fal

from fal import (
    App,
    HealthCheck,
    FalServerlessKeyCredentials,
    ContainerImage,
    function,
    cached,
    endpoint,
    realtime,
    sync_dir,
)

Classes

App

class fal.App

Create a fal serverless application. Subclass this to define your application with custom setup, endpoints, and configuration. The App class handles model loading, request routing, and lifecycle management. Example:

>>> class TextToImage(fal.App, machine_type="GPU"):
...     requirements = ["diffusers", "torch"]
...     
...     def setup(self):
...         self.pipe = StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
...     
...     @fal.endpoint("/")
...     def generate(self, prompt: str) -> dict:
...         image = self.pipe(prompt).images[0]
...         return {"url": fal.toolkit.upload_image(image)}

Inherits from: BaseServable

Constructor Parameters

Name	Type	Default	Description
`_allow_init`	`bool`	`False`	-

Class Variables

Name	Type	Default	Description
`requirements`	`ClassVar[list[str]]`	`[]`	List of pip packages to install in the environment. Supports standard pip syntax including version specifiers. Example: `["numpy==1.24.0", "torch\>=2.0.0"]`
`local_python_modules`	`ClassVar[list[str]]`	`[]`	List of local Python module names to include in the deployment. Use for custom code not available on PyPI. Example: `["my_utils", "models"]`
`machine_type`	`ClassVar[str \| list[str]]`	`'S'`	Compute instance type. Options: ‘S’ (small), ‘M’ (medium), ‘L’ (large), ‘GPU’ (T4), ‘A100’ (A100 40GB). Use a list for fallback. Example: `"GPU"` or `["A100", "GPU"]`
`num_gpus`	`ClassVar[int \| None]`	`None`	Number of GPUs to allocate. Only applies to GPU machine types.
`regions`	`ClassVar[Optional[list[str]]]`	`None`	Allowed regions for deployment. None means any region. Example: `["us-east", "eu-west"]`
`host_kwargs`	`ClassVar[dict[str, Any]]`	`\{'_scheduler': 'nomad', '_scheduler_options': \{'storage_region': 'us-east'\}, 'resolver': 'uv', 'keep_alive': 60\}`	Advanced configuration dictionary passed to the host. For internal use. Prefer using class attributes instead.
`app_name`	`ClassVar[Optional[str]]`	`None`	Custom name for the application. Defaults to class name.
`app_auth`	`ClassVar[Optional[AuthModeLiteral]]`	`None`	Authentication mode. Options: ‘private’ (API key required), ‘public’ (no auth), ‘shared’ (shareable link).
`app_files`	`ClassVar[list[str]]`	`[]`	List of files/directories to include in deployment. Example: `["./models", "./config.yaml"]`
`app_files_ignore`	`ClassVar[list[str]]`	`['\\.pyc$', '__pycache__/', '\\.git/', '\\.DS_Store$']`	Regex patterns to exclude from deployment. Default excludes `.pyc`, `__pycache__`, `.git`, `.DS_Store`.
`app_files_context_dir`	`ClassVar[Optional[str]]`	`None`	Base directory for resolving app_files paths. Defaults to the directory containing the app file.
`request_timeout`	`ClassVar[Optional[int]]`	`None`	Maximum seconds for a single request. None for default.
`startup_timeout`	`ClassVar[Optional[int]]`	`None`	Maximum seconds for app startup/setup. None for default.
`min_concurrency`	`ClassVar[Optional[int]]`	`None`	Minimum warm instances to keep running. Set to 1+ to avoid cold starts. Default is 0 (scale to zero).
`max_concurrency`	`ClassVar[Optional[int]]`	`None`	Maximum instances to scale up to.
`concurrency_buffer`	`ClassVar[Optional[int]]`	`None`	Additional instances to keep warm above current load.
`concurrency_buffer_perc`	`ClassVar[Optional[int]]`	`None`	Percentage buffer of instances above current load.
`scaling_delay`	`ClassVar[Optional[int]]`	`None`	Seconds to wait before scaling down idle instances.
`max_multiplexing`	`ClassVar[Optional[int]]`	`None`	Maximum concurrent requests per instance.
`kind`	`ClassVar[Optional[str]]`	`None`	Deployment kind. For internal use.
`image`	`ClassVar[Optional[ContainerImage]]`	`None`	Custom container image for the application. Use ContainerImage to specify a Dockerfile.
`local_file_path`	`ClassVar[Optional[str]]`	`None`	-
`isolate_channel`	`async_grpc.Channel \| None`	`None`	-

Methods

collect_routes

def collect_routes(self) -> 'dict[RouteSignature, Callable[..., Any]]'

Returns: dict[RouteSignature, Callable[Ellipsis, Any]]

get_endpoints

def get_endpoints(cls) -> 'list[str]'

Returns: list[str]

get_health_check_config

def get_health_check_config(cls) -> 'Optional[ApplicationHealthCheckConfig]'

Returns: Optional[ApplicationHealthCheckConfig]

health

def health(self)

lifespan

def lifespan(self, app: 'fastapi.FastAPI')

Parameter	Type	Default	Description
`app`	`FastAPI`	-	-

provide_hints

def provide_hints(self) -> 'list[str]'

Provide hints for routing the application.Returns: list[str]

setup

def setup(self)

Setup the application before serving.

teardown

def teardown(self)

Teardown the application after serving.

HealthCheck

class fal.HealthCheck

Constructor Parameters

Name	Type	Default	Description
`start_period_seconds`	`Optional[int]`	`None`	Minimum time the runner has been running before considering the runner unhealthy when health check fails. To prevent the health check from failing too early, this will be replaced by startup_timeout of the application if it’s less than it. Defaults to 30.
`timeout_seconds`	`Optional[int]`	`None`	Timeout in seconds for the health check request. Defaults to 5 seconds.
`failure_threshold`	`Optional[int]`	`None`	Number of consecutive failures before considering the runner as unhealthy. Defaults to 3.
`call_regularly`	`Optional[bool]`	`None`	Perform health check every 15s. If false, only do it when the x-fal-runner-health-check header is present. Defaults to True.

Class Variables

Name	Type	Default	Description
`start_period_seconds`	`Optional[int]`	`None`	-
`timeout_seconds`	`Optional[int]`	`None`	-
`failure_threshold`	`Optional[int]`	`None`	-
`call_regularly`	`Optional[bool]`	`None`	-

FalServerlessKeyCredentials

class fal.FalServerlessKeyCredentials

Inherits from: Credentials

Constructor Parameters

Name	Type	Default	Description
`key_id`	`str`	-	-
`key_secret`	`str`	-	-

Class Variables

Name	Type	Default	Description
`key_id`	`str`	-	-
`key_secret`	`str`	-	-

Methods

to_grpc

def to_grpc(self) -> 'grpc.ChannelCredentials'

Returns: ChannelCredentials

to_headers

def to_headers(self) -> 'dict[str, str]'

Returns: dict[str, str]

ContainerImage

class fal.ContainerImage

ContainerImage represents a Docker image that can be built from a Dockerfile.

Constructor Parameters

Name	Type	Default	Description
`dockerfile_str`	`str`	-	-
`build_args`	`dict[str, str]`	`\<factory\>`	-
`registries`	`dict[str, dict[str, str]]`	`\<factory\>`	-
`builder`	`Optional[Literal[depot, service, worker]]`	`None`	-
`compression`	`str`	`'gzip'`	-
`force_compression`	`bool`	`False`	-
`secrets`	`dict[str, str]`	`\<factory\>`	-

Class Variables

Name	Type	Default	Description
`dockerfile_str`	`str`	-	-
`build_args`	`dict[str, str]`	-	-
`registries`	`dict[str, dict[str, str]]`	-	-
`builder`	`Optional[Literal[depot, service, worker]]`	`None`	-
`compression`	`str`	`'gzip'`	-
`force_compression`	`bool`	`False`	-
`secrets`	`dict[str, str]`	-	-

Methods

from_dockerfile

def from_dockerfile(cls, path: str, **kwargs) -> 'ContainerImage'

Parameter	Type	Default	Description
`path`	`str`	-	-
`kwargs`	-	-	-

Returns: ContainerImage

from_dockerfile_str

def from_dockerfile_str(cls, text: str, **kwargs) -> 'ContainerImage'

Parameter	Type	Default	Description
`text`	`str`	-	-
`kwargs`	-	-	-

Returns: ContainerImage

to_dict

def to_dict(self) -> dict

Returns: dict

Functions

function

def function(kind: 'str' = 'virtualenv', *, host: 'Host | None' = None, local_python_modules: 'list[str] | None' = None, **config: 'Any')

Parameter	Type	Default	Description
`kind`	`str`	`'virtualenv'`	-
`host`	`fal.api.api.Host \| None`	`None`	-
`local_python_modules`	`list[str] \| None`	`None`	-
`config`	`Any`	-	-

cached

def cached(func: 'Callable[ArgsT, ReturnT]') -> 'Callable[ArgsT, ReturnT]'

Cache the result of the given function in-memory.

Parameter	Type	Default	Description
`func`	`Callable[ArgsT, ReturnT]`	-	-

Returns: Callable[ArgsT, ReturnT]

endpoint

def endpoint(path: 'str', *, is_websocket: 'bool' = False, health_check: 'HealthCheck | None' = None) -> 'Callable[[EndpointT], EndpointT]'

Designate the decorated function as an application endpoint.

Parameter	Type	Default	Description
`path`	`str`	-	-
`is_websocket`	`bool`	`False`	-
`health_check`	`fal.sdk.HealthCheck \| None`	`None`	-

Returns: Callable[EndpointT, EndpointT]

realtime

def realtime(path: 'str', *, buffering: 'int | None' = None, session_timeout: 'float | None' = None, input_modal: 'Any | None' = <object object at 0x102ffa3a0>, max_batch_size: 'int' = 1) -> 'Callable[[EndpointT], EndpointT]'

Designate the decorated function as a realtime application endpoint.

Parameter	Type	Default	Description
`path`	`str`	-	-
`buffering`	`int \| None`	`None`	-
`session_timeout`	`float \| None`	`None`	-
`input_modal`	`Any \| None`	`\<object object at 0x102ffa3a0\>`	-
`max_batch_size`	`int`	`1`	-

Returns: Callable[EndpointT, EndpointT]

sync_dir

def sync_dir(local_dir: 'str | Path', remote_dir: 'str', force_upload=False) -> 'str'

Parameter	Type	Default	Description
`local_dir`	`str \| pathlib.Path`	-	-
`remote_dir`	`str`	-	-
`force_upload`	-	`False`	-

Returns: str

Client Libraries

Serverless SDK

Platform APIs

Classes

App

collect_routes

get_endpoints

get_health_check_config

health

lifespan

provide_hints

setup

teardown

HealthCheck

FalServerlessKeyCredentials

to_grpc

to_headers

ContainerImage

from_dockerfile

from_dockerfile_str

to_dict

Functions

function

cached

endpoint

realtime

sync_dir

Client Libraries

Serverless SDK

Platform APIs

​Classes

​App

​collect_routes

​get_endpoints

​get_health_check_config

​health

​lifespan

​provide_hints

​setup

​teardown

​HealthCheck

​FalServerlessKeyCredentials

​to_grpc

​to_headers

​ContainerImage

​from_dockerfile

​from_dockerfile_str

​to_dict

​Functions

​function

​cached

​endpoint

​realtime

​sync_dir

Classes

App

collect_routes

get_endpoints

get_health_check_config

health

lifespan

provide_hints

setup

teardown

HealthCheck

FalServerlessKeyCredentials

to_grpc

to_headers

ContainerImage

from_dockerfile

from_dockerfile_str

to_dict

Functions

function

cached

endpoint

realtime

sync_dir