Rift

PostgreSQL CDC pipeline in a single Go binary. No Kafka. No JVM. No complexity.

---

The problem

Syncing data out of Postgres usually means:

Postgres → Debezium → Kafka → Kafka Connect → Destination

Four systems to run, monitor, and debug. Most small teams don't need that complexity.

Rift

Postgres → Rift → Destination

One binary. One config file.

---

How it works

Rift connects to Postgres using logical replication and streams every INSERT, UPDATE, DELETE to your destinations in real time.

If a destination goes down, Rift writes events to a local BoltDB disk queue and automatically drains them when the destination recovers.

No Kafka needed for that resilience.

---

Quickstart

1. Enable logical replication

In postgresql.conf:

wal_level = logical

2. Create rift.yaml

source:
  type: postgres
  url: postgres://user:pass@localhost:5432/mydb?replication=database
  slot: rift_slot
  publication: rift_pub

destinations:
  - name: my-webhook
    type: webhook
    url: https://myapp.com/webhook

  - name: analytics-db
    type: postgres
    url: postgres://user:pass@analytics:5432/analytics

  - name: cache
    type: redis
    url: redis://localhost:6379

queue:
  enabled: true
  path: ./rift-queue
  max_size_mb: 1000

filter:
  script: ./filter.js

3. Run

go run main.go

---

JS Filtering

Drop events at source before they consume bandwidth:

function filter(event) {
  // only sync enterprise users
  if (event.data.plan !== 'enterprise') return false

  // drop test emails
  if (event.data.email.includes('test@')) return false

  return true
}

Point to your script in rift.yaml:

filter:
  script: ./filter.js

---

Destinations

Type	Status	Description
Webhook	✅ v0.1.0	HTTP POST with JSON payload
HTTP	✅ v0.1.0	Generic HTTP endpoint
Postgres	✅ v0.3.0	Real-time DB to DB sync
Redis	✅ v0.4.0	Pub/sub + event list

---

Disk Queue

When a destination goes offline Rift switches to air-gap mode automatically.

Events write to local BoltDB instead of being dropped. When the destination recovers, Rift drains the queue and resumes. No events lost. No manual intervention.

---

Event format

{
  "table": "users",
  "operation": "INSERT",
  "data": {
    "id": "1",
    "name": "Mujib",
    "email": "mujib@example.com"
  },
  "lsn": "0/16C752F8",
  "timestamp": "2026-05-20T14:32:00Z"
}

---

Roadmap

v0.1.0  ✅  WAL reading + webhook destination
v0.2.0  ✅  Embedded BoltDB disk queue
v0.3.0  ✅  Postgres destination
v0.4.0  ✅  Redis destination + JS filtering
v0.5.0  →   DDL schema tracking
v1.0.0  →   Production ready — Debezium alternative

---

Requirements

• PostgreSQL 12+ with wal_level = logical
• Go 1.26+

---

License

MIT