🧬 Flow DNA — Genetic Optimization

Every flow execution leaves a "DNA fingerprint." Over hundreds of runs, the library evolves optimal retry, timeout, and caching configs using a micro genetic algorithm.

Flow DNA Evolution

Generation: 0

Timeout

5000ms

Max Retries

3

Stale Time

10.0s

Fitness Score (Latency Optimized)

42.0%

The Concept

Traditional async libraries require manual tuning of timeouts, retry counts, and cache durations. Flow DNA eliminates this by learning from production traffic.

After N executions, the engine:

• Analyzes P95 latency to suggest optimal timeouts
• Tracks success rates to evolve retry counts
• Monitors cache hit rates to tune stale times
• Adjusts debounce based on call frequency

Quick Start

import { useFlow } from "@asyncflowstate/react";

const { data, execute } = useFlow(fetchUser, {
  evolution: {
    enabled: true,
    generations: 50, // Begin optimizing after 50 executions
    mutationRate: 0.1, // How aggressively to tweak params
    fitness: "latency", // Optimize for: 'latency' | 'reliability' | 'bandwidth'
  },
});

How It Works

graph TD
    A["Execute Flow"] --> B["Record Genome"]
    B --> C{"N Executions?"}
    C -->|No| A
    C -->|Yes| D["Run Genetic Algorithm"]
    D --> E["Evolve Config"]
    E --> F["Apply Mutations"]
    F --> G["Track Fitness Score"]
    G --> A

Execution Genome

Each execution captures a performance fingerprint:

Field	Description
latency	Total execution time in ms
retries	Number of retry attempts used
cacheHit	Whether the result came from cache
success	Whether execution succeeded
payloadSize	Response payload size in bytes

Fitness Functions

Mode	Optimizes For	Best When
latency	Fastest response times	User-facing API calls
reliability	Highest success rate	Critical mutations
bandwidth	Smallest payloads	Mobile-first apps

Evolved Configuration

After enough generations, the engine suggests:

// The flow automatically adjusts:
// - timeout: 5000 → 3200ms (learned from P95 latency)
// - retry.maxAttempts: 3 → 2 (server recovers fast)
// - staleTime: 30000 → 45000ms (data rarely changes)
// - debounce: undefined → 150ms (rapid-fire calls detected)

API Reference

FlowDNA

import { FlowDNA } from "@asyncflowstate/core";

const dna = new FlowDNA("user-fetch", { enabled: true });

// Record execution telemetry
dna.record({
  latency: 230,
  retries: 0,
  cacheHit: false,
  success: true,
  payloadSize: 1024,
});

// Get evolved suggestions
const evolved = dna.getEvolved();
// { suggestedTimeout: 350, suggestedMaxRetries: 2, fitnessScore: 0.92, ... }

Persistence

DNA telemetry is stored in localStorage with the key prefix af_dna_. It persists across sessions so flows continue evolving even after page refreshes.