AbsoluteJS

Platform Tools

The building blocks for running Bun apps in production on your own infrastructure — the same primitives that power AbsoluteJS hosting. Each package does one job behind a small API: deploy releases, run isolated processes, route connections, broker secrets, meter cost, scale a fleet, and probe health. Use one on its own, or compose the full stack.

#The Packages

Standalone packages
Every tool below is a standalone published package. None of them require AbsoluteJS hosting — or each other — to be useful; each works against plain Bun projects on your own machines.

Runtimev0.6.1

@absolutejs/runtime

Run many isolated Bun apps on one host — spawn-or-reuse child processes with idle-kill, spawn back-off, and per-process CPU + RSS metrics.

Routerv0.5.2

@absolutejs/router

Route requests and WebSocket connections across backend shards — consistent hashing, per-tenant connection caps and rate limits, healthy-shard skip.

Deployv0.21.0

@absolutejs/deploy

Ship releases to any box you can exec + upload to — atomic symlink swap, release history, rollback, and a pluggable step pipeline.

Secretsv0.9.6

@absolutejs/secrets

Broker credentials host-side — pluggable adapters, safe-to-log fingerprints, rotation, and redaction before text reaches a log sink.

Meteringv0.6.0

@absolutejs/metering

Attribute CPU, egress, and request cost per tenant, and trip a circuit breaker the moment a budget dimension is exceeded.

Billingv0.6.0

@absolutejs/billing

Turn a metering usage snapshot into invoice line items — plans, graduated tiers, free allowances, all in integer micros so float drift is impossible.

Autoscalerv0.2.1

@absolutejs/autoscaler

Evaluate pluggable signals (CPU, queue depth, latency) against a declarative policy and drive your own actuator to spawn, drain, or terminate instances.

Healthv0.3.0

@absolutejs/health

Expose /healthz and /readyz from composable named checks — per-check timeouts, worst-of-any status, and a standard JSON envelope.

Ops CLIv0.1.0

@absolutejs/cli

Config-file-driven secrets / env / deploy verbs at the terminal — the ops sibling to the absolute framework CLI.

isolated-jscv0.12.4

@absolutejs/isolated-jsc

Heap-isolated JavaScriptCore sandbox for running untrusted code in Bun, with an isolated-vm-shaped API, hard memory limits, and timeouts.

Adminv0.2.0

@absolutejs/admin

Framework-neutral roles, capabilities, navigation, and security read models for AbsoluteJS project administration.

Attestv0.1.3

@absolutejs/attest

Keyless Sigstore attestation policy, provenance, signing, and verification for private CI pipelines.

Compliancev0.7.0

@absolutejs/compliance

Framework-agnostic compliance policy used by the hosted AbsoluteJS.ai platform: data classification, residency, retention, subject requests, erasure, and evidence bundles.

Rate Limitv0.5.1

@absolutejs/rate-limit

First-class 2026 rate limit for Bun + Elysia. GCRA by default (exact, O(1) memory per key, no boundary effect — the algorithm Stripe uses), IETF draft-09 RateLimit-* headers, IPv6 /64 grouping, X-Forwarded-For trust modes, BigInt nanosecond TAT, pluggable store (memory LRU bundled).

Reliabilityv0.1.0

@absolutejs/reliability

Provider-neutral atomic inbox, transaction, and idempotent-operation primitives for AbsoluteJS.

Slov0.1.0

@absolutejs/slo

Storage-neutral SLO, error-budget, burn-rate, release-gate, and incident-decision primitives for AbsoluteJS.

Vulnerabilitiesv0.14.1

@absolutejs/vulnerabilities

Scanner-neutral vulnerability normalization, policy gates, host inventory, and audit-ready evidence for AbsoluteJS applications.

Vulnerability Adaptersworkspace

@absolutejs/vulnerabilities-adapters

Feed and storage adapters for @absolutejs/vulnerabilities.

Vulnerability Modulesworkspace

@absolutejs/vulnerabilities-modules

Optional first-party modules for AbsoluteJS vulnerability management.

#How They Compose

The packages share deliberate seams rather than a framework: metering consumes lifecycle and observation events from the runtime, the router takes meter.allow as its one-line admission hook, billing prices a metering usage snapshot, deploy's verify step can probe the endpoints health exposes, and the CLI drives secrets and deploy from one config file. A typical production loop looks like this:

1
deployShip a release
Deploy ships a release — upload, install, build, atomic current symlink swap — and only reports green after verify passes.
2
runtimeRun the tenants
Runtime runs the apps — each one in an idle-killing, metric-emitting child process, spawned on first request and reused after.
3
routerRoute the traffic
Router routes the connections — consistent-hash tenant to shard, enforce connection caps and rate limits, skip unhealthy shards.
4
secretsBroker the credentials
Secrets brokers the credentials — resolve through an adapter, log fingerprints instead of plaintext, redact before log sinks.
5
metering + billingTrack the cost
Metering rolls up per-tenant usage and trips budgets; billing turns the usage snapshot into invoice line items.
6
autoscalerScale the fleet
Autoscaler compares signals like CPU and queue depth against a policy and asks your actuator to spawn or drain instances.
7
healthGate the rollout
Health exposes /healthz and /readyz — deploy verify, load balancers, and drain workflows all gate on the same two probes.

Four companion packages round out a production setup: Queue, Dispatch, Audit, Telemetry. Queue gives you durable background jobs, Dispatch fans work out to workers, Audit records who did what, and Telemetry gets traces and metrics off the box.