the function with the given arguments, without waiting for the results.
 
@@ -510,7 +510,7 @@ class Function(
             ...
 
     _experimental_spawn: ___experimental_spawn_spec[
-        modal._functions.P, modal._functions.ReturnType, typing_extensions.Self
+        modal._functions.ReturnType, modal._functions.P, typing_extensions.Self
     ]
 
     class ___spawn_map_inner_spec(typing_extensions.Protocol[P_INNER, SUPERSELF]):
@@ -519,7 +519,7 @@ class Function(
 
     _spawn_map_inner: ___spawn_map_inner_spec[modal._functions.P, typing_extensions.Self]
 
-    class __spawn_spec(typing_extensions.Protocol[P_INNER, ReturnType_INNER, SUPERSELF]):
+    class __spawn_spec(typing_extensions.Protocol[ReturnType_INNER, P_INNER, SUPERSELF]):
         def __call__(self, /, *args: P_INNER.args, **kwargs: P_INNER.kwargs) -> FunctionCall[ReturnType_INNER]:
             """Calls the function with the given arguments, without waiting for the results.
 
@@ -540,7 +540,7 @@ class Function(
             """
             ...
 
-    spawn: __spawn_spec[modal._functions.P, modal._functions.ReturnType, typing_extensions.Self]
+    spawn: __spawn_spec[modal._functions.ReturnType, modal._functions.P, typing_extensions.Self]
 
     def get_raw_f(self) -> collections.abc.Callable[..., typing.Any]:
         """Return the inner Python object wrapped by this Modal Function."""

{modal-1.1.5.dev49.dist-info → modal-1.1.5.dev52.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: modal
-Version: 1.1.5.dev49
+Version: 1.1.5.dev52
 Summary: Python client library for Modal
 Author-email: Modal Labs <support@modal.com>
 License: Apache-2.0

{modal-1.1.5.dev49.dist-info → modal-1.1.5.dev52.dist-info}/RECORD

@@ -7,7 +7,7 @@ modal/_functions.py,sha256=6e4rFdl8thTnRuhUOj_4ehOzC1wdvJHhYSoIwB0LXhU,91783
 modal/_ipython.py,sha256=TW1fkVOmZL3YYqdS2YlM1hqpf654Yf8ZyybHdBnlhSw,301
 modal/_location.py,sha256=joiX-0ZeutEUDTrrqLF1GHXCdVLF-rHzstocbMcd_-k,366
 modal/_object.py,sha256=gwsLdXb-Ecd8nH8LVCo8oVZPzzdyo9BrN1DjgQmsSuM,11967
-modal/_output.py,sha256=G9CeSQEBzjhveWWEzWmYa5Uwbu4lZf8N8IFH1UM4fU0,25803
+modal/_output.py,sha256=h7wyYPtHdWFgtiWg-5obZI07RaNwivoEiLCUPl8ZOKc,26522
 modal/_partial_function.py,sha256=Yqk97hLS6vi8nWWVpzS5TSWbndWMdCtkhccdnyDJgBk,37302
 modal/_pty.py,sha256=E58MQ8d5-wkbMatRKpQR-G9FdbCRcZGiZxOpGy__VuY,1481
 modal/_resolver.py,sha256=2RWvm34cNSnbv1v7izJMNZgfvpLDD6LzaBlr0lIrLnY,7364
@@ -22,7 +22,7 @@ modal/app.py,sha256=RRUz2NjAWIQLHtU2IEslOlnIOCxPiWts3IP3rTFArkY,49635
 modal/app.pyi,sha256=CDp_rlX3hBuFdv9VRsKvNKCgu_hS2IO2uNU5qhzmXps,44719
 modal/call_graph.py,sha256=1g2DGcMIJvRy-xKicuf63IVE98gJSnQsr8R_NVMptNc,2581
 modal/client.py,sha256=kyAIVB3Ay-XKJizQ_1ufUFB__EagV0MLmHJpyYyJ7J0,18636
-modal/client.pyi,sha256=K_pIkzUJpPR7aFiPQPfdIs8spI7by2FSpqjWbpHqmRg,15831
+modal/client.pyi,sha256=88q9g9WIFGDAWQpiFIyjAeyexpoG5-3npFQ1LHKS5us,15831
 modal/cloud_bucket_mount.py,sha256=I2GRXYhOWLIz2kJZjXu75jAm9EJkBNcutGc6jR2ReUw,5928
 modal/cloud_bucket_mount.pyi,sha256=VuUOipMIHqFXMkD-3g2bsoqpSxV5qswlFHDOqPQzYAo,7405
 modal/cls.py,sha256=R1uLQbdqWRRjvxs0I57a4hZZELZkBVCxOKxvKryU5_s,41639
@@ -39,7 +39,7 @@ modal/file_io.py,sha256=OSKr77TujcXGJW1iikzYiHckLSmv07QBgBHcxxYEkoI,21456
 modal/file_io.pyi,sha256=xtO6Glf_BFwDE7QiQQo24QqcMf_Vv-iz7WojcGVlLBU,15932
 modal/file_pattern_matcher.py,sha256=A_Kdkej6q7YQyhM_2-BvpFmPqJ0oHb54B6yf9VqvPVE,8116
 modal/functions.py,sha256=kcNHvqeGBxPI7Cgd57NIBBghkfbeFJzXO44WW0jSmao,325
-modal/functions.pyi,sha256=FU1F_75_-Y-YEWo6UNxvmuQ8P3T8TNgpqEqhgED4dhc,39597
+modal/functions.pyi,sha256=9hHRHIkzRtaUWtAHZEmGSUmufyPfgB1JK0kgBipfbys,39597
 modal/gpu.py,sha256=Fe5ORvVPDIstSq1xjmM6OoNgLYFWvogP9r5BgmD3hYg,6769
 modal/image.py,sha256=pCiIeDt-YDpzBZ7_uqPcuizRniZYG34Z_NDMCsIIjas,108084
 modal/image.pyi,sha256=ZNp48mVPzcQ6XNvxin1iO5XrZ89vfEZvU1Bi-V57jq0,76835
@@ -86,7 +86,7 @@ modal/_runtime/container_io_manager.py,sha256=HZJsAC7Vn1a3EXpyJAMuOibsFulHpWlqTE
 modal/_runtime/container_io_manager.pyi,sha256=GDNLirCcPMRc6gckInYKmGmJZY3LNcgUVXKftt9P9jI,23493
 modal/_runtime/execution_context.py,sha256=AYrNQRHHXEqX2MwMf8zxelKZnYf25RE_B-NRLWf93n8,3521
 modal/_runtime/execution_context.pyi,sha256=FVzakehz72ndL-ufe8-EC7TM4IHO_MEBcAdgWuU4W9k,2426
-modal/_runtime/gpu_memory_snapshot.py,sha256=BWIMKkH-UXTQOJJuXbM15UWCHHSYlJ0XxGlZunKb0Ug,11877
+modal/_runtime/gpu_memory_snapshot.py,sha256=8T85kyfGJ11AnhFRCZTeWj9zin9e9f5XbrRqU8_WjAk,11955
 modal/_runtime/telemetry.py,sha256=T1RoAGyjBDr1swiM6pPsGRSITm7LI5FDK18oNXxY08U,5163
 modal/_runtime/user_code_imports.py,sha256=1MlOgw810aj0MeDvFPvHBIz-aHd7jUX6dwRfIcM3-KE,16498
 modal/_utils/__init__.py,sha256=waLjl5c6IPDhSsdWAm9Bji4e2PVxamYABKAze6CHVXY,28
@@ -153,7 +153,7 @@ modal/experimental/__init__.py,sha256=fCqzo_f3vcY750vHtd7CtLs5dvdM_C0ZLLGb3zXuK9
 modal/experimental/flash.py,sha256=7qRAL2Nrwbb60YKobcnpM0zJ8vw4xGJqabLPFgEzMZE,28295
 modal/experimental/flash.pyi,sha256=R9VV0UDotiY9BRUjacB-xI4qhR3yBymAvEZFRFHztLs,15143
 modal/experimental/ipython.py,sha256=TrCfmol9LGsRZMeDoeMPx3Hv3BFqQhYnmD_iH0pqdhk,2904
-modal-1.1.5.dev49.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
+modal-1.1.5.dev52.dist-info/licenses/LICENSE,sha256=psuoW8kuDP96RQsdhzwOqi6fyWv0ct8CR6Jr7He_P_k,10173
 modal_docs/__init__.py,sha256=svYKtV8HDwDCN86zbdWqyq5T8sMdGDj0PVlzc2tIxDM,28
 modal_docs/gen_cli_docs.py,sha256=c1yfBS_x--gL5bs0N4ihMwqwX8l3IBWSkBAKNNIi6bQ,3801
 modal_docs/gen_reference_docs.py,sha256=d_CQUGQ0rfw28u75I2mov9AlS773z9rG40-yq5o7g2U,6359
@@ -161,13 +161,13 @@ modal_docs/mdmd/__init__.py,sha256=svYKtV8HDwDCN86zbdWqyq5T8sMdGDj0PVlzc2tIxDM,2
 modal_docs/mdmd/mdmd.py,sha256=tUTImNd4UMFk1opkaw8J672gX8AkBO5gbY2S_NMxsxs,7140
 modal_docs/mdmd/signatures.py,sha256=XJaZrK7Mdepk5fdX51A8uENiLFNil85Ud0d4MH8H5f0,3218
 modal_proto/__init__.py,sha256=MIEP8jhXUeGq_eCjYFcqN5b1bxBM4fdk0VESpjWR0fc,28
-modal_proto/api.proto,sha256=hN0KacVe7fpAAN0GPGVSEeWTTi4hReDfVmvegN2IsZ4,109060
-modal_proto/api_grpc.py,sha256=2KC4gGgqj7FahTrmZD5Gvxs2yV9_5l65yWjLuEd5dNQ,132154
-modal_proto/api_pb2.py,sha256=6bKElO_Pt3S6YTwGx2nxb0Cl694MxzYnPSvCTdL9csU,379863
-modal_proto/api_pb2.pyi,sha256=8I-Vo42yQjPoJolheJDk0vxrZx5uWdqdePlBYhQbHg0,525036
-modal_proto/api_pb2_grpc.py,sha256=nh3_94du1FMglF7sgVfk-s3XCl4vc07Mu-6qe6e0DdY,284763
-modal_proto/api_pb2_grpc.pyi,sha256=7X2WZeHyKJWsyEXyzoOdLlU3NT7jhWbaWDZw_-C6EeU,66726
-modal_proto/modal_api_grpc.py,sha256=UTgAj4GMOSZmwJBgrhZYnKOyWP1rQQS_HSnlfKkKT5s,19929
+modal_proto/api.proto,sha256=7VHunG16zrHWb6QZQ69SZoqMCwtbhhnE8N2KCUu9kUw,109391
+modal_proto/api_grpc.py,sha256=uA2UlaOWKR8vOq3MCRC_dbqB3vcl0jfpHtDzgDJd1uo,133054
+modal_proto/api_pb2.py,sha256=BDY2mO2b-Lb5RvP8o4tlOkJLbcIXK33IXgrThF5_hrk,381409
+modal_proto/api_pb2.pyi,sha256=-F95nswVuBe_k4FRKZ7f4RjpWs7p1Sgm_nwGrkPJE4c,526019
+modal_proto/api_pb2_grpc.py,sha256=TZRVnkM0KoXbPkV86bFGH-YVfEwMf1dXzJMfcDPURu0,286578
+modal_proto/api_pb2_grpc.pyi,sha256=KDTR99gf00Vrp2aA8FPDY3xxSqJ06HpL9uw2117PYTs,67175
+modal_proto/modal_api_grpc.py,sha256=MajAfiiJUJaZ3oRKDss4i-PDxNLDRTLkw9mptD90nxI,20063
 modal_proto/modal_options_grpc.py,sha256=qJ1cuwA54oRqrdTyPTbvfhFZYd9HhJKK5UCwt523r3Y,120
 modal_proto/options.proto,sha256=zp9h5r61ivsp0XwEWwNBsVqNTbRA1VSY_UtN7sEcHtE,549
 modal_proto/options_grpc.py,sha256=M18X3d-8F_cNYSVM3I25dUTO5rZ0rd-vCCfynfh13Nc,125
@@ -182,10 +182,10 @@ modal_proto/sandbox_router_pb2.py,sha256=INd9izYaIYqllESQt4MSv2Rj9Hf5bMjAvtCc9b4
 modal_proto/sandbox_router_pb2.pyi,sha256=YCK0WnCgRos3-p7t4USQQ7x6WAuM278yeQX2IeU5mLg,13295
 modal_proto/sandbox_router_pb2_grpc.py,sha256=zonC5flvCwxeZYJPENj1IJo2Mr0J58DpoC1_8IdPYik,8243
 modal_proto/sandbox_router_pb2_grpc.pyi,sha256=4QgCB9b7_ykvH8YD-hfnogVH9CLyHVDC5QNb03l4_X8,2735
-modal_version/__init__.py,sha256=mM4WmtQdr6bl4_q5Ng9gd1RJp3mSrMgTTZdiwzjaWU8,121
+modal_version/__init__.py,sha256=Xg8_Tklblx3LORhBdVEOf2vdRZLbs54ReBWogH_UTZY,121
 modal_version/__main__.py,sha256=2FO0yYQQwDTh6udt1h-cBnGd1c4ZyHnHSI4BksxzVac,105
-modal-1.1.5.dev49.dist-info/METADATA,sha256=HUxfOE14ZAHVPzaxsZZ49AVDi3_9iCgp6FfL1MPR_Z8,2481
-modal-1.1.5.dev49.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
-modal-1.1.5.dev49.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
-modal-1.1.5.dev49.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
-modal-1.1.5.dev49.dist-info/RECORD,,
+modal-1.1.5.dev52.dist-info/METADATA,sha256=HYMWn_-7J-j8-kK13Dp4N940_-LG1-smuLwvuNAZLJM,2481
+modal-1.1.5.dev52.dist-info/WHEEL,sha256=1tXe9gY0PYatrMPMDd6jXqjfpz_B-Wqm32CPfRC58XU,91
+modal-1.1.5.dev52.dist-info/entry_points.txt,sha256=An-wYgeEUnm6xzrAP9_NTSTSciYvvEWsMZILtYrvpAI,46
+modal-1.1.5.dev52.dist-info/top_level.txt,sha256=4BWzoKYREKUZ5iyPzZpjqx4G8uB5TWxXPDwibLcVa7k,43
+modal-1.1.5.dev52.dist-info/RECORD,,

modal_proto/api.proto

@@ -1345,6 +1345,14 @@ message FlashProxyUpstreamRequest {
   double timestamp = 2; 
 }
 
+message FlashSetTargetSlotsMetricsRequest {
+  // TODO(claudia): add other metrics to use in autoscaling decisions
+  string function_id = 1;
+  uint32 target_slots = 2;
+}
+
+message FlashSetTargetSlotsMetricsResponse {}
+
 message Function {
   string module_name = 1;
   string function_name = 2;
@@ -3629,6 +3637,7 @@ service ModalClient {
   rpc FlashContainerDeregister(FlashContainerDeregisterRequest) returns (google.protobuf.Empty);
   rpc FlashContainerList(FlashContainerListRequest) returns (FlashContainerListResponse);
   rpc FlashContainerRegister(FlashContainerRegisterRequest) returns (FlashContainerRegisterResponse);
+  rpc FlashSetTargetSlotsMetrics(FlashSetTargetSlotsMetricsRequest) returns (FlashSetTargetSlotsMetricsResponse);
 
   // Functions
   rpc FunctionAsyncInvoke(FunctionAsyncInvokeRequest) returns (FunctionAsyncInvokeResponse);

modal_proto/api_grpc.py

@@ -262,6 +262,10 @@ class ModalClientBase(abc.ABC):
     async def FlashContainerRegister(self, stream: 'grpclib.server.Stream[modal_proto.api_pb2.FlashContainerRegisterRequest, modal_proto.api_pb2.FlashContainerRegisterResponse]') -> None:
         pass
 
+    @abc.abstractmethod
+    async def FlashSetTargetSlotsMetrics(self, stream: 'grpclib.server.Stream[modal_proto.api_pb2.FlashSetTargetSlotsMetricsRequest, modal_proto.api_pb2.FlashSetTargetSlotsMetricsResponse]') -> None:
+        pass
+
     @abc.abstractmethod
     async def FunctionAsyncInvoke(self, stream: 'grpclib.server.Stream[modal_proto.api_pb2.FunctionAsyncInvokeRequest, modal_proto.api_pb2.FunctionAsyncInvokeResponse]') -> None:
         pass
@@ -1070,6 +1074,12 @@ class ModalClientBase(abc.ABC):
                 modal_proto.api_pb2.FlashContainerRegisterRequest,
                 modal_proto.api_pb2.FlashContainerRegisterResponse,
             ),
+            '/modal.client.ModalClient/FlashSetTargetSlotsMetrics': grpclib.const.Handler(
+                self.FlashSetTargetSlotsMetrics,
+                grpclib.const.Cardinality.UNARY_UNARY,
+                modal_proto.api_pb2.FlashSetTargetSlotsMetricsRequest,
+                modal_proto.api_pb2.FlashSetTargetSlotsMetricsResponse,
+            ),
             '/modal.client.ModalClient/FunctionAsyncInvoke': grpclib.const.Handler(
                 self.FunctionAsyncInvoke,
                 grpclib.const.Cardinality.UNARY_UNARY,
@@ -2102,6 +2112,12 @@ class ModalClientStub:
             modal_proto.api_pb2.FlashContainerRegisterRequest,
             modal_proto.api_pb2.FlashContainerRegisterResponse,
         )
+        self.FlashSetTargetSlotsMetrics = grpclib.client.UnaryUnaryMethod(
+            channel,
+            '/modal.client.ModalClient/FlashSetTargetSlotsMetrics',
+            modal_proto.api_pb2.FlashSetTargetSlotsMetricsRequest,
+            modal_proto.api_pb2.FlashSetTargetSlotsMetricsResponse,
+        )
         self.FunctionAsyncInvoke = grpclib.client.UnaryUnaryMethod(
             channel,
             '/modal.client.ModalClient/FunctionAsyncInvoke',