Configuration — Atrion

Workload Profiles

Different workloads need different baselines. A 100ms API response is unhealthy, but a 100ms video encoding step is insanely fast.

Profile	Baseline	Use Case
LIGHT	10ms	Health checks, cache reads
STANDARD	100ms	REST APIs, database queries
HEAVY	5s	Video transcode, ML inference
EXTREME	60s	Genome sequencing, AI swarms

// Default (STANDARD)
atrion.route('api/users', telemetry)

// Heavy computation
atrion.route('ml/inference', telemetry, {
  profile: 'HEAVY'
})

// Long-running with lease
const lease = await atrion.startTask('genom/sequence', {
  profile: 'EXTREME',
  abortController: controller,
})

lease.heartbeat({ progress: 0.5 })
await lease.release()

Auto-Tuning

The engine uses Z-Score based statistics to dynamically adjust thresholds:

dynamicBreak = μ(R) + 3σ(R)

Break point is mean resistance plus three standard deviations

Bootstrap

Initial samples collected. Safe defaults applied.

Operational

Thresholds track your traffic using EMA.

Adaptive

Seasonal patterns and traffic shifts handled.

Zero Configuration: Auto-tuning is enabled by default. No manual thresholds needed.

State Providers

Pluggable state backends for different deployment scenarios:

InMemory (Default)

Single Node

const atrion = new Atrion()
// Uses InMemoryProvider by default

Redis

Distributed

import { Atrion, RedisStateProvider } from 'atrion'

const atrion = new Atrion({
  stateProvider: new RedisStateProvider({
    url: 'redis://localhost:6379',
    keyPrefix: 'atrion:',
  })
})

All Options

Option	Default	Description
engine	'auto'	'wasm' \| 'js' \| 'auto'
stateProvider	InMemory	StateProvider instance
bootstrapSamples	100	Samples before operational mode
decayRate	0.1	Scar tissue forgiveness rate (λ)
scarFactor	5	Trauma weight (σ)

./typescript-sdk ./api