Guard prompt before model execution
POST/api/v1/guard
Evaluates prompt payload before model execution and returns allow/block verdict with event correlation fields.
Use this endpoint synchronously in front of LLM execution.
Request
Responses
- 200
- 400
- 401
- 403
- 413
- 429
- 500
- 503
Guard decision completed.
Invalid payload.
Missing, invalid, revoked, or unauthorized API key.
Request blocked by the guard policy engine.
Payload exceeds endpoint size limits.
Rate limit exceeded.
Response Headers
Retry-After
Seconds to wait before retrying.
Example:
2X-RateLimit-Limit
Max requests allowed for current window.
Example:
60X-RateLimit-Remaining
Remaining requests in current window.
Example:
0X-RateLimit-Reset
UNIX epoch seconds when current limit window resets.
Example:
1767225600Internal server processing failure.
Temporary infrastructure timeout or dependency outage.