Connector

class Connector

Entry point for connecting your agent process to the AsterQuanta platform. The Connector maintains the WebSocket connection, handles incoming optimization requests, and dispatches work to your executor implementation.

Import

from adk.connector import Connector

Typical usage (RL agent)

from adk.connector import Connector
from adk.executors.rl import RLExecutor
from my_agent import MyAgent

app = Connector(
    RLExecutor,
    rl_agent_env_class=MyAgent,
    model_handler=MyAgent,
    throw_errors=True,
)
app.start()

Typical usage (custom executor)

from adk.connector import Connector
from adk.model_handler import ModelHandler
from my_executor import MyOptimizerExecutor
from my_models import MyModelHandler

app = Connector(
    MyOptimizerExecutor,
    model_handler=MyModelHandler,
)
app.start()

Constructor

Connector(
    executor_class=RLExecutor,
    *,
    model_handler=None,
    throw_errors=False,
    **executor_kwargs,
)

Parameter	Description
executor_class	Executor type. Defaults to RLExecutor. Pass a BaseExecutor subclass for custom logic.
model_handler	Class implementing ModelHandler static methods. Required when the platform invokes model transfer, export, or import routes. Often the same class as your RL agent.
throw_errors	If True, WebSocket connection errors propagate instead of being logged and retried. Separate from RLExecutor's throw_errors kwarg, which controls optimization-time error handling.
**executor_kwargs	Forwarded to the executor constructor. For RLExecutor, common kwargs are rl_agent_env_class, config, and throw_errors.

Methods

start

def start(self) -> None

Blocking call. Runs startup checks, starts model sync for local models/ directories, connects to the platform, and processes requests until interrupted (Ctrl+C).

note

Ensure .env contains a valid SECRET from genie setup before calling start().

What happens on start

1. Startup routine validates project configuration (including registration and agent name).
2. Resource usage logging is initialized if enabled in settings.json.
3. Model sync watches models/ and keeps the platform in sync with local model files.
4. WebSocket connects using SECRET, ADK version, and instance-id headers when an instance id is known.
5. If no instance id is configured, the platform may assign one on connect and persist it to .env.
6. Incoming optimization requests instantiate your executor with the job's optimization payload.

See Agent Settings and Environment Settings for resource_logging_options and INSTANCE_ID.

See RLExecutor and BaseExecutor for implementing the optimization logic itself.