n8n AI Core

Multi-provider AI routing and centralized config for n8n — production-ready in 5 minutes.

Two sub-workflows that form the foundation of a production n8n AI stack. The Model Router handles 6 AI providers (Gemini, Claude, GPT-4o, Groq, Grok, Fal.ai) with automatic fallback — call one workflow, get the best provider for each task type. The Config Loader centralizes all API keys and secrets in one place using Docker secrets. Import both, wire your keys, and every workflow in your instance can route AI calls through a single, reliable pipeline. Built from 56 production workflows running daily on real infrastructure. Updated to claude-sonnet-4-6.

What's included

Model Router — 6 AI providers with automatic fallback (Gemini, Claude, GPT-4o, Groq, Grok, Fal.ai)
10 task types — research, writing, code, code-review, quiz, image, fast, embedding, summary, classify
Config Loader — centralized Docker secrets management, one place for all API keys
Primary + fallback routing — if one provider fails, the next picks up automatically
Latency tracking — every call reports model used, fallback status, and response time
Image generation via Fal.ai Flux Schnell — landscape, portrait, or square
Google Search grounding — Gemini research calls include live web results
Clean interface — { prompt, taskType, maxTokens } in, { text, model, success, latencyMs } out
No hardcoded secrets — all keys loaded from Docker secrets via Config Loader
Fully customizable routing table — swap providers or add task types in one code node

Requirements

n8n self-hosted (any recent version)
At least one AI provider API key (Gemini free tier works as primary)
Docker secrets support (standard in Docker Compose deployments)

model-router.js

// 6 providers. 10 task types. Automatic fallback.

const routing = {
  research:      { primary: 'gemini',  fallback: 'claude' },
  writing:       { primary: 'gemini',  fallback: 'claude' },
  code:          { primary: 'claude',  fallback: 'openai' },
  'code-review': { primary: 'claude',  fallback: 'openai' },
  fast:          { primary: 'groq',    fallback: 'grok'   },
  summary:       { primary: 'grok',    fallback: 'groq'   },
  image:         { primary: 'fal',     fallback: 'mermaid'},
};

// Call: { prompt, taskType: 'writing', maxTokens: 2000 }
// Returns: { text, model, taskType, success, usedFallback, latencyMs }