Client and Server model

As the monicker, "a full-stack database" implies, Triplit has components that run on both the client and the server. In particular, each client has its own database that is fully queryable just like the remote database. Thus, any query can be immediately fulfilled by the client without needing to go to the network. The client database is kept in sync with the remote database by setting up subscriptions to queries.

Along with reads, every client is able to write to its database and Triplit ensures the write is propagated to the remote database and other listening clients. Because writes go directly to the client database, your application gets an optimistic effect to all of its writes.

Client Database

On the client, Triplit keeps your data in two storage areas: an outbox and a cache. The outbox is a queue of data that is waiting to be sent to the server. The cache is a local copy of the data that has been confirmed to have been received by the server.

Although the outbox acts as a queue, it is actually a separate part of the database that is equally queryable. By default, results from the cache and outbox are merged together, however queries can be configured to only return data from the cache (to display only confirmed data) or only return data from the outbox (to denote unconfirmed data).

If a client is offline or is not connected to the remote database for any reason, then the data will remain in the outbox so you can sync that data can be synced at a later time.

Server Database

The server database provides durable centralized storage for clients and handles propagating writes to other listening clients. As well, the server acts as an authoritative figure for data as needed. For example, access control rules are checked on the server and unauthorized updates are rejected and not propagated.