A tiny Flask app (app.py) that imports the SDK straight from this repo's source (no build or install step) so you can manually exercise Featureflow.evaluate() against a real Featureflow account from a browser.

python3 -m pip install -r examples/server/requirements.txt
FEATUREFLOW_SERVER_KEY=srv-env-<your key> python3 examples/server/app.py

Open http://127.0.0.1:3456/ -- or hit the JSON endpoints directly:

• GET /health
• GET /api/ready -- true once the 30s poll loop has fetched flags at least once
• GET /api/features -- keys of every feature currently loaded from the API
• GET /api/evaluate?feature=<key>&userId=<id>&role=&tier=
• GET /api/evaluate-all?userId=<id>&role=&tier=

PORT overrides the default 3456.

• This SDK has no baseUrl override (unlike the Node SDK) -- it always polls and reports events against https://app.featureflow.io. There's no way to point this example at staging, and no "disable events" switch either: every /api/evaluate* request here sends a real evaluation event to your account.
• The app reads client._features (a private attribute) to list known feature keys and drive "evaluate all" -- the SDK doesn't expose a public accessor for the loaded feature set yet.
• Polling starts the moment Featureflow(api_key) is constructed, on a background thread, every 30s. If the key is wrong, the SDK prints PollingClient got error from API code ... to stderr and /api/ready stays false forever.