"""add_serving_feature.py: your turn, add one production-serving feature.

The service (../solution/app.py) has config, probes, graceful lifespan, request ids, and is stateless.
Add ONE more thing real services need. Ideas, easy to hard:

  1. A /metrics endpoint: count requests and errors in a dict and return them (the basis for monitoring).
  2. Concurrency safety: make `handle` async and `await asyncio.sleep(0)` so one slow request does not
     block others (then explain workers vs async in your own words).
  3. A simple in-memory rate limit per client (X requests per minute) returning 429 when exceeded.
  4. Wire real per-session memory: load/save state by session_id from a store (M21), keeping the
     PROCESS stateless so replicas still scale.

Steps:
  1. Pick one and implement it in a copy of app.py.
  2. Add a TestClient check (copy the pattern from ../solution/demo.py).
  3. Run your check.
"""

print(__doc__)