Client Interfaces
Choose an interface first by where the database engine will run. An ORM is an application abstraction, not a third deployment mode.
| Interface | Engine location | Application responsibility |
|---|---|---|
| Embedded Rust API | Inside the application process | Open the database, execute operations and close it |
| Rust TCP client | In a separate RadixDB server process | Connect, authenticate, select a database and consume results |
| ORM | Uses a database execution interface | Describe models or records and choose transaction boundaries |
Embedded Access
Section titled “Embedded Access”The local CLI used at the start of the tutorial demonstrates embedded access:
it opens a memory or file database without a network service. Rust applications
use the public database API provided by radixdb-api (also exposed through the
top-level radixdb crate). The application owns the engine lifecycle and needs
access to the database files.
TCP Access
Section titled “TCP Access”radixdb-client speaks the RadixDB binary protocol. It does not embed the
storage engine. Use a client compatible with the server protocol; do not assume
PostgreSQL clients or a PostgreSQL connection URL can connect to RadixDB.
The server exercise runs a supplied client.
Its source is examples/public/rust-client/basic.rs: connection, authentication,
database selection, SQL execution, cursor fetching and connection shutdown.
Fetch until EOF or close an unfinished cursor before sending an unrelated command.
The local CLI’s database path and the server’s database name are not interchangeable.
radixdb-orm provides model and record abstractions for application code.
It does not remove the need to understand SQL constraints, transaction boundaries
or errors. Choose embedded versus server deployment separately from whether
the application uses ORM operations. See Embedded Rust,
the Rust TCP client and the Rust ORM for the
detailed contracts and executable examples.
Scope of This Overview
Section titled “Scope of This Overview”This overview is not the complete API contract. The detailed chapters define parameter types, prepared statements, transaction errors and uncertain outcomes after a lost connection. In particular, do not automatically retry a write merely because its reply was not received.