Horizon — Queue Dashboard

Laravel-style queue dashboard: in-memory throughput stats, optional Redis-backed failed jobs, live Redis queue depths (pending / delayed / processing / in-flight), Prometheus metrics, and a configurable Gate for production.

§ Overview

• Dashboard HTML under HORIZON_PATH (default /horizon).
• Observer hooks increment global Prometheus counters (same registry as /metrics).
• With Options.RedisURL (typically REDIS_URL): failed job persistence + Horizon UI for list / forget / retry.
• Same Redis client powers live workloads on the Pending page and GET …/api/workloads using the same keys as queue.RedisAdapter.
• HORIZON_QUEUES — comma-separated queue names to always include in workload snapshots (even before traffic hits the observer).

§ Installation

$ nimbus plugin:install horizon

Or register manually in bin/server.go:

import (
    "os"
    "github.com/CodeSyncr/nimbus/plugins/horizon"
)

app.Use(horizon.NewWithOptions(horizon.Options{
    RedisURL: os.Getenv("REDIS_URL"),
    Gate:     nil, // optional: func(*http.Context) bool
}))

§ Environment

§ Worker config (horizon.Config)

Code-driven supervisors (environments, balance strategies, tries, timeout, backoff) live on github.com/CodeSyncr/nimbus/plugins/horizon. Pass a pointer via horizon.Options{ Config: &cfg } or use horizon.DefaultConfig(). Fields match Laravel-style naming: Environments, SupervisorConfig with Balance auto / simple / false, etc.

§ Failed jobs

When Redis is configured, the plugin registers queue.SetFailedJobStore. Failed jobs appear under Failed in the UI. From code, use queue.GetFailedJobStore() after boot (same interface the dashboard uses).

§ HTTP routes (relative to HORIZON_PATH)

§ Programmatic queue depths

Outside Horizon, use the core queue helper with a github.com/CodeSyncr/nimbus/redis client:

workloads, err := queue.RedisQueueWorkloads(ctx, rdb, []string{"default", "emails"})

See also: Queue, Production readiness, Reverb (WebSockets).