Generators

class sim_panel.generators.GeneratorConfig(schema_version='0.1.0', seed=0, n_periods=1, policy=<factory>, selection=<factory>, execution=<factory>, outcome=None, validate_on_finish=True, max_errors=50, include_panelist_features_in_events=True, include_product_features_in_events=True, include_product_features_in_selection_prompt=True, event_namespace='sim_panel.v0', max_workers=1, prompting_strategy='persona', row_meta=<factory>)

Bases: object

Top-level generation config (YAML-governed in the CLI).

Generators orchestrate:

• policy exposure decisions

• panelist selection/evaluation calls

• outcome parsing/validation

• schema event construction

• optional schema validation

Notes

• schema_version is stored per row and validated against schema registry.

• seed governs RNG for exposure decisions and deterministic event ids.

Parameters:

• schema_version (str)

• seed (int)

• n_periods (int)

• policy (PolicyConfig)

• selection (SelectionConfig)

• execution (ExecutionConfig)

• outcome (OutcomeConfig | None)

• validate_on_finish (bool)

• max_errors (int)

• include_panelist_features_in_events (bool)

• include_product_features_in_events (bool)

• include_product_features_in_selection_prompt (bool)

• event_namespace (str)

• max_workers (int)

• prompting_strategy (str)

• row_meta (Dict[str, Any])

schema_version: str = '0.1.0'

seed: int = 0

n_periods: int = 1

policy: PolicyConfig

selection: SelectionConfig

execution: ExecutionConfig

outcome: OutcomeConfig | None = None

validate_on_finish: bool = True

max_errors: int = 50

include_panelist_features_in_events: bool = True

include_product_features_in_events: bool = True

include_product_features_in_selection_prompt: bool = True

event_namespace: str = 'sim_panel.v0'

max_workers: int = 1

prompting_strategy: str = 'persona'

row_meta: Dict[str, Any]

class sim_panel.generators.ExecutionConfig(rules=ExecutionRules(enforce_subset_of_choice_set=True, max_evals_per_panelist_per_t=None, allow_empty=True, keep_strategy='keep_first'))

Bases: object

Generator-side operational rules (NOT panelist constraints). This controls how many evaluations we actually execute after parsing selection.

Parameters:

rules (ExecutionRules)

rules: ExecutionRules = ExecutionRules(enforce_subset_of_choice_set=True, max_evals_per_panelist_per_t=None, allow_empty=True, keep_strategy='keep_first')

class sim_panel.generators.EventGenerator(cfg)

Bases: object

Orchestrates policy exposure + panelist selection/evaluation + outcomes into schema rows.

Expectations:

• panelists: Sequence[Panelist] runtime agents (LLM-capable)

• products: Sequence[Product] runtime wrappers (record + display())

Parameters:

cfg (GeneratorConfig)

cfg: GeneratorConfig

generate(*, panelists, products, progress=True, resume_from_period=0, prior_rows=None, on_period_complete=None)

Return type:

List[Dict[str, Any]]

Parameters:

• panelists (Sequence[Panelist])

• products (Sequence[Product])

• progress (bool)

• resume_from_period (int)

• prior_rows (List[Dict[str, Any]] | None)

• on_period_complete (Callable[[int, List[Dict[str, Any]]], None] | None)

class sim_panel.generators.pipeline.EventGenerator(cfg)

Bases: object

Orchestrates policy exposure + panelist selection/evaluation + outcomes into schema rows.

Expectations:

• panelists: Sequence[Panelist] runtime agents (LLM-capable)

• products: Sequence[Product] runtime wrappers (record + display())

Parameters:

cfg (GeneratorConfig)

cfg: GeneratorConfig

generate(*, panelists, products, progress=True, resume_from_period=0, prior_rows=None, on_period_complete=None)

Return type:

List[Dict[str, Any]]

Parameters:

• panelists (Sequence[Panelist])

• products (Sequence[Product])

• progress (bool)

• resume_from_period (int)

• prior_rows (List[Dict[str, Any]] | None)

• on_period_complete (Callable[[int, List[Dict[str, Any]]], None] | None)

class sim_panel.generators.types.ExecutionConfig(rules=ExecutionRules(enforce_subset_of_choice_set=True, max_evals_per_panelist_per_t=None, allow_empty=True, keep_strategy='keep_first'))

Bases: object

Generator-side operational rules (NOT panelist constraints). This controls how many evaluations we actually execute after parsing selection.

Parameters:

rules (ExecutionRules)

rules: ExecutionRules = ExecutionRules(enforce_subset_of_choice_set=True, max_evals_per_panelist_per_t=None, allow_empty=True, keep_strategy='keep_first')

class sim_panel.generators.types.GeneratorConfig(schema_version='0.1.0', seed=0, n_periods=1, policy=<factory>, selection=<factory>, execution=<factory>, outcome=None, validate_on_finish=True, max_errors=50, include_panelist_features_in_events=True, include_product_features_in_events=True, include_product_features_in_selection_prompt=True, event_namespace='sim_panel.v0', max_workers=1, prompting_strategy='persona', row_meta=<factory>)

Bases: object

Top-level generation config (YAML-governed in the CLI).

Generators orchestrate:

• policy exposure decisions

• panelist selection/evaluation calls

• outcome parsing/validation

• schema event construction

• optional schema validation

Notes

• schema_version is stored per row and validated against schema registry.

• seed governs RNG for exposure decisions and deterministic event ids.

Parameters:

• schema_version (str)

• seed (int)

• n_periods (int)

• policy (PolicyConfig)

• selection (SelectionConfig)

• execution (ExecutionConfig)

• outcome (OutcomeConfig | None)

• validate_on_finish (bool)

• max_errors (int)

• include_panelist_features_in_events (bool)

• include_product_features_in_events (bool)

• include_product_features_in_selection_prompt (bool)

• event_namespace (str)

• max_workers (int)

• prompting_strategy (str)

• row_meta (Dict[str, Any])

schema_version: str = '0.1.0'

seed: int = 0

n_periods: int = 1

policy: PolicyConfig

selection: SelectionConfig

execution: ExecutionConfig

outcome: OutcomeConfig | None = None

validate_on_finish: bool = True

max_errors: int = 50

include_panelist_features_in_events: bool = True

include_product_features_in_events: bool = True

include_product_features_in_selection_prompt: bool = True

event_namespace: str = 'sim_panel.v0'

max_workers: int = 1

prompting_strategy: str = 'persona'

row_meta: Dict[str, Any]

sim_panel.generators.rng.make_rng(seed)

Return type:

Generator

Parameters:

seed (int)

sim_panel.generators.rng.stable_event_id(namespace, payload)

Deterministic event id from a namespace + canonicalized payload.

Return type:

str

Parameters:

• namespace (str)

• payload (Dict[str, Any])