Hi all. I maintain n8n-nodes-jev-classification, a community node that wraps Jev, TypeSafe’s System One decision model, for classify, score and check operations with calibrated probabilities.
0.3.0 is out:
- The credential now has a Base URL field (default https://api.typesafe.ai). Set https://openrouter.ai/api or https://ai-gateway.vercel.sh/typesafe and the node calls Jev through that gateway. Existing credentials are unchanged.
- The credential test now sends one tiny real request to /v1/systemone (about 270 input tokens) instead of listing models, because gateways serve /v1/models without a key.
The node behaves like the built-in Text Classifier: one output branch per category plus a Needs Review branch when confidence is below your threshold. Score returns a probability weighted score, most probable level, legend and confidence. Check routes to Yes or No. Ask Questions sends a mix of choice, score and noul questions and returns the answers map. It also works as an AI Agent tool.
Volume handling: up to 50 items per request, up to 20 parallel requests, retries on 429, 529 and 5xx with backoff that honors retry-after.
Note on verification: this package is self-hosted only. n8n declined verification in September 2026 because of overlap with the built-in classification node that is in progress. The package is MIT and fully usable when installed from Community Nodes.
Install: Settings > Community Nodes > Install, then n8n-nodes-jev-classification Repo: GitHub - khmuhtadin/n8n-nodes-jev-classification: n8n community node for Jev by TypeSafe AI: classify, score and check text with calibrated probabilities. Parallel requests and multi-item batching. · GitHub Four importable example workflows: n8n-nodes-jev-classification/examples at main · khmuhtadin/n8n-nodes-jev-classification · GitHub
Feedback and issues welcome, especially from people running it through a gateway.