API Reference
Safe AI Workbench provides enterprise AI capabilities with built-in PHI detection, policy enforcement, and safety controls. The API is organized around REST principles with JSON request/response bodies.
phiDetected flag in responses.AI Chat
Execute AI tasks with automatic PHI detection and policy enforcement. Supports file uploads and conversation continuity.
AI Tasks
List available AI tasks (20+ pre-built) and retrieve task definitions with variable schemas and requirements.
File Processing
Upload PDF, DOCX, and XLSX files for AI processing with automatic PHI detection and 24-hour retention.
Policies
Create and manage safety policies with warn, block, redact, or allow actions for sensitive content patterns.
Analytics
Retrieve usage statistics, cost estimates, PHI detection trends, and policy violation reports.
Groups
Manage user groups for task sharing, access control, and group-specific policy enforcement.
Common patterns
All endpoints require a Bearer API key in the Authorization header. Requests are rate limited per account. Handle 429 Too Many Requests by retrying with exponential backoff.
- Every collection endpoint is paginated with
limitandoffsetparameters. - Date filters expect ISO-8601 date strings (
YYYY-MM-DD). - Identifiers are UUIDs that map to Guardian Health’s unified records across EHR vendors.
- Metadata blocks include
source,fetched_at, and pagination details for traceability.
Collection endpoints
List endpoints return paginated arrays of resources. Use query parameters for filtering, searching, and pagination.
Resource endpoints
Detail endpoints return a single resource by ID. Combine with collection filters for robust data fetching workflows.