Note: orchestration

Context

The src/attune/orchestration/ module covers three related concerns: dynamic agent team composition, execution strategy selection, and workflow nesting. These are the building blocks that meta-orchestration patterns (Sequential, Parallel, Debate, Teaching, Refinement, Adaptive) are built on top of.

How the package surface is structured

The module exposes two complementary layers that are designed to compose with each other.

Execution strategy classes (defined under _strategies/) each extend ExecutionStrategy and implement a single execute(agents, context) -> StrategyResult signature:

Class	Where	What it does
`ExecutionStrategy`	`_strategies/base.py`	Abstract base for all composition strategies
`ToolEnhancedStrategy`	`_strategies/advanced_strategies.py`	Runs a single agent with a full tool bundle
`PromptCachedSequentialStrategy`	`_strategies/advanced_strategies.py`	Sequential execution with a shared cached context (default TTL: 3600 s)
`DelegationChainStrategy`	`_strategies/advanced_strategies.py`	Hierarchical delegation with a configurable max depth (default: 3)
`ConditionalStrategy`	`_strategies/conditional_strategies.py`	Branches on a single condition (if/else)
`MultiConditionalStrategy`	`_strategies/conditional_strategies.py`	Branches on multiple conditions (switch/case)
`NestedStrategy`	`_strategies/nesting.py`	Executes a referenced workflow as a nested unit
`NestedSequentialStrategy`	`_strategies/nesting.py`	Sequential steps where each step can be an agent or a workflow reference

Registry functions provide name-based lookup and registration so strategies and workflows can be composed without direct imports:

Function	Source	What it does
`get_strategy(name)`	`_strategies/__init__.py`	Returns a strategy instance by name; raises `ValueError` for unknown names
`register_strategy(name, cls)`	`_strategies/__init__.py`	Adds a custom `ExecutionStrategy` subclass to the registry
`register_workflow(workflow)`	`_strategies/nesting.py`	Makes a `WorkflowDefinition` available for nested references
`get_workflow(workflow_id)`	`_strategies/nesting.py`	Retrieves a registered workflow; raises `ValueError` if not found
`get_template(template_id)`	`agent_templates/registry.py`	Retrieves an `AgentTemplate` by ID
`register_custom_template(template)`	`agent_templates/registry.py`	Adds a user-defined template at runtime
`unregister_template(template_id)`	`agent_templates/registry.py`	Removes a template; returns `False` if the ID is not found

The pattern throughout is consistent: strategy classes hold execution logic, and the registry functions let MetaOrchestrator wire them together by name at runtime.

Source files

src/attune/orchestration/**
src/attune/coordination/**

Tags: orchestration, teams