Your ShipStation data, where work happens.

您询问时可以信赖的状态

发货状态是夜间复制总是出错的一项，因为它是在该行写入之后发生变化的内容。

• 按 ID 查询特定货件或标签，系统会在返回答案之前将这些行与 ShipStation 进行比对，因此取消或作废的标签会立即显示出来。
• 一份涵盖数月历史记录的报告以 SQL 的速度读取同步副本，而 ShipStation 端无需任何费用。
• ShipStation does not delete anything — a cancelled shipment stays as a row with its status changed, which is what you want when the question is how many were cancelled.

这15张桌子

• Movement — shipments, labels, fulfillments, manifests, pickups, batches.
• Setup — carriers, carrier_services, warehouses, users, tags, webhooks.
• Stock — products, inventory, inventory_locations.
• 每一行都将完整的 ShipStation 响应保存在 raw_data JSON 列中，因此即使包裹重量、尺寸、海关申报和完整地址不是顶级列，它们仍然都在那里。

谨慎使用共享预算

ShipStation 的速率限制是针对整个账户的，这意味着如果使用会频繁触发该限制的报告工具，就会减慢您的实际发货速度。而这款工具则特意避免了这种情况。

• 代理程序会控制自己使用账户限额的大约六分之一，其余的则留给打印标签的软件。
• 实时刷新使用相同的预算，并且出于同样的原因，每次查询的刷新次数也有限制。
• 发货单带有真实的修改时间光标，因此已编辑的发货单会在下一次增量运行中被拾取。
• Labels and fulfillments are tracked by creation date, which is ShipStation's own filter for them — a label voided after it was synced updates on the next full sync or the moment a live query touches it.

每个 API 连接器的工作原理

• Your account becomes tables — endpoints are mapped to typed SQL tables with primary keys and indexes, so you write SELECT, JOIN and GROUP BY instead of paging through JSON.
• You choose what syncs, and how often — each table has its own interval, or none at all. Where the vendor's API supports it, only records that changed since the last run are fetched.
• The copy is yours — it lives in an AES-256-GCM encrypted DuckDB file next to the agent, on your machine. Not in our cloud.
• The vendor only ever hears from you — the agent calls the API directly from your own network, so your token is not needed by Query Streams to answer a query.

知道这个很有用

• It is a synced copy, not a live proxy — a query reads what has already been fetched, so it returns at SQL speed and never spends your API rate limit. Selected tables can additionally refresh a stale row as you read it.
• Rate limits are respected for you — each connector paces itself well inside the vendor's published limits, so a report cannot get your account throttled for everything else that uses it.
• Nothing is thrown away — fields the table does not name are kept in a raw_data JSON column you can still query.
• Join it to your databases — an API table and a SQL Server table in the same statement is an ordinary query here, not an integration project.