flagsman 1.0.0

Flagsman server SDK: zero-dependency, in-process feature-flag evaluation with streaming updates.

Flagsman — Python server SDK

Zero-dependency, in-process feature-flag evaluation for Python. The SDK streams your environment's ruleset into memory and evaluates flags locally in microseconds — no network call on your hot path, no per-evaluation charges.

• Local evaluation — decisions are computed in-process from a streamed ruleset.
• Streaming updates — flag changes converge in well under a second (falls back to polling).
• Rigorous — the same decide() spec every Flagsman SDK implements, asserted by a shared conformance suite.
• No dependencies — pure standard library.

Install

pip install flagsman

Usage

from flagsman import FlagsmanClient

client = FlagsmanClient(
    base_url="https://api.flagsman.com",
    server_key="flags-server-...",   # your environment's server-side SDK key
)
client.start()

ctx = {"key": "user-123", "country": "US", "plan": "pro"}

# Boolean flag
if client.bool_variation("checkout.new-payment-flow", ctx, default=False):
    render_new_checkout()

# Multivariate flag (string / number / json)
color = client.variation("ui.button-color", ctx, default="blue")

# With evaluation reason (for debugging)
detail = client.variation_detail("ui.button-color", ctx, default="blue")

# Every flag for a context
flags = client.all_flags(ctx)

# Track an experiment goal / conversion
client.track("signup_completed", ctx)

client.stop()   # flushes buffered events and stops the stream

Configuration

FlagsmanClient(base_url, server_key, use_stream=True, poll_interval=30.0, request_timeout=10.0, send_events=False, flush_interval=5.0, event_batch_size=25)

• use_stream — stream the ruleset (recommended); set False to poll.
• send_events — send exposure/goal analytics events back to Flagsman.

Links

• Documentation: https://docs.flagsman.com
• Homepage: https://flagsman.com

MIT licensed.