Check a model loop's calls for spinning

Say, from the tool calls a model loop made in order, when it should be nudged for repeating itself and when the job should be given back because it keeps repeating or its tools keep failing

Give calls (each { name, args, failed, malformed }) in the order they were made and, if wanted, rules ({ repeat_nudge_after, repeat_give_back_after, error_budget, words }; the shipped ones in read-data/loop-guards.json otherwise). A call is the same call when its name and its arguments match, whatever their key order. A malformed call (arguments that were not JSON, a tool the loop may not call) never counts. Failures count only in a row; one success clears them. Answers steps (each { at, nudge } or { at, give_back }) and, when the job should go back, give_back and the call index at.

What it asks for

Nothing: your AI starts it with your words.

Use it from your AI

Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.

Say to your AI: “Use Trillion to say, from the tool calls a model loop made in order, when it should be nudged for repeating itself and when the job should be given back because it keeps repeating or its tools keep failing”