Clients and Frontends

The basic ClientManager can support any kind of frontend or client. However, writing your own frontend may not be the fastest path to running your experiment with CHEESE. It is reccomended to instead use the GradioClientManager and to create your frontend in Gradio. If you should choose to do this, the GradioFront object is all you will need for the frontend of your experiment.

class cheese.client.ClientManager[source]

add_client(id: int, client_cls, **kwargs) → str[source]: Add a client to the ClientManager and return a url to the frontend.

get_idle_clients() → int[source]: Get count of how many clients are idle.

init_connection(connection: b_rabbit.b_rabbit.BRabbit)[source]: Initialize message channel and consumption callbacks.

notify_completion(id: int)[source]: Notify the ClientManager that a client has completed a task.

queue_task(id: int)[source]: Given a client id, queues the task that was assigned to that client and marks client as free or active accordingly.

remove_client(id: int)[source]: Remove a client from the ClientManager. Note: this drops any task that client is working on.

class cheese.client.gradio_client.GradioClientManager(no_login: bool = False)[source]

ClientManager for frontends made in Gradio

add_client(id: int)[source]

Add a new client. Creates a user/pass combo.

Parameters: id – ID for the new client

await_new_task(id: int) → cheese.tasks.Task [source]

GradioFront should call this with ID of client. It will return a new task if one is available. Otherwise, it will loop and wait for one.

Parameters: id – ID of the client awaiting a task
Returns: A task, as soon as it is available

load_user_info(path)[source]: Load info on users from a previously saved file

queue_task(id: int, task: cheese.tasks.Task)[source]: Given a client id, queues the task that was assigned to that client and marks client as free or active accordingly.

remove_client(id: int)[source]: Remove a client from the ClientManager. Note: this drops any task that client is working on.

save_user_info(path)[source]: Save info on current users to a file

submit_task(id: int, task: cheese.tasks.Task)[source]

GradioFront should call this to submit a finished task.

Parameters

id – ID of the client submitting a task
task – The finished task

class cheese.client.gradio_client.GradioFront(no_login: bool = False)[source]

Frontend for CHEESE using Gradio

Parameters: no_login (bool) – If True, will not require login. Useful for testing

handle_input_exception(*args) → Any[source]

Handle invalid input exceptions. Default behavior is to just present same data again.

Parameters

*args –

List of arguments that caused the exception

Returns

Outputs for gradio demo

Return type

Iterable[Any] or Any

launch()[source]: Launch Gradio demo

login()[source]: Returns basic components for login screen

abstract main() → List[gradio.components.Component][source]

Gradio interface for collecting data can be written here. Should call GradioFront.response() with self. Please read the documentation of GradioFront.response for information on which specific inputs and outputs must go to/come out of the function.

Returns: List of all output components in the gradio demo

abstract present(task: cheese.tasks.Task) → List[gradio.components.Component][source]: Present data in the task to user. Should take task and return outputs to gradio functions in list form

abstract receive(*inp) → cheese.tasks.Task [source]

Receive input from user and modify the data in the task with it. Can enforce input validity by raising InvalidInputException. Assumes first two parameters in inp are ID and Task respectively.

Returns: task modified to reflect user input

response(*inp) → Any[source]

Submit input from user then stall until we have an output ready for them. Assumes first two parameters in inp are ID and Task respectively

Parameters: inp – Inputs from gradio components
Returns: New task, Outputs to give to gradio components

set_manager(manager: cheese.client.gradio_client.GradioClientManager)[source]: Set the manager for the frontend. This is how it will communicate with cheese. Must be set before calling launch

wrap_event(event: Callable)[source]: For any gradio event (i.e. gr.Button.click, gr.Slider.change), wrap with this method to ensure id and task are passed properly.