FlatRatePricingOption.Parameters is a shared grab-bag type across all pricing models

[bokelley]

Summary

When constructing a FlatRatePricingOption with custom parameters, the parameters field accepts a Parameters type that contains fields from every pricing model combined:

from adcp.types import FlatRatePricingOption

po = FlatRatePricingOption(
    pricing_option_id="newsletter_weekly",
    pricing_model="flat_rate",
    fixed_price=250.00,
    currency="USD",
    parameters={"unit": "week", "sends_per_week": 5},
)

# The validated parameters object contains unrelated fields:
print(po.parameters)
# Parameters(
#     daypart=None,
#     duration_hours=None,
#     estimated_impressions=None,
#     loop_duration_seconds=None,     # DOOH field
#     min_plays_per_hour=None,        # DOOH field
#     sov_percentage=None,            # DOOH field
#     venue_package=None,             # DOOH field
#     unit='week',
#     sends_per_week=5,
# )

Problem

1. Autocomplete is misleading — IDE suggests sov_percentage and venue_package on a flat-rate email newsletter pricing option
2. No validation — You can set loop_duration_seconds on a flat-rate option and it silently accepts it
3. Schema is confusing — A consumer reading the schema can't tell which parameters apply to which pricing model

Suggested Fix

Discriminated parameter types per pricing model:

class FlatRateParameters(BaseModel):
    unit: str  # "week", "month", "event", "quarter"
    # Other flat-rate-specific fields

class DoohParameters(BaseModel):
    loop_duration_seconds: int | None = None
    min_plays_per_hour: int | None = None
    sov_percentage: float | None = None
    venue_package: str | None = None

class TimeBasedParameters(BaseModel):
    duration_hours: int | None = None
    daypart: str | None = None

This is likely a spec-level change (affects pricing-option.json), but the Python library could add validation even if the shared schema stays flat — e.g., a model validator that warns/strips inapplicable fields based on pricing_model.

Impact

Low-to-medium. This is more of a code quality / developer confusion issue than a blocker. But it makes the library feel auto-generated rather than designed, which hurts trust when building against it.

Environment

• adcp: 3.6.0
• Python: 3.12

[bokelley]

Moved upstream: adcontextprotocol/adcp#1373 — this is a spec-level concern (the Parameters schema is defined in the AdCP spec).