Skip to main content
EventDBX is intentionally opinionated about append-only storage yet lenient everywhere else. Domains, schema modes, and plugin payload modes form building blocks you can rearrange without touching the core event log.

​
Carve up work with domains

Use domains whenever a team, product surface, or environment needs its own truth:
Each domain keeps its own RocksDB store, Merkle trees, replication offsets, and plugin queues. Teams can experiment safely by creating *-dev domains that reuse only the schemas they trust.

​
Switch domains with dbx checkout

dbx checkout activates a domain (and can create or delete it) without touching other domains:
  • dbx checkout <name> --create to spin up a new domain for experiments, CQRS bounded contexts, or per-environment isolation.
  • --delete [--force] to remove a domain deliberately; use when retiring experiments.
  • --remote <host[:port]> --token <token> [--remote-tenant <tenant>] to bind a domain to a remote control socket, useful when your write node lives elsewhere.
This keeps iteration friction low—you can hop between domains or point a domain at a remote daemon without rebuilding schemas or replaying history.

​
Evolve schemas gradually

Start in a permissive validation mode while schemas evolve, then restart the daemon with a stricter mode once shapes settle (e.g., dbx start --restrict=strict).
EventDBX never mutates historical payloads. Instead, new columns surface through snapshots and plugins once they appear often enough. Re-snapshotting an aggregate with dbx aggregate snapshot <aggregate> <id> lets you adopt the updated shape without replaying the full stream.

​
Route only the slices you need

Plugins choose the minimal payload per domain when you run dbx plugin config <type> … --payload <mode>. Modes include all (default), event-only, state-only, schema-only, event-and-schema, and extensions-only (clears payload but keeps metadata/extensions). For process plugins, add --emit-events=false if you want the worker supervised without enqueuing jobs.
Because payloads are set per plugin, you can bolt on new read models without modifying the write path or exposing data they do not need.

​
Sample workflow

  1. Create an experiment domain (payments-fx-dev) for foreign-exchange work.
  2. Copy the core schema and relax validation on new FX fields.
  3. Attach a log plugin that mirrors only FX aggregates for analysts.
  4. Once the feature stabilizes, promote the schema to payments-prod and rehydrate downstream stores by replaying the FX events.
This pattern lets you iterate quickly while preserving EventDBX’s guarantees around durability and integrity.