BreezeData management for JavaScript clients

How the client and server talk ​

1. Metadata. The client asks once for the model: entity types, keys, relationships and validation. A Breeze .NET server reads it from your Entity Framework Core or NHibernate mapping, so you never describe the model twice. See Metadata. The same metadata is what breeze-gen-entities turns into a TypeScript class per entity type, which is what gives you typed entities and typed queries — see Generating entity classes.
2. Query. The client composes a query — filter, sort, page, expand — and sends it in the URL. The server applies it to the IQueryable your action returns, so it runs in the database as SQL. The entities that come back are merged into the cache. See Querying.
3. Save. The client sends every pending change in one request. The server saves them in one transaction, and sends back the saved entities with the real keys in place of the temporary ones. See Saving changes.

The server side is covered in Using a Breeze .NET server. Breeze can also talk to other back ends — see Talking to the server.

Breeze 3 ​

Version 3 is a rewrite of the 2.x codebase with the same public API. What changed:

• ESM only, one package, one npm tag. No CommonJS build, no UMD bundle, no mjs/cjs dist-tags to choose between.
• No runtime dependencies.
• No adapter setup — Breeze uses its standard adapters unless you register others. When you do, configureBreeze replaces the stringly-typed adapter registration.
• Knockout, jQuery, AngularJS and OData support removed. See Migrating from 2.x for what to do if you use them.
• Built under strictNullChecks and noImplicitAny.

If you are coming from 2.x, start with Migrating from 2.x — most applications need only a handful of changes, and they are listed there.