How can I minimize costs for a low‑traffic Azure Cosmos DB Table API workload using the serverless option?

I am designing an application with intermittent, low‑traffic patterns (average‑to‑peak ratio < 10 %) that will store key‑value data in Azure Cosmos DB for Table. To keep expenses predictable and low, I plan to use a serverless account so that I pay only for the request units (RUs) actually consumed and for storage, without any minimum throughput commitment. However, I am uncertain how to estimate the RU consumption for my expected mix of point reads and queries, and how to configure monitoring and alerts to stay within a target monthly cost while respecting the single‑region limitation of serverless accounts.

What methods should I use to estimate RU usage for point reads versus queries in a serverless Table container? Which monitoring metrics and alert thresholds in Azure Portal or Azure Monitor are appropriate for tracking consumption and avoiding unexpected charges? Are there any serverless‑specific constraints (e.g., single‑region deployment) that could affect cost predictability for this workload?