Reference

API Reference

Every route the app uses internally, grouped by resource. All routes are relative to the app's origin, accept and return JSON, and require no authentication.

Base URL

In the preview, that's https://preview.devx-ai.xyz. Locally it's whatever host your dev server runs on (typically http://localhost:3000).

Resources

Resource	Covers
Datasets	Golden test case collections.
Evaluations	Runs, comparisons, and the real scoring engine.
Traces	Logged LLM calls with spans and metadata.
Prompts	Versioned prompt templates.
Annotations	Human ratings on traces and results.
Alerts	Threshold rules and triggered events.
Playground	Interactive prompt runs.

Conventions

Successful reads return 200 with the resource wrapped in a named key, e.g. { "dataset": {...} }.
Successful creates return 201 with the created resource.
Nothing is ever paginated — every list route returns the full collection, since datasets are small and mock.
Dynamic segments (:datasetId, :runId, etc.) are the resource's id field.

Errors

Errors return a JSON body with a single error string and an appropriate status code:

Status	Meaning	Example cause
400	Bad request	Missing a required field, or an empty dataset.
404	Not found	The id in the URL doesn't match any resource.
500	Server error	Unexpected failure — check server logs.

{ "error": "Dataset not found" }

No auth, no rate limits

This is a demo workspace over an in-memory store — there's no API key, no session, and no throttling. See About for what that means for persistence.

Try it now

New to the API? Start with the Quickstart, which walks through triggering a real evaluation run with a single curl command.