POST /v1/verify - PromptWall

POST https://api.prompt-wall.com/v1/verify Use this endpoint after your application’s LLM has produced an answer. PromptWall scans the input, checks the answer against the grounding tool_result, and applies any policies you’ve configured.

Request

Required headers

Header	Value
`X-API-Key`	Your PromptWall API key (`pw_…`)
`Content-Type`	`application/json`

Authorization: Bearer pw_… is also accepted.

Optional headers

Header	Default	Notes
`X-Use-Case`	`default`	Dashboard dimension
`X-Team-Id`	(unset)	Dashboard dimension
`X-Integration-Id`	`api:v1_verify`	Identifies the calling integration
`X-End-User-Id`	(unset)	End-user pseudonym
`X-Session-Id`	(unset)	Multi-turn session correlation

Body

{
  "prompt":      "What is our Q3 revenue?",
  "answer":      "Q3 revenue was $12.4M.",
  "tool_result": "Q3 revenue: $12.4M (from internal BI)",
  "model":       "gpt-4o-mini",
  "provider":    "openai"
}

Response

{
  "ok":              true,
  "request_id":      "1a2b3c…",
  "answer":          "Q3 revenue was $12.4M.",
  "answer_changed":  false,
  "governance":      "allow",
  "mismatch_type":   "none",
  "policy_hits":     [],
  "explanation":     { … }
}

`governance` field

Value	Meaning
`allow`	Answer passed every check — returned unchanged
`rewrite`	Answer modified (e.g. PII redacted) — `answer_changed: true`
`regenerate`	Answer violated a policy; a new answer was produced
`block`	Answer cannot be returned — replaced with a safe fallback

`mismatch_type` field

Value	Meaning
`none`	No grounding mismatch
`numeric`	Number in the answer doesn’t match `tool_result`
`contradiction`	Direct contradiction with grounding source
`unsupported`	Claim isn’t present in the grounding source

Errors

Status	Body
400	`{"ok": false, "error": "answer is required"}`
401	`{"ok": false, "error": "Invalid or missing API key"}`

The full request and decision are recorded as a trace and appear on the Observability dashboard.