Configuration

Configuration values are loaded from pyproject.toml under [tool.auto_workflow] merged with defaults and overridable via environment variables prefixed with AUTO_WORKFLOW_.

Defaults

Key	Default	Description
log_level	INFO	Baseline logging level (middleware dependent)
max_dynamic_tasks	2048	Guardrail for dynamic expansion
artifact_store	memory	Artifact backend (memory/filesystem)
artifact_store_path	.aw_artifacts	Directory for file store
artifact_serializer	pickle	Serializer for filesystem artifacts (pickle/json)
result_cache	memory	Result cache backend
result_cache_path	.aw_cache	Directory for file cache
result_cache_max_entries	None	Optional LRU bound for in-memory cache
process_pool_max_workers	None	Max workers for process pool

Example pyproject.toml

[tool.auto_workflow]
artifact_store = "filesystem"
artifact_store_path = ".data/artifacts"
result_cache = "filesystem"
result_cache_path = ".data/cache"
max_dynamic_tasks = 4096

Environment Overrides

export AUTO_WORKFLOW_RESULT_CACHE=filesystem
export AUTO_WORKFLOW_RESULT_CACHE_PATH=/tmp/aw_cache
# Logging controls
export AUTO_WORKFLOW_DISABLE_STRUCTURED_LOGS=1
export AUTO_WORKFLOW_LOG_LEVEL=DEBUG

# Process pool tuning
export AUTO_WORKFLOW_PROCESS_POOL_MAX_WORKERS=8

# Cache bounding
export AUTO_WORKFLOW_RESULT_CACHE_MAX_ENTRIES=1000

Reloading Config

from auto_workflow.config import reload_config
reload_config()

Caches are cleared (using lru_cache invalidation).