Guardrail Detect Image
Analyze an image with text using a saved guardrail with input/output mode selection
Authorizations
Headers
The guardrail name
"My Guardrail"
Whether to apply input (prompt) or output (response) detectors
prompt, response "prompt"
Body
Request body for guardrail/policy-based image detection. Detectors are defined by the guardrail or policy; no inline detectors config is needed.
The text prompt to accompany the image for analysis.
"Help me with the content in this image"
Base64-encoded image file content.
"<base64-encoded-image-data>"
Optional arbitrary metadata object echoed back in the response. Limits: at most 16 top-level keys, key names up to 64 characters, nesting depth up to 3, total serialized size up to 4 KB. When provided and valid, this object is also echoed as a top-level user_metadata key on error responses (4xx/5xx) of endpoints that accept it; invalid or oversized metadata is never echoed, including on the 422 response it causes.
Response
Multimodal image detection results.
Per-detector integer flags (1 = detected, 0 = not detected).
Per-detector detail objects containing human-readable results and any extra fields (e.g. entities for PII, explanation for policy_violation).
Optional arbitrary metadata object echoed back in the response. Limits: at most 16 top-level keys, key names up to 64 characters, nesting depth up to 3, total serialized size up to 4 KB. When provided and valid, this object is also echoed as a top-level user_metadata key on error responses (4xx/5xx) of endpoints that accept it; invalid or oversized metadata is never echoed, including on the 422 response it causes.

