Dedup, matching & cheap classification at volume

Batch pipelines that ask "same or different?", "which bucket?", "route left or right?" millions of times a day are the worst fit for pay-per-output generative pricing. Decision models flip the economics: you pay for the input you send, and the verdict is free.

Why decision models

• Output tokens free — verdicts cost nothing
• Input pricing from $0.01/M tokens
• Per-key RPM/TPM limits for batch control
• Token-precise logs for cost attribution

Combine with per-key rate limits and token-precise usage logs to run predictable batch jobs, and switch models per request when a cheaper checkpoint fits the bucket.

Runnable example

curl https://api.easyjev.online/v1/systemone \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "typesafe/jev-latest",
    "state": {"user": {"plan": "pro"}},
    "questions": {
      "is_premium": {"type": "noul", "criteria": {"true": "…", "false": "…"}}
    }
  }'