Python SDK

feat-sdk (PyPI) is the Python SDK. Standard library only. Evaluates flags locally against a polled datafile.

Install

pip install feat-sdk

Python 3.10 or newer.

Usage

from feat import Client, ClientConfig, EvalContext

client = Client(ClientConfig(
    api_key="feat_sdk_...",
    data_plane_url="https://data.feat.so",
))
client.ready()

ctx = EvalContext(
    targeting_key="user-123",
    kinds={"user": {"plan": "pro", "email": "alice@example.com"}},
)

if client.get_boolean_value("checkout-v2", False, ctx):
    # ...
    pass

client.close()

Use a server key (feat_sdk_…).

API

enabled = client.get_boolean_value("checkout-v2", False, ctx)
variant = client.get_string_value("hero-copy", "control", ctx)
limit   = client.get_number_value("rate-limit", 100, ctx)
config  = client.get_object_value("layout-config", {}, ctx)

result  = client.evaluate("checkout-v2", False, ctx)
# result.value, result.variation_id, result.reason, result.error_message

reason is one of TARGETING_MATCH, SPLIT, FALLTHROUGH, DEFAULT, DISABLED, ERROR. See Evaluation order.

Lifecycle

Client(config) constructs and starts a background daemon thread that polls the datafile.
client.ready() blocks until the first datafile arrives.
client.close() stops the thread cleanly.

The daemon thread does not block process exit, but close() should still be called for orderly shutdown.

Configuration

ClientConfig(
    api_key="feat_sdk_...",
    data_plane_url="https://data.feat.so",
    poll_interval=30.0,    # seconds; default 30, minimum 5
    bootstrap=None,        # optional: a parsed datafile dict to skip the first fetch
)

Semantics

The Python SDK ports the same evaluation algorithm as @feathq/feat-eval bit-for-bit. The same datafile, the same context, and the same flag key always give the same result across Node.js, Go, Python, and Ruby.