configuration.md

Configuration

Dimos provides a Configurable base class. See service/spec.py.

This allows using dataclasses to specify configuration structure and default values per module.

from dimos.protocol.service import Configurable
from rich import print
from dataclasses import dataclass

@dataclass
class Config():
    x: int = 3
    hello: str = "world"

class MyClass(Configurable):
    default_config = Config
    config: Config
    def __init__(self, **kwargs) -> None:
        super().__init__(**kwargs)

myclass1 = MyClass()
print(myclass1.config)

# can easily override
myclass2 = MyClass(hello="override")
print(myclass2.config)

# we will raise an error for unspecified keys
try:
    myclass3 = MyClass(something="else")
except TypeError as e:
    print(f"Error: {e}")

Config(x=3, hello='world')
Config(x=3, hello='override')
Error: Config.__init__() got an unexpected keyword argument 'something'

Configurable Modules

Modules inherit from Configurable, so all of the above applies. Module configs should inherit from ModuleConfig (core/module.py), which includes shared configuration for all modules like transport protocols, frame IDs, etc.

from dataclasses import dataclass
from dimos.core.core import rpc
from dimos.core.module import Module, ModuleConfig
from dimos.core.stream import In, Out
from rich import print

@dataclass
class Config(ModuleConfig):
    frame_id: str = "world"
    publish_interval: float = 0
    voxel_size: float = 0.05
    device: str = "CUDA:0"

class MyModule(Module):
    default_config = Config
    config: Config

    def __init__(self, **kwargs) -> None:
        super().__init__(**kwargs)
        print(self.config)


myModule = MyModule(frame_id="frame_id_override", device="CPU")

# In production, use dimos.deploy() instead:
# myModule = dimos.deploy(MyModule, frame_id="frame_id_override")

Config(
    rpc_transport=<class 'dimos.protocol.rpc.pubsubrpc.LCMRPC'>,
    tf_transport=<class 'dimos.protocol.tf.tf.LCMTF'>,
    frame_id_prefix=None,
    frame_id='frame_id_override',
    publish_interval=0,
    voxel_size=0.05,
    device='CPU'
)