operations

Operations models for CrewMaster v2.0.0.

MODULE	DESCRIPTION
assemble_disclosure_test	Tests for assemble_disclosure integration.
operation	Operation model for CrewMaster v2.0.0.
operation_test	Tests for Operation model.
plans	Execution plan models for CrewMaster v2.0.0.
plans_resolve_test	Tests for resolve_plan — the ExecutionPlan resolver.
plans_test	Tests for ExecutionPlan, PlanNode, and DisclosurePayload models.

CLASS	DESCRIPTION
Operation	A typed, recursive unit of work in the execution plan.
DisclosurePayload	The fully assembled disclosure for a single plan node.
ExecutionPlan	The resolved, immutable execution plan.
PlanBuildError	Raised when an ExecutionPlan cannot be built from an Operation tree.
PlanNode	A single node in the ExecutionPlan DAG.

FUNCTION	DESCRIPTION
resolve_plan	Resolve an Operation tree into an immutable ExecutionPlan.

__all__ module-attribute

__all__ = ['DisclosurePayload', 'ExecutionPlan', 'Operation', 'PlanBuildError', 'PlanNode', 'resolve_plan']

Operation

A typed, recursive unit of work in the execution plan.

Operations form the nodes of the ExecutionPlan DAG. Each operation declares what type it produces, what context types it consumes, which agent executes it, and what sub-operations it contains.

ATTRIBUTE	DESCRIPTION
name	Unique operation name within the plan. TYPE: str
produces	The type produced by this operation. TYPE: type[Any]
kind	Semantic category of the operation. TYPE: Literal['artifact', 'cognition', 'communication']
agent	The AgentConfig for the executing agent (optional for sub-ops). TYPE: Any | None
consumes	List of context types this operation requires from upstream. TYPE: list[type[Any]]
task_blocks	List of blocks:// URIs for task prompt blocks. TYPE: list[str]
sub_operations	Recursive list of child operations. TYPE: list[Operation]
collaboration	Optional CollaborationProtocol for multi-agent execution. TYPE: Any | None
max_retrieval_rounds	Max rounds of request_capability (0 = disabled). TYPE: int
required_slots	Names of context variables that must be present. TYPE: list[str]

METHOD	DESCRIPTION
validate_sub_operations_names_unique	Ensure sub-operation names are unique within this operation.

name instance-attribute

name: str

produces instance-attribute

produces: type[Any]

kind instance-attribute

kind: Literal['artifact', 'cognition', 'communication']

agent class-attribute instance-attribute

agent: Any | None = Field(default=None)

consumes class-attribute instance-attribute

consumes: list[type[Any]] = Field(default_factory=list)

task_blocks class-attribute instance-attribute

task_blocks: list[str] = Field(default_factory=list)

sub_operations class-attribute instance-attribute

sub_operations: list[Operation] = Field(default_factory=list)

collaboration class-attribute instance-attribute

collaboration: Any | None = Field(default=None)

max_retrieval_rounds class-attribute instance-attribute

max_retrieval_rounds: int = Field(default=0, ge=0)

required_slots class-attribute instance-attribute

required_slots: list[str] = Field(default_factory=list)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

validate_sub_operations_names_unique

validate_sub_operations_names_unique() -> 'Operation'

Ensure sub-operation names are unique within this operation.

DisclosurePayload

The fully assembled disclosure for a single plan node.

Contains everything an agent needs to execute its assigned node: identity blocks, task blocks, resolved context, resolved tools, and the rendered prompt text.

ATTRIBUTE	DESCRIPTION
identity_blocks	List of resolved identity PromptBlock ids. TYPE: list[str]
task_blocks	List of resolved task PromptBlock ids. TYPE: list[str]
resolved_context	The flattened context dictionary for template rendering. TYPE: dict[str, Any]
resolved_tools	The list of ToolSchema instances visible to this node. TYPE: list[Any]
rendered_prompt	The final rendered prompt text. .. deprecated:: 2.0.0 Rendering is now done in :func:crewmaster.api.execute via :class:PromptEngine. This field is kept for backward compatibility with pre-built plans that already contain rendered prompts. TYPE: str

identity_blocks class-attribute instance-attribute

identity_blocks: list[str] = Field(default_factory=list)

task_blocks class-attribute instance-attribute

task_blocks: list[str] = Field(default_factory=list)

resolved_context class-attribute instance-attribute

resolved_context: dict[str, Any] = Field(default_factory=dict)

resolved_tools class-attribute instance-attribute

resolved_tools: list[Any] = Field(default_factory=list)

rendered_prompt class-attribute instance-attribute

rendered_prompt: str = ''

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

ExecutionPlan

The resolved, immutable execution plan.

Built from an Operation tree. The plan contains all nodes in topological order, their dependency edges, the artifact type registry, and per-node disclosure payloads.

ATTRIBUTE	DESCRIPTION
nodes	All PlanNodes in topological order. TYPE: list[PlanNode]
edges	Dependency edges as (from_node, to_node) tuples. TYPE: list[tuple[str, str]]
artifacts	Mapping of operation name to output type. TYPE: dict[str, type[Any]]
disclosure	Mapping of operation name to DisclosurePayload. TYPE: dict[str, DisclosurePayload]

nodes class-attribute instance-attribute

nodes: list[PlanNode] = Field(default_factory=list)

edges class-attribute instance-attribute

edges: list[tuple[str, str]] = Field(default_factory=list)

artifacts class-attribute instance-attribute

artifacts: dict[str, type[Any]] = Field(default_factory=dict)

disclosure class-attribute instance-attribute

disclosure: dict[str, DisclosurePayload] = Field(default_factory=dict)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

PlanBuildError

Raised when an ExecutionPlan cannot be built from an Operation tree.

PlanNode

A single node in the ExecutionPlan DAG.

Each PlanNode corresponds to one Operation and carries its agent, runtime driver, disclosure payload, and dependency information.

ATTRIBUTE	DESCRIPTION
operation_name	The name of the Operation this node executes. TYPE: str
agent	The AgentConfig for this node (resolved from Operation or defaults). TYPE: Any | None
runtime	The RuntimeDriver instance for this node. TYPE: Any | None
disclosure	The assembled DisclosurePayload. TYPE: DisclosurePayload
produces	The type this node produces as output. TYPE: type[Any] | None
depends_on	List of operation names this node depends on. TYPE: list[str]

operation_name instance-attribute

operation_name: str

agent class-attribute instance-attribute

agent: Any | None = Field(default=None)

runtime class-attribute instance-attribute

runtime: Any | None = Field(default=None)

disclosure class-attribute instance-attribute

disclosure: DisclosurePayload = Field(default_factory=DisclosurePayload)

produces class-attribute instance-attribute

produces: type[Any] | None = Field(default=None)

depends_on class-attribute instance-attribute

depends_on: list[str] = Field(default_factory=list)

max_retrieval_rounds class-attribute instance-attribute

max_retrieval_rounds: int = Field(default=0, ge=0)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

resolve_plan

resolve_plan(operation: Operation, context_store: ContextStore, block_store: BlockStore, default_runtime: Any | None = None, tool_registry: ToolRegistry | None = None) -> ExecutionPlan

Resolve an Operation tree into an immutable ExecutionPlan.

Walks the Operation tree to collect leaf operations (those with an agent assigned), resolves consumes dependencies by type — checking both upstream node artifacts and the ContextStore for domain projectors — topologically sorts the resulting nodes, assembles per-node disclosure payloads, and returns a complete ExecutionPlan.

Dependency resolution rules:

1. If a consumed type is a :class:ContextProjector, resolve it via the context_store. The projected context becomes part of the ctx namespace.
2. If a consumed type matches the produces of another leaf operation in the same tree, it is an artifact dependency — the downstream node depends on the upstream node.
3. If neither matches, the plan fails with a PlanBuildError.

PARAMETER	DESCRIPTION
operation	The root Operation to resolve. TYPE: Operation
context_store	The ContextStore holding domain instances for ContextProjector resolution. TYPE: ContextStore
block_store	The BlockStore for resolving blocks:// URIs during disclosure assembly. TYPE: BlockStore
default_runtime	Fallback RuntimeDriver when an AgentConfig does not specify its own. TYPE: Any | None DEFAULT: None
tool_registry	Optional ToolRegistry for populating resolved_tools in each node's disclosure. TYPE: ToolRegistry | None DEFAULT: None

RETURNS	DESCRIPTION
ExecutionPlan	A fully resolved ExecutionPlan with nodes in topological order.

RAISES	DESCRIPTION
PlanBuildError	If any consumes cannot be resolved, if there is a cycle in dependencies, or if no leaf operations are found.

assemble_disclosure_test

Tests for assemble_disclosure integration.

CLASS	DESCRIPTION
FakeOutput
TestAssembleDisclosure	Tests for assemble_disclosure: Operation + AgentConfig -> DisclosurePayload.

FakeOutput

TestAssembleDisclosure

Tests for assemble_disclosure: Operation + AgentConfig -> DisclosurePayload.

METHOD	DESCRIPTION
test_combined_prompt_with_identity_and_task_blocks	Disclosure should contain both identity and task blocks rendered.
test_rendered_prompt_includes_both_block_types	The rendered prompt should combine identity + task blocks.
test_ctx_variables_rendered_correctly	ctx.* variables should be rendered in the prompt.
test_deps_variables_rendered_correctly	deps.* variables should be rendered in the prompt.
test_cfg_variables_rendered_correctly	cfg.* variables should be rendered in the prompt.
test_missing_ctx_requires_raises_descriptive_error	Fails with descriptive error when a ctx.* require is missing.
test_missing_deps_requires_raises_descriptive_error	Fails with descriptive error when a deps.* require is missing.
test_missing_requires_on_identity_block	Validation should also check requires on identity blocks.
test_requires_satisfied_on_both_block_types	Validation should pass when requires are satisfied in all blocks.
test_missing_required_slot_raises_error	Fails when Operation required_slots are not covered by block provides.
test_required_slots_satisfied_by_task_block	Should pass when a task block provides the required slot.
test_required_slots_satisfied_by_identity_block	Should pass when an identity block provides the required slot.
test_no_blocks_returns_empty_payload	When there are no blocks, returns empty disclosure.
test_no_required_slots_passes	Empty required_slots should pass validation.
test_resolved_context_preserved_in_payload	The context dict should be preserved in the DisclosurePayload.

test_combined_prompt_with_identity_and_task_blocks

test_combined_prompt_with_identity_and_task_blocks()

Disclosure should contain both identity and task blocks rendered.

test_rendered_prompt_includes_both_block_types

test_rendered_prompt_includes_both_block_types()

The rendered prompt should combine identity + task blocks.

test_ctx_variables_rendered_correctly

test_ctx_variables_rendered_correctly()

ctx.* variables should be rendered in the prompt.

test_deps_variables_rendered_correctly

test_deps_variables_rendered_correctly()

deps.* variables should be rendered in the prompt.

test_cfg_variables_rendered_correctly

test_cfg_variables_rendered_correctly()

cfg.* variables should be rendered in the prompt.

test_missing_ctx_requires_raises_descriptive_error

test_missing_ctx_requires_raises_descriptive_error()

Fails with descriptive error when a ctx.* require is missing.

test_missing_deps_requires_raises_descriptive_error

test_missing_deps_requires_raises_descriptive_error()

Fails with descriptive error when a deps.* require is missing.

test_missing_requires_on_identity_block

test_missing_requires_on_identity_block()

Validation should also check requires on identity blocks.

test_requires_satisfied_on_both_block_types

test_requires_satisfied_on_both_block_types()

Validation should pass when requires are satisfied in all blocks.

test_missing_required_slot_raises_error

test_missing_required_slot_raises_error()

Fails when Operation required_slots are not covered by block provides.

test_required_slots_satisfied_by_task_block

test_required_slots_satisfied_by_task_block()

Should pass when a task block provides the required slot.

test_required_slots_satisfied_by_identity_block

test_required_slots_satisfied_by_identity_block()

Should pass when an identity block provides the required slot.

test_no_blocks_returns_empty_payload

test_no_blocks_returns_empty_payload()

When there are no blocks, returns empty disclosure.

test_no_required_slots_passes

test_no_required_slots_passes()

Empty required_slots should pass validation.

test_resolved_context_preserved_in_payload

test_resolved_context_preserved_in_payload()

The context dict should be preserved in the DisclosurePayload.

operation

Operation model for CrewMaster v2.0.0.

Operation is the fundamental unit of work — a typed, recursive task that declares what it produces, what it consumes, and how to execute it.

CLASS	DESCRIPTION
Operation	A typed, recursive unit of work in the execution plan.

Operation

A typed, recursive unit of work in the execution plan.

Operations form the nodes of the ExecutionPlan DAG. Each operation declares what type it produces, what context types it consumes, which agent executes it, and what sub-operations it contains.

ATTRIBUTE	DESCRIPTION
name	Unique operation name within the plan. TYPE: str
produces	The type produced by this operation. TYPE: type[Any]
kind	Semantic category of the operation. TYPE: Literal['artifact', 'cognition', 'communication']
agent	The AgentConfig for the executing agent (optional for sub-ops). TYPE: Any | None
consumes	List of context types this operation requires from upstream. TYPE: list[type[Any]]
task_blocks	List of blocks:// URIs for task prompt blocks. TYPE: list[str]
sub_operations	Recursive list of child operations. TYPE: list[Operation]
collaboration	Optional CollaborationProtocol for multi-agent execution. TYPE: Any | None
max_retrieval_rounds	Max rounds of request_capability (0 = disabled). TYPE: int
required_slots	Names of context variables that must be present. TYPE: list[str]

METHOD	DESCRIPTION
validate_sub_operations_names_unique	Ensure sub-operation names are unique within this operation.

name instance-attribute

name: str

produces instance-attribute

produces: type[Any]

kind instance-attribute

kind: Literal['artifact', 'cognition', 'communication']

agent class-attribute instance-attribute

agent: Any | None = Field(default=None)

consumes class-attribute instance-attribute

consumes: list[type[Any]] = Field(default_factory=list)

task_blocks class-attribute instance-attribute

task_blocks: list[str] = Field(default_factory=list)

sub_operations class-attribute instance-attribute

sub_operations: list[Operation] = Field(default_factory=list)

collaboration class-attribute instance-attribute

collaboration: Any | None = Field(default=None)

max_retrieval_rounds class-attribute instance-attribute

max_retrieval_rounds: int = Field(default=0, ge=0)

required_slots class-attribute instance-attribute

required_slots: list[str] = Field(default_factory=list)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

validate_sub_operations_names_unique

validate_sub_operations_names_unique() -> 'Operation'

Ensure sub-operation names are unique within this operation.

operation_test

Tests for Operation model.

CLASS	DESCRIPTION
FakeOutput
FakeContext
TestOperation

FakeOutput

FakeContext

TestOperation

METHOD	DESCRIPTION
test_minimal_construction	Operation should construct with minimum required fields.
test_full_construction	Operation should accept all fields.
test_name_is_required	Operation should require a name.
test_produces_is_required	Operation should require a produces type.
test_kind_is_required	Operation should require a kind.
test_kind_rejects_invalid_values	Operation should reject invalid kind values.
test_kind_accepts_all_valid_values	Operation should accept all three kind values.
test_sub_operations_recursion	Operation should accept recursive sub_operations.
test_max_retrieval_rounds_non_negative	Operation should reject negative max_retrieval_rounds.
test_default_max_retrieval_rounds_is_zero	Operation should default max_retrieval_rounds to 0.
test_duplicate_sub_operation_names_rejected	Operation should reject duplicate names in sub_operations.
test_serialization_roundtrip	Operation should serialize and deserialize.

test_minimal_construction

test_minimal_construction()

Operation should construct with minimum required fields.

test_full_construction

test_full_construction()

Operation should accept all fields.

test_name_is_required

test_name_is_required()

Operation should require a name.

test_produces_is_required

test_produces_is_required()

Operation should require a produces type.

test_kind_is_required

test_kind_is_required()

Operation should require a kind.

test_kind_rejects_invalid_values

test_kind_rejects_invalid_values()

Operation should reject invalid kind values.

test_kind_accepts_all_valid_values

test_kind_accepts_all_valid_values()

Operation should accept all three kind values.

test_sub_operations_recursion

test_sub_operations_recursion()

Operation should accept recursive sub_operations.

test_max_retrieval_rounds_non_negative

test_max_retrieval_rounds_non_negative()

Operation should reject negative max_retrieval_rounds.

test_default_max_retrieval_rounds_is_zero

test_default_max_retrieval_rounds_is_zero()

Operation should default max_retrieval_rounds to 0.

test_duplicate_sub_operation_names_rejected

test_duplicate_sub_operation_names_rejected()

Operation should reject duplicate names in sub_operations.

test_serialization_roundtrip

test_serialization_roundtrip()

Operation should serialize and deserialize.

plans

Execution plan models for CrewMaster v2.0.0.

The ExecutionPlan is the resolved, immutable DAG that drives execution. It is built from an Operation tree by resolving dependencies, ordering nodes topologically, and assembling per-node disclosure payloads.

CLASS	DESCRIPTION
DisclosurePayload	The fully assembled disclosure for a single plan node.
PlanNode	A single node in the ExecutionPlan DAG.
ExecutionPlan	The resolved, immutable execution plan.
PlanBuildError	Raised when an ExecutionPlan cannot be built from an Operation tree.

FUNCTION	DESCRIPTION
assemble_disclosure	Assemble a DisclosurePayload from an Operation and AgentConfig.
resolve_plan	Resolve an Operation tree into an immutable ExecutionPlan.

DisclosurePayload

The fully assembled disclosure for a single plan node.

Contains everything an agent needs to execute its assigned node: identity blocks, task blocks, resolved context, resolved tools, and the rendered prompt text.

ATTRIBUTE	DESCRIPTION
identity_blocks	List of resolved identity PromptBlock ids. TYPE: list[str]
task_blocks	List of resolved task PromptBlock ids. TYPE: list[str]
resolved_context	The flattened context dictionary for template rendering. TYPE: dict[str, Any]
resolved_tools	The list of ToolSchema instances visible to this node. TYPE: list[Any]
rendered_prompt	The final rendered prompt text. .. deprecated:: 2.0.0 Rendering is now done in :func:crewmaster.api.execute via :class:PromptEngine. This field is kept for backward compatibility with pre-built plans that already contain rendered prompts. TYPE: str

identity_blocks class-attribute instance-attribute

identity_blocks: list[str] = Field(default_factory=list)

task_blocks class-attribute instance-attribute

task_blocks: list[str] = Field(default_factory=list)

resolved_context class-attribute instance-attribute

resolved_context: dict[str, Any] = Field(default_factory=dict)

resolved_tools class-attribute instance-attribute

resolved_tools: list[Any] = Field(default_factory=list)

rendered_prompt class-attribute instance-attribute

rendered_prompt: str = ''

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

PlanNode

A single node in the ExecutionPlan DAG.

Each PlanNode corresponds to one Operation and carries its agent, runtime driver, disclosure payload, and dependency information.

ATTRIBUTE	DESCRIPTION
operation_name	The name of the Operation this node executes. TYPE: str
agent	The AgentConfig for this node (resolved from Operation or defaults). TYPE: Any | None
runtime	The RuntimeDriver instance for this node. TYPE: Any | None
disclosure	The assembled DisclosurePayload. TYPE: DisclosurePayload
produces	The type this node produces as output. TYPE: type[Any] | None
depends_on	List of operation names this node depends on. TYPE: list[str]

operation_name instance-attribute

operation_name: str

agent class-attribute instance-attribute

agent: Any | None = Field(default=None)

runtime class-attribute instance-attribute

runtime: Any | None = Field(default=None)

disclosure class-attribute instance-attribute

disclosure: DisclosurePayload = Field(default_factory=DisclosurePayload)

produces class-attribute instance-attribute

produces: type[Any] | None = Field(default=None)

depends_on class-attribute instance-attribute

depends_on: list[str] = Field(default_factory=list)

max_retrieval_rounds class-attribute instance-attribute

max_retrieval_rounds: int = Field(default=0, ge=0)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

ExecutionPlan

The resolved, immutable execution plan.

Built from an Operation tree. The plan contains all nodes in topological order, their dependency edges, the artifact type registry, and per-node disclosure payloads.

ATTRIBUTE	DESCRIPTION
nodes	All PlanNodes in topological order. TYPE: list[PlanNode]
edges	Dependency edges as (from_node, to_node) tuples. TYPE: list[tuple[str, str]]
artifacts	Mapping of operation name to output type. TYPE: dict[str, type[Any]]
disclosure	Mapping of operation name to DisclosurePayload. TYPE: dict[str, DisclosurePayload]

nodes class-attribute instance-attribute

nodes: list[PlanNode] = Field(default_factory=list)

edges class-attribute instance-attribute

edges: list[tuple[str, str]] = Field(default_factory=list)

artifacts class-attribute instance-attribute

artifacts: dict[str, type[Any]] = Field(default_factory=dict)

disclosure class-attribute instance-attribute

disclosure: dict[str, DisclosurePayload] = Field(default_factory=dict)

model_config class-attribute instance-attribute

model_config = {'arbitrary_types_allowed': True}

PlanBuildError

Raised when an ExecutionPlan cannot be built from an Operation tree.

assemble_disclosure

assemble_disclosure(operation: Operation, agent: AgentConfig, context: dict[str, Any], block_store: BlockStore) -> DisclosurePayload

Assemble a DisclosurePayload from an Operation and AgentConfig.

Resolves the agent's identity blocks and the operation's task blocks through the block_store, validates that all frontmatter requires are satisfied by the available context variables, validates that all required_slots on the Operation are covered by block provides, renders all blocks via Jinja2, and returns a DisclosurePayload.

This is called at plan-build time. If any validation fails, the plan fails before any LLM call is made.

PARAMETER	DESCRIPTION
operation	The Operation being assembled into a disclosure. TYPE: Operation
agent	The AgentConfig that will execute this operation. TYPE: AgentConfig
context	Variables structured as {"ctx": {...}, "deps": {...}, "cfg": {...}} for template rendering. TYPE: dict[str, Any]
block_store	The BlockStore implementation for resolving blocks:// URIs. TYPE: BlockStore

RETURNS	DESCRIPTION
DisclosurePayload	A DisclosurePayload with identity blocks, task blocks, resolved
DisclosurePayload	context, and the fully rendered prompt.

RAISES	DESCRIPTION
ValueError	If any block require is unsatisfied by the context, or if any required_slot is not provided by any block.
FileNotFoundError	If a block URI cannot be resolved.

resolve_plan

resolve_plan(operation: Operation, context_store: ContextStore, block_store: BlockStore, default_runtime: Any | None = None, tool_registry: ToolRegistry | None = None) -> ExecutionPlan

Resolve an Operation tree into an immutable ExecutionPlan.

Walks the Operation tree to collect leaf operations (those with an agent assigned), resolves consumes dependencies by type — checking both upstream node artifacts and the ContextStore for domain projectors — topologically sorts the resulting nodes, assembles per-node disclosure payloads, and returns a complete ExecutionPlan.

Dependency resolution rules:

1. If a consumed type is a :class:ContextProjector, resolve it via the context_store. The projected context becomes part of the ctx namespace.
2. If a consumed type matches the produces of another leaf operation in the same tree, it is an artifact dependency — the downstream node depends on the upstream node.
3. If neither matches, the plan fails with a PlanBuildError.

PARAMETER	DESCRIPTION
operation	The root Operation to resolve. TYPE: Operation
context_store	The ContextStore holding domain instances for ContextProjector resolution. TYPE: ContextStore
block_store	The BlockStore for resolving blocks:// URIs during disclosure assembly. TYPE: BlockStore
default_runtime	Fallback RuntimeDriver when an AgentConfig does not specify its own. TYPE: Any | None DEFAULT: None
tool_registry	Optional ToolRegistry for populating resolved_tools in each node's disclosure. TYPE: ToolRegistry | None DEFAULT: None

RETURNS	DESCRIPTION
ExecutionPlan	A fully resolved ExecutionPlan with nodes in topological order.

RAISES	DESCRIPTION
PlanBuildError	If any consumes cannot be resolved, if there is a cycle in dependencies, or if no leaf operations are found.

plans_resolve_test

Tests for resolve_plan — the ExecutionPlan resolver.

Covers: - Walking the Operation tree and collecting leaf nodes - Resolving consumes by type (ContextProjectors and upstream artifacts) - Topological sort - Error on unresolved consumes - Error on cycles - Error on missing agents - Disclosure assembly per node - Runtime assignment (AgentConfig.runtime vs default_runtime) - Tool retrieval from ToolRegistry

CLASS	DESCRIPTION
Greeting
UserInfo
Colors
Typography
FinalDesign
InvalidOutput
User
UserNameProjector	Projects a User to context.
NotAProjector	Does NOT implement ContextProjector.
FakeOutput
FakeDriver	A fake RuntimeDriver that records calls and returns canned responses.
TestLeafCollection	Tests for leaf operation collection from the Operation tree.
TestConsumesResolution	Tests for resolving consumes via ContextStore and upstream artifacts.
TestTopologicalSort	Tests for topological ordering of plan nodes.
TestRuntimeAssignment	Tests for runtime driver assignment to each node.
TestToolRegistryIntegration	Tests for tool retrieval during plan building.
TestPlanStructure	Tests for ExecutionPlan structure correctness.
TestErrorMessages	Tests for descriptive error messages.

Greeting

UserInfo

Colors

Typography

FinalDesign

InvalidOutput

User

User(name: str)

ATTRIBUTE	DESCRIPTION
name

name instance-attribute

name = name

UserNameProjector

UserNameProjector(user_name: str)

Projects a User to context.

METHOD	DESCRIPTION
from_domain
to_context

from_domain classmethod

from_domain(data: User) -> 'UserNameProjector'

to_context

to_context() -> dict

NotAProjector

Does NOT implement ContextProjector.

FakeOutput

FakeDriver

FakeDriver(output: object = None)

A fake RuntimeDriver that records calls and returns canned responses.

METHOD	DESCRIPTION
execute
astream

ATTRIBUTE	DESCRIPTION
calls	TYPE: list[RuntimeRequest]

calls instance-attribute

calls: list[RuntimeRequest] = []

execute async

execute(request: RuntimeRequest) -> RuntimeResponse

astream async

astream(request: RuntimeRequest)

TestLeafCollection

Tests for leaf operation collection from the Operation tree.

METHOD	DESCRIPTION
test_single_leaf_root_with_agent	Root with agent → single leaf.
test_sub_operations_as_leaves	Container root with sub_operations → each sub-op is a leaf.
test_deeply_nested_sub_operations	Three-level nested operations → collect only those with agents.
test_no_agent_in_tree_raises	A tree with no agents at all should fail.

test_single_leaf_root_with_agent

test_single_leaf_root_with_agent()

Root with agent → single leaf.

test_sub_operations_as_leaves

test_sub_operations_as_leaves()

Container root with sub_operations → each sub-op is a leaf.

test_deeply_nested_sub_operations

test_deeply_nested_sub_operations()

Three-level nested operations → collect only those with agents.

test_no_agent_in_tree_raises

test_no_agent_in_tree_raises()

A tree with no agents at all should fail.

TestConsumesResolution

Tests for resolving consumes via ContextStore and upstream artifacts.

METHOD	DESCRIPTION
test_context_projector_resolved	consumes=[UserNameProjector] → resolved via ContextStore.
test_upstream_artifact_dependency	consumes=[Colors] where upstream node produces Colors → dependency.
test_mixed_consumes_projectors_and_artifacts	consumes=[UserNameProjector, Colors] → both resolved.
test_consumes_not_resolvable_raises	A consumed type that is neither a projector nor produced → error.
test_self_consuming_raises	Operation consuming its own output type → error.

test_context_projector_resolved

test_context_projector_resolved()

consumes=[UserNameProjector] → resolved via ContextStore.

test_upstream_artifact_dependency

test_upstream_artifact_dependency()

consumes=[Colors] where upstream node produces Colors → dependency.

test_mixed_consumes_projectors_and_artifacts

test_mixed_consumes_projectors_and_artifacts()

consumes=[UserNameProjector, Colors] → both resolved.

test_consumes_not_resolvable_raises

test_consumes_not_resolvable_raises()

A consumed type that is neither a projector nor produced → error.

test_self_consuming_raises

test_self_consuming_raises()

Operation consuming its own output type → error.

TestTopologicalSort

Tests for topological ordering of plan nodes.

METHOD	DESCRIPTION
test_linear_chain	A → B → C linear chain should order correctly.
test_diamond_dependency	step1 → step2, step1 → step3 → step4 (diamond) should work.
test_cycle_detection	Cyclic dependencies should be detected and fail.
test_independent_nodes_any_order	Two independent nodes → both should appear, no dependency edges.

test_linear_chain

test_linear_chain()

A → B → C linear chain should order correctly.

test_diamond_dependency

test_diamond_dependency()

step1 → step2, step1 → step3 → step4 (diamond) should work.

test_cycle_detection

test_cycle_detection()

Cyclic dependencies should be detected and fail.

test_independent_nodes_any_order

test_independent_nodes_any_order()

Two independent nodes → both should appear, no dependency edges.

TestRuntimeAssignment

Tests for runtime driver assignment to each node.

METHOD	DESCRIPTION
test_agent_runtime_overrides_default	AgentConfig.runtime should be used when set, not default_runtime.
test_fallback_to_default_runtime	When AgentConfig has no runtime, use default_runtime.
test_missing_runtime_raises	No runtime anywhere → error.

test_agent_runtime_overrides_default

test_agent_runtime_overrides_default()

AgentConfig.runtime should be used when set, not default_runtime.

test_fallback_to_default_runtime

test_fallback_to_default_runtime()

When AgentConfig has no runtime, use default_runtime.

test_missing_runtime_raises

test_missing_runtime_raises()

No runtime anywhere → error.

TestToolRegistryIntegration

Tests for tool retrieval during plan building.

METHOD	DESCRIPTION
test_tools_retrieved_for_agent_scope	Agent with default_tool_scope → tools should appear in disclosure.
test_no_tools_without_registry	No ToolRegistry → resolved_tools should be empty.
test_global_scope_tools_visible	Tools registered under '*' should be visible to any scope.

test_tools_retrieved_for_agent_scope

test_tools_retrieved_for_agent_scope()

Agent with default_tool_scope → tools should appear in disclosure.

test_no_tools_without_registry

test_no_tools_without_registry()

No ToolRegistry → resolved_tools should be empty.

test_global_scope_tools_visible

test_global_scope_tools_visible()

Tools registered under '*' should be visible to any scope.

TestPlanStructure

Tests for ExecutionPlan structure correctness.

METHOD	DESCRIPTION
test_plan_has_nodes_edges_artifacts_disclosure	Plan should have all top-level fields populated.
test_disclosure_rendered_prompt_empty_after_resolve	After resolve_plan, rendered_prompt is empty — rendering moved to api.execute().

test_plan_has_nodes_edges_artifacts_disclosure

test_plan_has_nodes_edges_artifacts_disclosure()

Plan should have all top-level fields populated.

test_disclosure_rendered_prompt_empty_after_resolve

test_disclosure_rendered_prompt_empty_after_resolve()

After resolve_plan, rendered_prompt is empty — rendering moved to api.execute().

TestErrorMessages

Tests for descriptive error messages.

METHOD	DESCRIPTION
test_duplicate_operation_names	Duplicate leaf operation names → descriptive error.
test_duplicate_produces_types	Duplicate produces types → descriptive error.

test_duplicate_operation_names

test_duplicate_operation_names()

Duplicate leaf operation names → descriptive error.

test_duplicate_produces_types

test_duplicate_produces_types()

Duplicate produces types → descriptive error.

plans_test

Tests for ExecutionPlan, PlanNode, and DisclosurePayload models.

CLASS	DESCRIPTION
FakeOutput
TestDisclosurePayload
TestPlanNode
TestExecutionPlan

FakeOutput

TestDisclosurePayload

METHOD	DESCRIPTION
test_defaults	DisclosurePayload should have sensible defaults.
test_full_construction	DisclosurePayload should accept all fields.
test_serialization_roundtrip	DisclosurePayload should serialize and deserialize.

test_defaults

test_defaults()

DisclosurePayload should have sensible defaults.

test_full_construction

test_full_construction()

DisclosurePayload should accept all fields.

test_serialization_roundtrip

test_serialization_roundtrip()

DisclosurePayload should serialize and deserialize.

TestPlanNode

METHOD	DESCRIPTION
test_defaults	PlanNode should have sensible defaults.
test_operation_name_required	PlanNode should require operation_name.
test_full_construction	PlanNode should accept all fields.
test_serialization_roundtrip	PlanNode should serialize and deserialize.

test_defaults

test_defaults()

PlanNode should have sensible defaults.

test_operation_name_required

test_operation_name_required()

PlanNode should require operation_name.

test_full_construction

test_full_construction()

PlanNode should accept all fields.

test_serialization_roundtrip

test_serialization_roundtrip()

PlanNode should serialize and deserialize.

TestExecutionPlan

METHOD	DESCRIPTION
test_defaults	ExecutionPlan should have sensible defaults.
test_full_construction	ExecutionPlan should accept all fields.
test_serialization_roundtrip	ExecutionPlan should serialize and deserialize.

test_defaults

test_defaults()

ExecutionPlan should have sensible defaults.

test_full_construction

test_full_construction()

ExecutionPlan should accept all fields.

test_serialization_roundtrip

test_serialization_roundtrip()

ExecutionPlan should serialize and deserialize.