Spam

Rspamd spam scanner — scores inbound mail for the inbound MX receiver via HTTP /checkv2.

Run bun setup for core Docker first. Spam starts with the core Compose stack (not the edge profile) because inbound depends on it.

Overview

Property	Value
Directory	`apps/backend/spam`
Docker service	`reloop-spam` (container `reloop-spam`)
Scan ports	`11332`, `11333`
Web UI	`11334`
Stack	Rspamd · Redis

bun docker:up

Or rebuild after config changes:

docker compose -f local/docker-compose.yml up -d --build reloop-spam

Controller / scan endpoint: http://localhost:11333.

Local overrides live in apps/backend/spam/local.d/ (mounted into the container).

Spam Redis (set on the spam container — native Rspamd RSPAMD_* env):

Variable	Default
`RSPAMD_REDIS_SERVERS`	`reloop-redis:6379`
`RSPAMD_REDIS_PASSWORD`	`reloop123`

Inbound wires to this service with:

Variable	Default
`KUMOMTA_RSPAMD_URL`	`http://reloop-spam:11333/checkv2`

Command	Description
`bun docker:up`	Start core stack including spam
`docker compose -f local/docker-compose.yml up -d --build reloop-spam`	Rebuild and start spam only
`docker logs -f reloop-spam`	Stream logs
`docker compose -f local/docker-compose.yml stop reloop-spam`	Stop container

Inbound (KumoMTA) → HTTP POST /checkv2 → Spam (Rspamd) → score / action / symbols

Layer	Detail
Scan API	Accepts raw message bodies on `/checkv2` (port `11333`)
Headers	Inbound injects `X-Spam-*` from the response for inbox
State	Uses Redis (`reloop-redis`) for learning / fuzzy storage
Separate deploy	Not baked into the inbound image — own Dockerfile and CI (`be-spam`)

One-command local bootstrap

Scan ports and conflicts

MX receiver that calls this service

Persists X-Spam-* scores from NATS