autoverse

AutoVerse — production-grade multitasking & parallel AI custom nodes for n8n. Phase 1: Parallel Task Runner with bounded concurrency, AIMD rate-limit awareness, per-task timeout, retry with exponential backoff, deterministic ordering and full failure isol

Package Information

Released: 1 mo ago
Downloads: 0 weekly / 0 monthly
Latest Version: 0.1.0
Author: AutoVerse

Available Nodes

Documentation generating…

AutoVerse Parallel Task Runner
Runs the incoming items in parallel with a bounded concurrency pool, retries, 429/Retry-After awareness, per-task timeout and deterministic output ordering.

Documentation

n8n-nodes-autoverse

Production-grade multitasking & parallel AI custom nodes for n8n.

Phase 1 (this release, v0.1.0): AutoVerse Parallel Task Runner.

Run incoming n8n items in parallel with a bounded worker pool, AIMD adaptive
rate-limit reaction, per-task timeout, retry with exponential backoff,
deterministic output ordering and full failure isolation.

Why

The n8n community forum is full of threads asking how to run LLM / HTTP calls
in parallel without triggering 429 storms. Today the workaround is 8–15 nodes,
sub-workflows, static-data counters and webhook callbacks. This node reduces
all of that to a single step.

Install

Community-nodes settings inside n8n → npm Package Name →
n8n-nodes-autoverse

Or manually:

cd ~/.n8n/nodes
npm install n8n-nodes-autoverse

Node: AutoVerse Parallel Task Runner

Option Default Meaning
Mode http http = URL/body per item, code = inline JS body
Concurrency 5 max in-flight tasks
Adaptive (AIMD on 429) true shrink on 429, grow on success streak
Max Retries per Task 2 on top of first attempt
Base Backoff (ms) 500 with full jitter
Backoff Cap (ms) 30 000
Task Timeout (ms) 60 000 via per-task AbortController
Max Execution Time (ms) 0 (disabled) global deadline; remaining items become skipped:deadline
Output Ordering input input (preserved) or completion (fastest first)
Emit summary item true adds { __autoverse_summary: metrics } at the end

Output per item

{
  "__autoverse": {
    "index": 0,
    "status": "success | failed | skipped",
    "attempts": 1,
    "durationMs": 234,
    "retried": false,
    "skipReason": null,
    "error": null
  },
  "input":  { ... original item json ... },
  "output": { ... task result ... }
}

Summary item

{
  "__autoverse_summary": {
    "totalTasks": 100,
    "successful": 97,
    "failed": 2,
    "skipped": 1,
    "retries": 6,
    "rateLimits": 4,
    "timeouts": 2,
    "cancelled": false,
    "concurrencyRequested": 10,
    "concurrencyFinal": 5,
    "totalDurationMs": 2410,
    "averageDurationMs": 218,
    "adaptive": true
  }
}

Benchmark

100 tasks × 200 ms simulated work:

Scenario Duration Speed-up
Sequential (n8n Loop Over Items today) 20.05 s 1.00×
Parallel concurrency=10 2.01 s 9.99×
Parallel + 20% 429 + AIMD 2.41 s 8.32×
Parallel + failures + timeouts 2.55 s 7.91×

License

MIT

Roadmap

  • v0.2 — AutoVerse Batch LLM (OpenAI / Anthropic / Gemini / Groq Batch APIs, 50% discount).
  • v0.3 — AutoVerse Smart Cache & Router (prompt caching + model routing).

See ARCHITECTURE.md and PHASE-1-REPORT.md.

Discussion