Add user stats caching and invalidation functionality

[tenkus47]

• Introduced caching for user statistics with a new timeout configuration.
• Implemented functions to set, get, and invalidate user stats cache in the daily log cache service.
• Updated daily log service to utilize the new caching mechanism for user stats.
• Modified accumulator service to invalidate user stats cache upon updates.
• Enhanced tests to cover new caching behavior and ensure proper functionality of user stats retrieval and invalidation.

[greptile-apps]

Confidence Score: 3/5

Needs one fix before merging: cache invalidation in delete_task_service fires before the DB transaction commits, which can leave stale stats cached for up to 5 minutes.

The cache invalidation in delete_task_service is placed inside the with SessionLocal() block. Because delete_task_service is a sync function that runs in a FastAPI thread-pool worker, asyncio.run() is called synchronously there — meaning the Redis DELETE completes before the surrounding DB transaction commits. A concurrent call to get_user_stats_service can race into that window, read pre-commit data from the DB, and cache it as the authoritative result. The rest of the caching logic (timer service, accumulator service, daily-log service, series-day-sync) is correctly ordered with invalidation happening after commits.

pecha_api/plans/users/plan_users_service.py — delete_task_service and check_day_completion both call schedule_invalidate_user_stats_cache inside an open DB session.

_{Reviews (2): Last reviewed commit: "Remove rate limiting functionality and r..." | Re-trigger Greptile}

[greptile-apps]

UserStatsResponse(**cache_data) will raise a Pydantic ValidationError if the cached dict doesn't conform to the current model schema — for example, after a deployment that adds a new required field while Redis still holds stale entries from the previous schema. The exception is not caught here, so it propagates through get_user_stats_service and surfaces as a 500 for every user whose stats are cached until all entries expire (up to 5 minutes). Returning None on validation failure causes a graceful cache miss and fresh DB computation instead.

Suggested change

	async def get_user_stats_cache(user_id: UUID) -> Optional[UserStatsResponse]:
	cache_key = _build_user_stats_cache_key(user_id=user_id)
	cache_data = await get_cache_data(hash_key=cache_key)
	if cache_data and isinstance(cache_data, dict):
	return UserStatsResponse(**cache_data)
	return None
	async def get_user_stats_cache(user_id: UUID) -> Optional[UserStatsResponse]:
	cache_key = _build_user_stats_cache_key(user_id=user_id)
	cache_data = await get_cache_data(hash_key=cache_key)
	if cache_data and isinstance(cache_data, dict):
	try:
	return UserStatsResponse(**cache_data)
	except Exception:
	logging.warning("Failed to deserialize user stats cache; treating as cache miss", exc_info=True)
	return None
	return None

[greptile-apps]

loop.create_task() returns a Task but the reference is immediately discarded. When the garbage collector eventually collects the task and it raised an exception, Python emits a "Task exception was never retrieved" warning rather than surfacing the error. Storing the returned task (or attaching a done-callback) ensures unhandled exceptions are at minimum logged.

Suggested change

	def schedule_invalidate_user_stats_cache(user_id: UUID) -> None:
	"""Invalidate stats cache from sync callers without blocking."""
	try:
	loop = asyncio.get_running_loop()
	loop.create_task(invalidate_user_stats_cache(user_id))
	except RuntimeError:
	asyncio.run(invalidate_user_stats_cache(user_id))
	def schedule_invalidate_user_stats_cache(user_id: UUID) -> None:
	"""Invalidate stats cache from sync callers without blocking."""
	try:
	loop = asyncio.get_running_loop()
	task = loop.create_task(invalidate_user_stats_cache(user_id))
	task.add_done_callback(
	lambda t: logging.error("Failed to invalidate user stats cache", exc_info=t.exception())
	if not t.cancelled() and t.exception() is not None
	else None
	)
	except RuntimeError:
	asyncio.run(invalidate_user_stats_cache(user_id))