queue-ti

Appearance

queue-tiSelf-hosted distributed message queue backed by PostgreSQL

No Kafka, Redis, or RabbitMQ. If you have Postgres, you have a production-ready queue.

Get Started
View on GitHub
🗄️

PostgreSQL Only

No additional infrastructure. If you run Postgres already, queue-ti is a drop-in message queue with one table.

✅

At-Least-Once Delivery

Messages are never lost. Visibility timeouts ensure unacked messages are retried. Dead-letter queue automatically contains exhausted messages.

📊

Built for Observability

Prometheus metrics out of the box; live queue depth via REST API; admin UI shows message status, retry counts, and expiry times.

⚡

High Performance

gRPC protocol with FOR UPDATE SKIP LOCKED dequeue. Throughput tested at 1500+ ops/sec per consumer.

🎯

Admin UI Included

Inspect messages, manually enqueue test data, requeue from DLQ, manage topics and users—all without writing code.

🔒

JWT Authentication

Optional JWT-based auth with user accounts, role-based access, and per-topic grants. OAuth-ready.

📦

Multi-Language Clients

Official clients for Go, Node.js, Python, and Java — auto-reconnect, token refresh, and batch consumption included.

Quick Start ​

bash
docker-compose up

Access the admin UI at http://localhost:8081 (login: admin / secret).

Why queue-ti? ​

queue-ti is designed for teams who want reliable, observable message processing without the operational overhead of a separate queue broker. It features:

  • Built-in dead-letter queues — Failed messages are automatically moved to a DLQ for manual inspection and requeue
  • Per-topic schema validation — Avro schemas enforce payload contracts at enqueue time
  • Fine-grained JWT grants — Control access per topic, topic pattern, or consumer group
  • Throughput throttling — Per-topic rate limits using PostgreSQL token-bucket algorithm
  • Consumer groups — Multiple independent systems process the same messages in parallel
  • Message keys & upsert — Deduplication and idempotent enqueue operations
  • Automatic reapers — Expiry reaper marks old messages; delete reaper cleans up disk space
  • Multi-language clients — Go, Node.js, and Python with auto-reconnect and token refresh

Core Features ​

  • gRPC API — High-performance queue operations (enqueue, dequeue, acknowledge, nack) over gRPC
  • HTTP Admin API — REST endpoints for queue inspection, management, user/grant administration, and schema configuration
  • Topic-based routing — Multiple independent queues share a single PostgreSQL table, partitioned by topic
  • Message keys — Optional deduplication keys allow upsert semantics
  • Automatic retries — Failed messages are automatically retried up to a configurable limit
  • Dead-letter queue — Messages that exhaust their retry limit are automatically promoted to <topic>.dlq
  • Message TTL — Messages expire after a configurable duration
  • Contention-free dequeue — Uses FOR UPDATE SKIP LOCKED for lock-free concurrent consumption
  • JWT authentication — Optional JWT-based auth with user accounts, role-based access, and per-topic grants
  • Avro schema validation — Optional per-topic Avro schema registration; payloads validated at enqueue time
  • Per-topic configuration — Override retry count, TTL, queue depth limits, and throughput caps per topic via HTTP API or admin UI
  • Throughput throttling — Optional per-topic message rate limits (messages/second) enforced at dequeue time
  • Admin UI — Angular web interface for message inspection, manual enqueue, DLQ requeue, and topic management
  • Prometheus metrics — Real-time counters and gauges (/metrics endpoint, unauthenticated)
  • Client libraries — Go, Node.js, and Python with async-first design, auto-reconnection, and token refresh