AI

This page covers the AI test case assistant and AI settings endpoints.

All paths are relative to /api/v1. Requests require authentication. AI endpoints also require the project to have effective AI settings.

Test Case Assistant

Method	Path	Purpose
GET	/ai/test-cases/status	Get AI feature status for the current user and optional project.
POST	/ai/test-cases/generate	Generate draft test cases from a prompt or source context.
POST	/ai/test-cases/{test_case_id}/review	Review an existing test case.
POST	/ai/test-cases/duplicates/check	Check candidate test cases for duplicates.

GET /ai/test-cases/status accepts optional project_id.

Duplicate checks require viewer access to the payload project and configured AI provider settings.

AI Settings

Method	Path	Purpose
GET	/settings/ai	List AI settings overview for manageable projects.
GET	/settings/ai/global	Get global AI settings.
PUT	/settings/ai/global	Update global AI settings.
GET	/settings/ai/{project_id}	Get project AI settings.
PUT	/settings/ai	Update project AI settings.
DELETE	/settings/ai/{project_id}	Remove project AI settings and fall back to global/env config.

Global AI settings are restricted to superadmin users. Project AI settings are restricted to users who can manage that project.

Required Permissions

Unless a more specific rule is stated above, read endpoints require viewer, create and update endpoints require tester, and destructive endpoints require lead or the stricter role enforced by the owning service.

Request and Response Schemas

Use the examples on this page together with the generated OpenAPI schema at /docs or /redoc for exact field types. JSON endpoints use application/json; upload endpoints use multipart/form-data; resource ids are passed as path parameters.

Status Codes and Errors

Common responses are 200 for reads and updates, 201 for creates, 204 for successful deletes, 401 for missing authentication, 403 for insufficient project role, 404 for missing or inaccessible resources, and 422 for validation errors.

{
  "type": "https://tms.local/errors/validation_error",
  "title": "Validation failed",
  "status": 422,
  "detail": "Request contains invalid fields",
  "instance": "/api/v1/example",
  "code": "validation_error",
  "request_id": "req_123",
  "errors": {
    "field": ["Field is required"]
  }
}

Pagination and Filtering

List endpoints that expose page and page_size use 1-based pagination. Many list endpoints cap page_size at 200; row-oriented endpoints may document a different cap. Filtering examples on this page use repeated query parameters when multiple values are supported.

Idempotency and Retries

GET requests are safe to retry. POST, PATCH, DELETE, bulk operations, and imports are not guaranteed to be idempotent unless the endpoint explicitly says so. After a network timeout on a write request, read the affected resource before retrying.

Limits and Destructive Operations

Karvio does not currently enforce application-level rate limits. File upload limits are documented on upload-specific pages. Delete, archive, unlink, and bulk delete operations can remove data or hide it from active workflows; confirm project scope and permissions before calling them from automation.