In Prefect 3.4, you can access flow failure parameters in your failure handler through the arguments provided to the handler function. Here's how to implement it:

python
from prefect import flow
from prefect.client.schemas.objects import Flow, FlowRun, State

def failure_handler(flow: Flow, flow_run: FlowRun, state: State):
    # Access flow parameters
    flow_params = flow_run.parameters
    
    # Access error information
    error_message = state.message
    error_type = state.type
    
    # Access flow metadata
    flow_name = flow.name
    flow_run_id = flow_run.id
    
    print(f"Flow '{flow_name}' failed with parameters: {flow_params}")
    print(f"Error: {error_message}")
    print(f"Flow run ID: {flow_run_id}")

@flow(on_failure=[failure_handler])
def my_flow(x: int, y: str):
    raise ValueError("Something went wrong")

# Example usage
my_flow(x=1, y="test")

The failure handler receives three important objects: 1. `flow`: The Flow object containing flow metadata - `flow.name`: Name of the flow - `flow.version`: Flow version - Other flow-level attributes 2. `flow_run`: The FlowRun object containing runtime information - `flow_run.id`: Unique identifier for this run - `flow_run.parameters`: Dictionary of parameters passed to the flow - `flow_run.start_time`: When the flow started - `flow_run.expected_start_time`: Scheduled start time 3. `state`: The State object containing failure details - `state.message`: Error message - `state.type`: State type (e.g., "FAILED") - `state.timestamp`: When the failure occurred You can also create an async failure handler:

python
from prefect import flow

async def async_failure_handler(flow: Flow, flow_run: FlowRun, state: State):
    # Access same parameters as sync handler
    params = flow_run.parameters
    error = state.message
    
    # Perform async operations
    await notify_slack(f"Flow failed with params {params}, error: {error}")

@flow(on_failure=[async_failure_handler])
def my_flow(x: int, y: str):
    raise ValueError("Something went wrong")

The handler will be called automatically when the flow fails, and you'll have access to all the flow's runtime information to handle the failure appropriately.