Study

design_research_agents.study is the stable public facade for running agent participants from study orchestration code and normalizing their outputs into a shared envelope.

The older design_research_agents.integration module remains available for compatibility. New study runners should prefer the typed request object and facade exports documented here.

class design_research_agents.study.AgentRunRequest(*, agent_ref, prompt, request_id=None, dependencies=<factory>, agent_bindings=<factory>)

Typed request object for running one agent inside a study.

agent_ref

Public agent reference, executable object, or binding key.

Type:

Any

prompt

Prompt or problem-like payload for the run.

Type:

Any

request_id

Optional run identifier.

Type:

str | None

dependencies

Optional dependencies exposed to the agent.

Type:

collections.abc.Mapping[str, object]

agent_bindings

Optional mapping of binding keys to executable agents or factories.

Type:

collections.abc.Mapping[str, AgentBinding]

class design_research_agents.study.AgentExecutionEnvelope(output=<factory>, metrics=<factory>, events=<factory>, trace_refs=<factory>, metadata=<factory>)

Normalized execution envelope used by study-orchestration consumers.

class design_research_agents.study.StudyCondition(*, condition_id, label=None, metadata=<factory>)

Minimal public condition descriptor for study orchestration.

condition_id

Stable condition identifier.

Type:

str

label

Optional human-readable condition label.

Type:

str | None

metadata

Optional condition metadata for downstream analysis.

Type:

collections.abc.Mapping[str, object]

design_research_agents.study.execute_agent_request(request)

Execute one typed agent-run request.

Parameters:

request – Agent-run request to execute.

Returns:

Normalized execution envelope.

design_research_agents.study.execute_agent_run(agent_ref, *, prompt, request_id, dependencies, agent_bindings=None)

Execute one public agent reference through the stable prompt/dependencies contract.

design_research_agents.study.normalize_agent_execution(raw, *, request_id=None)

Normalize raw agent output into the study-facing execution envelope.