Configuration Module

Configuration package - uses varlord for config management.

IMPORTANT: All configuration models are defined in models.py. No module should define its own config model. All modules access the same AtloopConfig model for type safety.

class atloop.config.ConfigLoader

Bases: object

Configuration loader - uses varlord for lib/api.

static get() → AtloopConfig

Get configuration - access from anywhere in lib/api.

Returns:

Loaded AtloopConfig instance (type-safe, validated against AtloopConfig model)

Raises:

• KeyError – If config not initialized (call setup() first)

• RequiredFieldError – If required fields missing (varlord validation)

• TypeError – If types don’t match model (varlord validation)

static get_atloop_dir() → Path

Get the found .atloop directory path.

static setup(atloop_dir: str | None = None) → Config

Setup configuration - call once at application startup.

Parameters:

atloop_dir – Custom atloop directory (for testing)

Returns:

Config instance (also registered globally)

class atloop.config.AtloopConfig(ai: ~atloop.config.models.AIConfig, sandbox: ~atloop.config.models.SandboxConfig = <factory>, memory: ~atloop.config.models.MemoryConfig = <factory>, limits: ~atloop.config.models.LimitsConfig = <factory>, runtime: ~atloop.config.models.RuntimeConfig = <factory>, debug: ~atloop.config.models.DebugConfig = <factory>, output: ~atloop.config.models.OutputConfig = <factory>)

Bases: object

Main atloop configuration.

__init__(ai: ~atloop.config.models.AIConfig, sandbox: ~atloop.config.models.SandboxConfig = <factory>, memory: ~atloop.config.models.MemoryConfig = <factory>, limits: ~atloop.config.models.LimitsConfig = <factory>, runtime: ~atloop.config.models.RuntimeConfig = <factory>, debug: ~atloop.config.models.DebugConfig = <factory>, output: ~atloop.config.models.OutputConfig = <factory>) → None

__post_init__()

Validate configuration.

ai: AIConfig

sandbox: SandboxConfig

memory: MemoryConfig

limits: LimitsConfig

runtime: RuntimeConfig

debug: DebugConfig

output: OutputConfig

ConfigLoader

class atloop.config.loader.ConfigLoader

Bases: object

Configuration loader - uses varlord for lib/api.

static setup(atloop_dir: str | None = None) → Config

Setup configuration - call once at application startup.

Parameters:

atloop_dir – Custom atloop directory (for testing)

Returns:

Config instance (also registered globally)

static get() → AtloopConfig

Get configuration - access from anywhere in lib/api.

Returns:

Loaded AtloopConfig instance (type-safe, validated against AtloopConfig model)

Raises:

• KeyError – If config not initialized (call setup() first)

• RequiredFieldError – If required fields missing (varlord validation)

• TypeError – If types don’t match model (varlord validation)

static get_atloop_dir() → Path

Get the found .atloop directory path.

Configuration Models

class atloop.config.models.AtloopConfig(ai: ~atloop.config.models.AIConfig, sandbox: ~atloop.config.models.SandboxConfig = <factory>, memory: ~atloop.config.models.MemoryConfig = <factory>, limits: ~atloop.config.models.LimitsConfig = <factory>, runtime: ~atloop.config.models.RuntimeConfig = <factory>, debug: ~atloop.config.models.DebugConfig = <factory>, output: ~atloop.config.models.OutputConfig = <factory>)

Bases: object

Main atloop configuration.

ai: AIConfig

sandbox: SandboxConfig

memory: MemoryConfig

limits: LimitsConfig

runtime: RuntimeConfig

debug: DebugConfig

output: OutputConfig

__post_init__()

Validate configuration.

__init__(ai: ~atloop.config.models.AIConfig, sandbox: ~atloop.config.models.SandboxConfig = <factory>, memory: ~atloop.config.models.MemoryConfig = <factory>, limits: ~atloop.config.models.LimitsConfig = <factory>, runtime: ~atloop.config.models.RuntimeConfig = <factory>, debug: ~atloop.config.models.DebugConfig = <factory>, output: ~atloop.config.models.OutputConfig = <factory>) → None

class atloop.config.models.TaskSpec(task_id: str, goal: str, workspace_root: str, budget: ~atloop.config.models.Budget = <factory>)

Bases: object

Task specification.

task_id: str

goal: str

workspace_root: str

budget: Budget

__post_init__()

Validate task specification.

__init__(task_id: str, goal: str, workspace_root: str, budget: ~atloop.config.models.Budget = <factory>) → None

class atloop.config.models.Budget(max_llm_calls: int = 80, max_tool_calls: int = 300, max_wall_time_sec: int = 1800)

Bases: object

Budget constraints for task execution.

max_llm_calls: int = 80

max_tool_calls: int = 300

max_wall_time_sec: int = 1800

to_dict() → dict

Convert Budget to dictionary.

__init__(max_llm_calls: int = 80, max_tool_calls: int = 300, max_wall_time_sec: int = 1800) → None

class atloop.config.models.SandboxConfig(base_url: str | None = None, local_test: bool = False, timeout: int = 30, session_ttl_seconds: int = 900, image: str | None = None, cpu_limit: str | None = None, memory_limit: str | None = None, ephemeral_storage_limit: str | None = None, default_session_id: str | None = None)

Bases: object

Sandbox execution configuration.

base_url: str | None = None

local_test: bool = False

timeout: int = 30

session_ttl_seconds: int = 900

image: str | None = None

cpu_limit: str | None = None

memory_limit: str | None = None

ephemeral_storage_limit: str | None = None

default_session_id: str | None = None

__post_init__()

Validate configuration.

__init__(base_url: str | None = None, local_test: bool = False, timeout: int = 30, session_ttl_seconds: int = 900, image: str | None = None, cpu_limit: str | None = None, memory_limit: str | None = None, ephemeral_storage_limit: str | None = None, default_session_id: str | None = None) → None