QUOTA_EXCEEDED is not mapped to TooManyRequestsError, so rate limiting arrives untyped

response_handler.py の3つのエラーマップと check_response から始め、特に60行目付近にある既存の TooManyRequestsError のマッピングを確認します。Issue にある FakeResp のケースを再現し、QUOTA_EXCEEDED レスポンスで汎用の OverkizError ではなく TooManyRequestsError が発生することを確認します。実際のレート制限レスポンスが、想定されるバックオフパスと一貫した型になることを確認できれば完了です。

🐛 The problem

The Overkiz cloud reports its polling quota with errorCode: QUOTA_EXCEEDED, which matches none of the three error maps in response_handler.py, so it falls through to the generic raise at the end of check_response:

pyoverkiz.exceptions.OverkizError: {'errorCode': 'QUOTA_EXCEEDED', 'error': 'Too many requests, try again later'}

TooManyRequestsError is only reachable for a different errorCode (response_handler.py:60):

("AUTHENTICATION_ERROR", "Too many requests", TooManyRequestsError),

So the one exception type that means "you are being rate limited" is never raised for the server's actual rate-limit response.

🔁 Reproduction

import asyncio
from pyoverkiz.response_handler import check_response
from pyoverkiz.exceptions import OverkizError, TooManyRequestsError

class FakeResp:
    status = 403
    url = "https://example/events/x/fetch"
    async def json(self, content_type=None):
        return {"errorCode": "QUOTA_EXCEEDED",
                "error": "Too many requests, try again later"}

async def main():
    try:
        await check_response(FakeResp())
    except TooManyRequestsError:
        print("typed")
    except OverkizError as err:
        print("untyped:", err)

asyncio.run(main())   # -> untyped: {'errorCode': 'QUOTA_EXCEEDED', ...}

Why none of the maps catch it:

• _ERROR_CODE_MESSAGE_MAP (:47) — has no QUOTA_EXCEEDED entry
• _MESSAGE_FALLBACK_MAP (:104) — has "Too many executions" and "Too many attempts with an invalid token", but not "Too many requests"
• _ERROR_CODE_FALLBACK_MAP (:121) — only AUTHENTICATION_ERROR and RESOURCE_ACCESS_DENIED

📉 Why it matters

Callers cannot distinguish rate limiting from an unknown server error without string-matching the payload, so any back-off keyed on TooManyRequestsError silently never runs.

Measured on an Atlantic Cozytouch V2 hub (cloud API), eight commands issued in parallel to eight heaters, Home Assistant's overkiz integration:

Because the error arrives untyped:

• it surfaces as Unexpected error fetching device events data with a full traceback, rather than a handled rate-limit path
• the coordinator keeps polling at the interval that hit the limit instead of slowing down
• the failed update marks every entity unavailable, and Home Assistant then refuses the remaining service calls — Referenced entities ... are missing or not currently available — so three commands were silently dropped
• worst state-delivery lag in that run was 28.97 s

✅ Suggested fix

Either map the code directly in _ERROR_CODE_MESSAGE_MAP:

("QUOTA_EXCEEDED", None, TooManyRequestsError),

…or, more robustly, add the message to _MESSAGE_FALLBACK_MAP so it matches whatever errorCode carries it:

("Too many requests", TooManyRequestsError),

The second also covers the existing AUTHENTICATION_ERROR variant, and any future code the cloud attaches to the same message.

🧾 Environment

• pyoverkiz 2.1.0
• Hub: Atlantic Cozytouch V2 (io://****-****-9776), cloud API — local API not available for this server
• Seen on the events/{listener_id}/fetch endpoint, documented as "1 calls per 1 SECONDS period for this particular operation (polling)"

Happy to open a PR for whichever form you prefer.