The MADSci Experiment Application provides a high-level framework for creating and managing scientific experiments within the MADSci ecosystem. This package serves as the primary interface for scientists and researchers to design, execute, and monitor automated and autonomous experimental campaigns.
Overview¶
The ExperimentApplication class provides a comprehensive framework for experiment management that can:
Manage complete experiment lifecycle (start, pause, resume, cancel, end)
Integrate with all MADSci manager services (experiment, data, resource, workcell, event, location, lab)
Evaluate complex resource and location conditions before and during experiments
Operate in both standalone execution and server modes as a REST node
Provide automatic experiment context management and error handling
Key Features¶
Experiment Lifecycle Management: Complete control over experiment states with proper error handling
Condition Evaluation System: Advanced validation of resource availability, locations, and field constraints
Context Management: Automatic experiment setup and teardown with context managers
Dual Operation Modes: Standalone execution or REST node server for remote control
Exception Handling: Built-in handling for experiment failures, cancellations, and pauses
Event Logging: Integrated logging and monitoring through the MADSci event system
Resource Management: Real-time resource tracking and validation
Dependencies¶
madsci.common: Shared types, utilities, and base classesmadsci.client: Client libraries for all MADSci manager servicesmadsci.node_module: REST node framework for server mode operationrich: Enhanced console output and formattingpydantic: Data validation and configuration management
Installation¶
# Using PDM (for development)
pdm install
# Using pip (for production use)
pip install madsci.experiment_applicationQuick Start¶
Basic Usage¶
from madsci.experiment_application import ExperimentApplication
from madsci.common.types.experiment_types import ExperimentDesign
# Option 1: Create with experiment design
experiment_design = ExperimentDesign.from_yaml("my_experiment.yaml")
app = ExperimentApplication(
lab_server_url="http://localhost:8000",
experiment_design=experiment_design
)
# Start and manage an experiment using context manager
with app.manage_experiment(run_name="Test Run", run_description="Initial testing"):
# Your experiment code here
app.logger.info("Running experiment steps...")
# Experiment automatically ends on successful completion
# or fails on exceptionClass Methods for Different Scenarios¶
# Start a completely new experiment
app = ExperimentApplication.start_new(
lab_server_url="http://localhost:8000",
experiment_design=experiment_design
)
# Continue an existing experiment
existing_experiment = experiment_client.get_experiment("01ARZ3NDEKTSV4RRFFQ69G5FAV")
app = ExperimentApplication.continue_experiment(
experiment=existing_experiment,
lab_server_url="http://localhost:8000"
)Manual Experiment Control¶
app = ExperimentApplication(
lab_server_url="http://localhost:8000",
experiment_design=experiment_design
)
# Manual lifecycle management
app.start_experiment_run(run_name="Manual Control Run")
try:
# Your experiment logic
app.check_experiment_status() # Checks for pause/cancel/fail
# ... more experiment steps
except ExperimentCancelledError:
app.logger.info("Experiment was cancelled")
except ExperimentFailedError:
app.logger.info("Experiment failed")
else:
app.end_experiment()Condition Evaluation System¶
The application includes a powerful condition evaluation system for validating resources, locations, and field values before and during experiments.
Condition Types¶
from madsci.common.types.condition_types import Condition
# Resource presence at location
condition = Condition(
condition_name="sample_present",
condition_type="resource_present",
location_name="sample_holder_1",
key="0", # Position in location
resource_class="Sample"
)
# Resource field validation
condition = Condition(
condition_name="temperature_check",
condition_type="resource_field_check",
resource_name="furnace_1",
field="temperature",
operator="is_greater_than",
target_value=500.0
)
# Resource child field validation
condition = Condition(
condition_name="sample_volume_check",
condition_type="resource_child_field_check",
resource_name="sample_rack",
key="position_1",
field="volume",
operator="is_greater_than_or_equal_to",
target_value=1.0
)
# Evaluate conditions
if app.evaluate_condition(condition):
app.logger.info("Condition satisfied")Supported Operators¶
is_greater_thanis_less_thanis_equal_tois_greater_than_or_equal_tois_less_than_or_equal_to
Server Mode Operation¶
The application can operate as a REST node for remote experiment execution:
from madsci.experiment_application import ExperimentApplicationConfig
# Configure for server mode
config = ExperimentApplicationConfig(
server_mode=True,
node_name="experiment_app_server",
node_description="Automated experiment server",
port=8010
)
class MyExperimentApp(ExperimentApplication):
def __init__(self):
super().__init__(config=config)
def run_experiment(self, parameter1: str, parameter2: float):
"""Custom experiment implementation"""
with self.manage_experiment(run_name=f"Auto_{parameter1}"):
# Experiment logic using parameters
self.logger.info(f"Running with {parameter1}, {parameter2}")
# Start server
app = MyExperimentApp()
app.start_app() # Starts REST server on configured portConfiguration¶
Environment Variables¶
All configuration can be set via environment variables with the EXPERIMENT_ prefix:
export EXPERIMENT_SERVER_MODE=true
export EXPERIMENT_LAB_SERVER_URL=http://localhost:8000
export EXPERIMENT_NODE_NAME=my_experiment_app
export EXPERIMENT_PORT=8010Configuration Files¶
Supports multiple configuration file formats:
.envorexperiment.envsettings.tomlorexperiment.settings.tomlsettings.yamlorexperiment.settings.yamlsettings.jsonorexperiment.settings.json
Example experiment.settings.toml:
server_mode = false
lab_server_url = "http://localhost:8000"
node_name = "experiment_application"
port = 8010
[run_args]
# Arguments for run_experiment when not in server mode
[run_kwargs]
# Keyword arguments for run_experiment when not in server mode
parameter1 = "default_value"
parameter2 = 100.0Advanced Usage¶
Custom Exception Handling¶
class MyExperimentApp(ExperimentApplication):
def handle_exception(self, exception: Exception):
"""Custom exception handling"""
if isinstance(exception, KeyboardInterrupt):
self.logger.info("Experiment interrupted by user")
self.pause_experiment()
else:
# Use default behavior
super().handle_exception(exception)Experiment Loop for Autonomous Operation¶
class AutonomousApp(ExperimentApplication):
@threaded_daemon
def loop(self):
"""Continuous experiment loop"""
while True:
# Check conditions
if self.evaluate_condition(my_condition):
with self.manage_experiment():
# Run experiment iteration
pass
time.sleep(60) # Wait before next iterationUsing the Decorator Pattern¶
app = ExperimentApplication(experiment_design=design)
@app.add_experiment_management
def my_experiment_function(param1: str, param2: int):
"""This function will automatically be wrapped with experiment management"""
app.logger.info(f"Running experiment with {param1}, {param2}")
# Function automatically starts/ends experiment
# Call the decorated function
my_experiment_function("test", 42)Manager Service Integration¶
The application provides direct access to all MADSci manager services:
app = ExperimentApplication(lab_server_url="http://localhost:8000")
# Access to all manager clients
app.experiment_client # Experiment management (port 8002)
app.data_client # Data storage and retrieval (port 8004)
app.resource_client # Resource and inventory tracking (port 8003)
app.workcell_client # Workflow coordination (port 8005)
app.event_client # Event logging (port 8001)
app.location_client # Location management (port 8006)
app.lab_client # Lab configuration
# Example usage
resources = app.resource_client.list_resources()
app.data_client.store_data(experiment_id=app.experiment.experiment_id, data=results)Error Handling¶
The application provides comprehensive error handling for experiment states:
from madsci.common.exceptions import ExperimentCancelledError, ExperimentFailedError
try:
with app.manage_experiment():
app.check_experiment_status() # Raises exception if cancelled/failed
# Experiment logic
except ExperimentCancelledError:
app.logger.info("Experiment was cancelled externally")
except ExperimentFailedError:
app.logger.info("Experiment failed externally")
except Exception as e:
app.logger.error(f"Unexpected error: {e}")
# Application will automatically mark experiment as failedULID Usage¶
MADSci uses ULID (Universally Unique Lexicographically Sortable Identifier) for all ID generation:
from madsci.common.utils import new_ulid_str
# All experiment IDs, resource IDs, etc. are ULIDs
experiment_id = app.experiment.experiment_id # ULID string
new_resource_id = new_ulid_str() # Generate new ULID