Skip to main content

Python SDK

The package name is redpennon. It requires Python 3.12+ and uses httpx. Requests go to the production API at https://api.redpennon.dev (see DEFAULT_API_BASE_URL in the client module); you only configure the environment API key (and optional timeout).

Install​

When the package is on PyPI:

pip install redpennon

Configure​

from redpennon import Client, UserContext

with Client(api_key="YOUR_ENVIRONMENT_API_KEY") as client:
    ...

You can inject an httpx.Client (for example with httpx.MockTransport) so unit tests never call the real API.

Evaluate a variable​

variable_value — value or default​

enabled = client.variable_value(
    "show-banner",
    default=False,
    user=UserContext(id="user-123"),
)

default is keyword-only; user is a UserContext. Returns the served value, or default when value is None, the variable is unknown, or a transport/governance error occurs.

variable — full result​

result = client.variable("checkout-flow", user=UserContext(id="user-123"))
# result.value, result.variation, result.reason, result.feature,
# result.evaluation_trace  (str | None)

Returns a VariableResult dataclass with all response fields. Use when you need variation, reason, or evaluation_trace.

variables — batch​

results = client.variables(
    ["show-banner", "discount-pct"],
    user=UserContext(id="user-123"),
)
banner = results["show-banner"].value or False
discount = results["discount-pct"].value or 0

Resolves multiple variables in one round-trip. Returns dict[str, VariableResult].

VariableResult dataclass​

@dataclass
class VariableResult:
    key: str
    value: Any          # None when not served
    variation: str | None
    reason: str
    feature: str | None
    evaluation_trace: str | None = None  # opaque signed token — forward to track_events

Track events​

from redpennon import EventPayload, UserContext

# Evaluate first to capture the trace
user = UserContext(id="user-123")
result = client.variable("checkout-flow", user=user)

# Send the event, forwarding the trace for verified attribution
client.track_events([
    EventPayload(
        event="purchase",
        variable="checkout-flow",
        variation=result.variation,
        user=user,
        value=49.99,
        evaluation_trace=result.evaluation_trace,
    )
])

track_events POSTs to POST /v1/events and returns a TrackEventsResult(accepted=N). Raises APIError on non-202.

EventPayload dataclass​

@dataclass
class EventPayload:
    event: str           # required
    variable: str        # required
    variation: str       # required
    user: UserContext | None = None
    value: float | None = None
    occurred_at: str | None = None   # ISO-8601
    evaluation_trace: str | None = None  # from VariableResult.evaluation_trace

See Ingest events for full field semantics and error codes.

Errors​

Failed requests raise APIError with status_code and message. See API errors.

See also​