Developers

VeroniQA for developers

VeroniQA's public API is a free, keyless sandbox: ask questions about the demo API and get cited answers, read the measured results, and fetch the demo API's OpenAPI spec. Models are simulated, so answers are extracts from the documents. Nothing you send is stored.

Base URL: https://veroniqa.vercel.app/api
OpenAPI 3.1: /openapi.json (also YAML)
API catalog (RFC 9727): /.well-known/api-catalog
Official CLI: veroniqa (Python, no dependencies)

Quickstart

Check the service, then ask a question:

curl https://veroniqa.vercel.app/api/v1/status

curl -X POST https://veroniqa.vercel.app/api/v1/ask \
  -H 'Content-Type: application/json' \
  -d '{"question": "What is the p95 latency target for the order list?"}'

Or use the CLI:

pipx install "git+https://github.com/Zapkid/agentqa#subdirectory=clients/veroniqa-cli"
veroniqa status
veroniqa ask "What is the p95 latency target for the order list?"
veroniqa results --json

Run it once without installing: uvx --from "git+https://github.com/Zapkid/agentqa#subdirectory=clients/veroniqa-cli" veroniqa status

Authentication and API keys

None. The public API is a sandbox for the demo, so there are no accounts or keys; each client is rate limited instead. To test your own API with real models, run VeroniQA yourself from the repository.

Endpoints

method	path	what it does
`GET`	`/api/`	API index
`GET`	`/api/v1/status`	Service status
`GET`	`/api/v1/results`	Measured results
`POST`	`/api/v1/ask`	Ask about the demo API
`GET`	`/api/v1/demo/openapi.json`	OpenAPI spec of the demo API

Request and response schemas are in the OpenAPI description.

Errors

Every error under /api is an RFC 9457 problem document, served as application/problem+json, with a stable code and a hint:

{
  "type": "https://veroniqa.vercel.app/developers#error-invalid_request",
  "title": "Invalid request",
  "status": 422,
  "detail": "The request body is not valid.",
  "instance": "/api/v1/ask",
  "code": "invalid_request",
  "hint": "Fix the fields listed in `errors`; the request schema is in /openapi.json.",
  "errors": [
    {
      "field": "question",
      "message": "Field required"
    }
  ]
}

code	status	title	how to fix
`not_found`	404	Not found	Check the path against the OpenAPI description at /openapi.json.
`method_not_allowed`	405	Method not allowed	Use the method listed for this path in /openapi.json (see the Allow header).
`invalid_request`	422	Invalid request	Fix the fields listed in `errors`; the request schema is in /openapi.json.
`rate_limited`	429	Too many requests	Wait for the number of seconds in Retry-After, then retry.
`internal_error`	500	Internal error	Retry later; if it persists, open an issue at https://github.com/Zapkid/agentqa/issues.

Rate limits

default: 60 requests per 60 seconds per client
ask: 10 requests per 60 seconds per client

Every response carries RateLimit-Policy and RateLimit headers, for example RateLimit: "default";r=59;t=60. A 429 rate_limited response adds Retry-After.

Sandbox

The interactive app: /talk/ (no sign-up; a private, temporary workspace)
The demo API that VeroniQA tests: its OpenAPI spec

Versioning

The API is versioned in the path (/v1); the current release is 1.0.0. Breaking changes get a new version path.