fal 1.59.0__tar.gz → 1.61.0__tar.gz

{fal-1.59.0/fal.egg-info → fal-1.61.0}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.4
 Name: fal
-Version: 1.59.0
+Version: 1.61.0
 Summary: fal is an easy-to-use Serverless Python Framework
 Author: Features & Labels <support@fal.ai>
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
-Requires-Dist: isolate[build]<0.22.0,>=0.18.0
-Requires-Dist: isolate-proto>=0.29.2
+Requires-Dist: isolate[build]<0.23.0,>=0.22.0
+Requires-Dist: isolate-proto>=0.29.4
 Requires-Dist: grpcio<2,>=1.64.0
 Requires-Dist: dill==0.3.7
 Requires-Dist: cloudpickle==3.0.0

{fal-1.59.0 → fal-1.61.0/fal.egg-info}/PKG-INFO

@@ -1,12 +1,12 @@
 Metadata-Version: 2.4
 Name: fal
-Version: 1.59.0
+Version: 1.61.0
 Summary: fal is an easy-to-use Serverless Python Framework
 Author: Features & Labels <support@fal.ai>
 Requires-Python: >=3.8
 Description-Content-Type: text/markdown
-Requires-Dist: isolate[build]<0.22.0,>=0.18.0
-Requires-Dist: isolate-proto>=0.29.2
+Requires-Dist: isolate[build]<0.23.0,>=0.22.0
+Requires-Dist: isolate-proto>=0.29.4
 Requires-Dist: grpcio<2,>=1.64.0
 Requires-Dist: dill==0.3.7
 Requires-Dist: cloudpickle==3.0.0

{fal-1.59.0 → fal-1.61.0}/fal.egg-info/SOURCES.txt

@@ -92,6 +92,7 @@ src/fal/project.py
 src/fal/py.typed
 src/fal/sdk.py
 src/fal/sync.py
+src/fal/upload.py
 src/fal/utils.py
 src/fal/workflows.py
 src/fal/api/__init__.py

{fal-1.59.0 → fal-1.61.0}/fal.egg-info/requires.txt

@@ -1,5 +1,5 @@
-isolate[build]<0.22.0,>=0.18.0
-isolate-proto>=0.29.2
+isolate[build]<0.23.0,>=0.22.0
+isolate-proto>=0.29.4
 grpcio<2,>=1.64.0
 dill==0.3.7
 cloudpickle==3.0.0

{fal-1.59.0 → fal-1.61.0}/pyproject.toml

@@ -22,8 +22,8 @@ authors = [{ name = "Features & Labels <support@fal.ai>"}]
 readme = "README.md"
 requires-python = ">=3.8"
 dependencies = [
-    "isolate[build]>=0.18.0,<0.22.0",
-    "isolate-proto>=0.29.2",
+    "isolate[build]>=0.22.0,<0.23.0",
+    "isolate-proto>=0.29.4",
     "grpcio>=1.64.0,<2",
     "dill==0.3.7",
     "cloudpickle==3.0.0",

{fal-1.59.0 → fal-1.61.0}/src/fal/_fal_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '1.59.0'
-__version_tuple__ = version_tuple = (1, 59, 0)
+__version__ = version = '1.61.0'
+__version_tuple__ = version_tuple = (1, 61, 0)
 
 __commit_id__ = commit_id = None

fal-1.61.0/src/fal/api/__init__.py

@@ -0,0 +1,16 @@
+from .api import *  # noqa: F403
+from .client import (  # noqa: F401
+    AppsNamespace,
+    KeysNamespace,
+    RunnersNamespace,
+    SecretsNamespace,
+    SyncServerlessClient,
+)
+
+__all__ = [
+    "SyncServerlessClient",
+    "AppsNamespace",
+    "RunnersNamespace",
+    "KeysNamespace",
+    "SecretsNamespace",
+]

{fal-1.59.0 → fal-1.61.0}/src/fal/api/api.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import asyncio
 import inspect
 import os
 import sys
@@ -450,6 +451,7 @@ class FalServerlessHost(Host):
             "app_files_ignore",
             "app_files_context_dir",
             "health_check_config",
+            "skip_retry_conditions",
         }
     )
 
@@ -548,6 +550,8 @@ class FalServerlessHost(Host):
         request_timeout = options.host.get("request_timeout")
         startup_timeout = options.host.get("startup_timeout")
         regions = options.host.get("regions")
+        health_check_config = options.host.get("health_check_config")
+        skip_retry_conditions = options.host.get("skip_retry_conditions")
         machine_requirements = MachineRequirements(
             machine_types=machine_type,  # type: ignore
             num_gpus=options.host.get("num_gpus"),
@@ -567,8 +571,6 @@ class FalServerlessHost(Host):
             valid_regions=regions,
         )
 
-        health_check_config = options.host.get("health_check_config")
-
         app_files = self._app_files_sync(options)
 
         partial_func = _prepare_partial_func(func)
@@ -596,6 +598,7 @@ class FalServerlessHost(Host):
             # By default, logs are public
             private_logs=options.host.get("private_logs", False),
             files=app_files,
+            skip_retry_conditions=skip_retry_conditions,
         ):
             for log in partial_result.logs:
                 self._log_printer.print(log)
@@ -699,16 +702,20 @@ class FalServerlessHost(Host):
             from fal.console import console
 
             if service_urls := partial_result.service_urls:
-                console.print("Playground:")
+                from fal.flags import URL_OUTPUT
+
                 endpoints = getattr(func, "_routes", ["/"])  # type: ignore[attr-defined]
-                for endpoint in endpoints:
-                    console.print(f"\t{service_urls.playground}{endpoint}")
-                console.print("Synchronous Endpoints:")
-                for endpoint in endpoints:
-                    console.print(f"\t{service_urls.run}{endpoint}")
-                console.print("Asynchronous Endpoints (Recommended):")
-                for endpoint in endpoints:
-                    console.print(f"\t{service_urls.queue}{endpoint}")
+                if URL_OUTPUT != "none":
+                    console.print("Playground:")
+                    for endpoint in endpoints:
+                        console.print(f"\t{service_urls.playground}{endpoint}")
+                if URL_OUTPUT == "all":
+                    console.print("Synchronous Endpoints:")
+                    for endpoint in endpoints:
+                        console.print(f"\t{service_urls.run}{endpoint}")
+                    console.print("Asynchronous Endpoints (Recommended):")
+                    for endpoint in endpoints:
+                        console.print(f"\t{service_urls.queue}{endpoint}")
 
             for log in partial_result.logs:
                 if (
@@ -801,6 +808,7 @@ def function(
     exposed_port: int | None = None,
     max_concurrency: int | None = None,
     local_python_modules: list[str] | None = None,
+    force_env_build: bool = False,
 ) -> Callable[
     [Callable[Concatenate[ArgsT], ReturnT]], IsolatedFunction[ArgsT, ReturnT]
 ]: ...
@@ -818,6 +826,7 @@ def function(
     exposed_port: int | None = None,
     max_concurrency: int | None = None,
     local_python_modules: list[str] | None = None,
+    force_env_build: bool = False,
 ) -> Callable[
     [Callable[Concatenate[ArgsT], ReturnT]], ServedIsolatedFunction[ArgsT, ReturnT]
 ]: ...
@@ -850,6 +859,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -883,6 +893,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -968,6 +979,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -1006,6 +1018,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -1038,6 +1051,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -1070,6 +1084,7 @@ def function(
     request_timeout: int | None = None,
     startup_timeout: int | None = None,
     setup_function: Callable[..., None] | None = None,
+    force_env_build: bool = False,
     _base_image: str | None = None,
     _scheduler: str | None = None,
 ) -> Callable[
@@ -1095,6 +1110,10 @@ def function(  # type: ignore
     if kind == "container" and config.get("app_files"):
         raise ValueError("app_files is not supported for container apps.")
 
+    if config.get("force_env_build") is not None:
+        force_env_build = config.pop("force_env_build")
+        config["force"] = force_env_build
+
     options = host.parse_options(kind=kind, **config)
 
     def wrapper(func: Callable[ArgsT, ReturnT]):
@@ -1235,8 +1254,16 @@ class BaseServable:
             # This is supposed to make it easier to understand to the user
             # that the error comes from the app and not our platform.
             if exc.detail == "Not Found":
+                # For 404 errors (non-existent endpoints), set billable units to 0.
+                # This prevents users from being charged when they hit endpoints that
+                # don't exist. Without this, the platform would use the default billable
+                # units for the endpoint, incorrectly charging users for failed requests
+                headers = dict(exc.headers) if exc.headers else {}
+                headers["x-fal-billable-units"] = "0"
                 return JSONResponse(
-                    {"detail": f"Path {request.url.path} not found"}, 404
+                    {"detail": f"Path {request.url.path} not found"},
+                    404,
+                    headers=headers,
                 )
             else:
                 # If it's not a generic 404, just return the original message.
@@ -1248,7 +1275,17 @@ class BaseServable:
 
         @_app.exception_handler(FieldException)
         async def field_exception_handler(request: Request, exc: FieldException):
-            return JSONResponse(exc.to_pydantic_format(), exc.status_code)
+            headers = {}
+            if exc.billable_units:
+                # poor man's validation. we dont want people to pass in
+                # non-numeric values.
+                units_float = float(exc.billable_units)
+                # we dont want to add 8 decimal places for ints.
+                format_string = ".0f" if isinstance(exc.billable_units, int) else ".8f"
+                headers["x-fal-billable-units"] = format(units_float, format_string)
+            return JSONResponse(
+                exc.to_pydantic_format(), exc.status_code, headers=headers
+            )
 
         # ref: https://github.com/fastapi/fastapi/blob/37c8e7d76b4b47eb2c4cced6b4de59eb3d5f08eb/fastapi/exception_handlers.py#L20
         @_app.exception_handler(RequestValidationError)
@@ -1331,12 +1368,9 @@ class BaseServable:
                 "Failed to generate OpenAPI metadata for function"
             ) from e
 
-    def serve(self) -> None:
-        import asyncio
-
+    async def serve(self) -> None:
         from prometheus_client import Gauge
         from starlette_exporter import handle_metrics
-        from uvicorn import Config
 
         # NOTE: this uses the global prometheus registry
         app_info = Gauge("fal_app_info", "Fal application information", ["version"])
@@ -1346,40 +1380,32 @@ class BaseServable:
 
         # We use the default workers=1 config because setup function can be heavy
         # and it runs once per worker.
-        server = Server(
-            config=Config(app, host="0.0.0.0", port=8080, timeout_keep_alive=300)
+        server = uvicorn.Server(
+            config=uvicorn.Config(
+                app, host="0.0.0.0", port=8080, timeout_keep_alive=300, lifespan="on"
+            )
         )
         metrics_app = FastAPI()
         metrics_app.add_route("/metrics", handle_metrics)
-        metrics_server = Server(config=Config(metrics_app, host="0.0.0.0", port=9090))
+        metrics_server = uvicorn.Server(
+            config=uvicorn.Config(metrics_app, host="0.0.0.0", port=9090)
+        )
 
         async def _serve() -> None:
             tasks = {
-                asyncio.create_task(server.serve()): server,
-                asyncio.create_task(metrics_server.serve()): metrics_server,
+                asyncio.create_task(server.serve()),
+                asyncio.create_task(metrics_server.serve()),
             }
 
-            _, pending = await asyncio.wait(
-                tasks.keys(),
-                return_when=asyncio.FIRST_COMPLETED,
+            await asyncio.wait(
+                tasks,
+                return_when=asyncio.ALL_COMPLETED,
             )
-            if not pending:
-                return
-
-            # try graceful shutdown
-            for task in pending:
-                tasks[task].should_exit = True
-            _, pending = await asyncio.wait(pending, timeout=2)
-            if not pending:
-                return
+            # we do not take care of pending tasks here.
+            # each task should be responsible for its own cleanup.
+            # graceful termination and timeout should be handled by external scheduler.
 
-            for task in pending:
-                task.cancel()
-            await asyncio.wait(pending)
-
-        with suppress(asyncio.CancelledError):
-            asyncio.set_event_loop(asyncio.new_event_loop())
-            asyncio.run(_serve())
+        await _serve()
 
 
 class ServeWrapper(BaseServable):
@@ -1393,13 +1419,13 @@ class ServeWrapper(BaseServable):
             RouteSignature("/"): self._func,
         }
 
-    def __call__(self, *args, **kwargs) -> None:
+    async def __call__(self, *args, **kwargs) -> None:
         if len(args) != 0 or len(kwargs) != 0:
             print(
                 f"[warning] {self._func.__name__} function is served with no arguments."
             )
 
-        self.serve()
+        await self.serve()
 
 
 @dataclass
@@ -1519,7 +1545,7 @@ class IsolatedFunction(Generic[ArgsT, ReturnT]):
         if serve_mode:
             # This type can be safely ignored because this case only happens when it
             # is a ServedIsolatedFunction
-            serve_func: Callable[[], None] = ServeWrapper(self.raw_func)
+            serve_func = ServeWrapper(self.raw_func)
             return serve_func  # type: ignore
         else:
             return self.raw_func
@@ -1545,15 +1571,3 @@ class ServedIsolatedFunction(
     def on(
         self, host: Host | None = None, *, serve: Literal[False], **config: Any
     ) -> IsolatedFunction[ArgsT, ReturnT]: ...
-
-
-class Server(uvicorn.Server):
-    """Server is a uvicorn.Server that actually plays nicely with signals.
-    By default, uvicorn's Server class overwrites the signal handler for SIGINT,
-    swallowing the signal and preventing other tasks from cancelling.
-    This class allows the task to be gracefully cancelled using asyncio's built-in task
-    cancellation or with an event, like aiohttp.
-    """
-
-    def install_signal_handlers(self) -> None:
-        pass