Using the APIs

Introduction

Start here for auth, base URLs, and the shape of every request and response.

Overview

Redeo exposes two families of API endpoints, both served from https://api.redeo.ai/v1:

Direct Model Access — POST /v1/chat/completions passes your prompt straight to an LLM provider (OpenAI, Anthropic, etc.) using your own API key. It behaves identically to calling the provider directly. Use it when you want a single LLM response with no orchestration.

Stilt Execution — POST /v1/{author}/{stilt}/chat/completions runs a LukiScript stilt, which may involve multiple steps, loops, gates, and parallel calls. The response is returned in the same OpenAI-compatible format, so drop-in replacement works.

Both endpoints share the same authentication, error format, and response shape.

API Endpoint

Base URL: https://api.redeo.ai/v1

All requests must use HTTPS. Requests over HTTP are rejected.

Authentication

All API endpoints require authentication via a bearer token. Create an API key from your Redeo dashboard, then include it on every request:

text

Authorization: Bearer rk_live_abc123...

API keys are scoped to your account. Keys can be configured with per-stilt access controls (whitelist, blacklist, or unrestricted) from the dashboard.

Requests without a valid key receive a 401 with authentication_error.

Error Format

Errors are returned as JSON with a consistent shape:

json

{
  "error": {
    "message": "Human-readable description of what went wrong.",
    "type": "invalid_request_error",
    "code": "missing_required_field"
  }
}

Common error types: authentication_error (401), invalid_request_error (400), insufficient_quota (402), api_error (500).

Retry idempotent requests with exponential backoff on 429, 502, 503, and 504. Do not retry 4xx errors — they indicate a problem with the request itself.