Skip to main content

Python API

Generator​

from leadforge.api import Generator

Generator.from_recipe(recipe_id, *, seed, exposure_mode, **kwargs)​

Create a generator bound to a recipe.

gen = Generator.from_recipe(
    "b2b_saas_procurement_v1",
    seed=42,
    exposure_mode="student_public",   # or "research_instructor"
)
ParameterTypeDefault
recipe_idstrrequired
seedint42
exposure_modestr"student_public"

gen.generate(*, n_leads, difficulty, n_accounts, n_contacts)​

Run the full generation pipeline and return a WorldBundle.

bundle = gen.generate(
    n_leads=5000,
    difficulty="intermediate",   # "intro" / "intermediate" / "advanced"
    n_accounts=1500,             # optional — overrides recipe default
    n_contacts=4200,             # optional — overrides recipe default
)

gen.world_spec​

Access the WorldSpec (narrative, config, hidden structure) after construction.

gen = Generator.from_recipe("b2b_saas_procurement_v1", seed=42)
print(gen.world_spec.narrative.company.name)  # "Veridian Technologies"

WorldBundle​

Returned by gen.generate().

bundle.save(path)​

Write the full bundle to disk at path.

bundle.save("./out/bundle")

bundle.to_dataframes()​

Return all 9 relational tables as a dict[str, pd.DataFrame].

tables = bundle.to_dataframes()
leads = tables["leads"]

bundle.task_splits​

Access the train/valid/test DataFrames directly:

train = bundle.task_splits["converted_within_90_days"]["train"]

list_recipes()​

from leadforge.api import list_recipes

recipes = list_recipes()
print(recipes)  # ["b2b_saas_procurement_v1", ...]