Skip to main content
Plugins turn EventDBX into a hub: the write path stays lean while plugins fan events out to search, caches, analytics, or custom workers. Jobs are enqueued per enabled plugin when events commit; payload modes trim what each plugin receives.

​
How it works

  1. Events commit.
  2. EventDBX enqueues a job per enabled plugin (respecting its payload mode and emit_events flag).
  3. A short-lived delivery is attempted; failures stay in the queue and retry with backoff until plugin_max_attempts.

​
Explicit publish (opt-in)

You can override the default “fan out to all enabled plugins” on a per-write basis. Provide publish targets on the write command to select plugins, override payload mode, and set queue priority: 
dbx aggregate apply orders order-1 order_created \
  --payload '{"status":"pending"}' \
  --publish search-indexer:all:high \
  --publish analytics-engine:state-only
More examples: 
# Patch an event and send only to the ETL plugin with default priority
dbx aggregate patch orders order-1 order_created \
  --patch '[{ "op": "replace", "path": "/status", "value": "paid" }]' \
  --publish etl:state-only

# Create and fan out to two plugins with different modes and priority
dbx aggregate create orders order-2 order_created \
  --payload '{"status":"pending"}' \
  --publish search-indexer:event-only:high \
  --publish audit-log:all
Each --publish entry is PLUGIN[:MODE[:PRIORITY]]:
  • PLUGIN – configured plugin name (must be enabled).
  • MODE (optional) – overrides payload shape for this write (all, event-only, state-only, schema-only, event-and-schema, extensions-only); defaults to the plugin’s configured mode.
  • PRIORITY (optional) – low|normal|high for queue ordering; defaults to normal.
If no publish targets are provided, EventDBX fans out to all enabled plugins using their configured payload modes.

​
Configure plugins

Use dbx plugin config <type> to create or update an instance (saved at the system data root, e.g. ~/.eventdbx/plugins.json): 
dbx plugin config http --name search --endpoint https://search.internal/hooks --payload event-only
dbx plugin config tcp --name fraud --host fraud.internal --port 7075 --payload state-only
dbx plugin config log --name audit --level info --template "agg={aggregate} id={id} event={event}"
dbx plugin config process --name enrich --plugin dbx_enrich --version 1.0.0 --payload all
Start/stop managed process workers: 
dbx plugin start enrich
dbx plugin status enrich
dbx plugin stop enrich
List and test: 
dbx plugin list
dbx plugin test [<name>…]

​
Payload modes

  • all (default): event + state + schema.
  • event-only: event document only.
  • state-only: latest state only.
  • schema-only: aggregate schema only.
  • event-and-schema: event + schema, no state.
  • extensions-only: identifiers + namespaced extensions; payload/state nulled out.

​
Retries and backpressure

  • Jobs retry automatically up to plugin_max_attempts (set via dbx config --plugin-max-attempts <n>).
  • When the persistent queue is unavailable, EventDBX attempts direct delivery and logs failures.
  • Use dbx queue to inspect pending and dead jobs; dbx queue retry replays failures.

​
Deployment options

  • TCP/HTTP/Capnp/Log plugins run as lightweight emitters inside the server/CLI process.
  • Process plugins run as supervised binaries; EventDBX starts/stops them and writes status files under the domain data directory.
Dive into transport-specific guides for configuration details and payload examples.