audit_ai_inference_escrow
[Private] Were you overbilled or served a cheaper model? Checks LLM output for token inflation, repetition collapse and missing required terms. Example: billedOutputTokens=32.
Tool call template
Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.
{
"id": 1,
"jsonrpc": "2.0",
"method": "tools/call",
"params": {
"arguments": {
"responseText": "<responseText>"
},
"name": "audit_ai_inference_escrow"
}
}Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| responseText | Yes | Generated text output returned by the inference provider. | |
| forbiddenTerms | No | Terms that must not appear in the output. | |
| mustIncludeTerms | No | Keywords or phrases that must appear in the output. | |
| billedOutputTokens | No | Number of output tokens billed by the provider. | |
| minUniqueWordRatio | No | Minimum lexical diversity ratio (0 to 1) to detect model downgrade. |