# ORVYX > ORVYX is an AI gateway that exposes more than 500 language, vision and reasoning models behind a single OpenAI-compatible endpoint, one API key and one bill. Base URL: https://api.orvyx-ai.com/v1. Models are addressed as orvyx/. Authentication is a Bearer API key. ## Documentation (31 pages, in navigation order) - [ORVYX](https://docs.orvyx-ai.com/) — Since ORVYX speaks the same language as OpenAI, switching over is one address and one key. Nothing else in your project changes. Here is a support assistant that streams an answer, in the language you already use. - [Chat Completions](https://docs.orvyx-ai.com/api/chat-completions) — POST /v1/chat/completions1Creates a model response for a conversation. Supports streaming, vision inputs and tool calling. - [Embeddings](https://docs.orvyx-ai.com/api/embeddings) — POST /v1/embeddings1Creates vector representations of input text. - [Errors](https://docs.orvyx-ai.com/api/errors) — Every ORVYX error uses one OpenAI-compatible envelope, carries a stable code, and includes the request id for support. - [API Reference](https://docs.orvyx-ai.com/api/index) — The ORVYX public API is OpenAI-compatible. Every customer endpoint lives under a single base URL and speaks the same contract your OpenAI client already understands. - [Models](https://docs.orvyx-ai.com/api/models) — GET /v1/models1Lists every model available to your account. This endpoint requires no authentication,it is safe to call as a connectivity probe and during an incident. - [Rate Limits](https://docs.orvyx-ai.com/api/rate-limits) — Rate limits protect both your account and the platform. When you exceed a limit ORVYX returns 429 with a retry-after value,it never silently drops or queues your request indefinitely. - [Billing & Credits](https://docs.orvyx-ai.com/concepts/billing) — ORVYX uses a prepaid credit model. You top up a balance, requests draw it down, and you can see exactly what every request cost. This page explains how the money side works so there are no surprises. - [Model Catalog](https://docs.orvyx-ai.com/concepts/model-catalog) — ORVYX exposes a catalogue of models through one API. This page explains how to browse it and how to pick a model for a task. - [Reliability](https://docs.orvyx-ai.com/concepts/reliability) — ORVYX is built so that routine upstream instability does not become your incident. This page explains the guarantees you can rely on. - [Security](https://docs.orvyx-ai.com/concepts/security) — This page describes the security properties you can rely on when building on ORVYX, and what you are responsible for on your side. - [Chat Completion](https://docs.orvyx-ai.com/examples/chat-completion) — POST /v1/chat/completions sends a conversation to a model and returns the assistant reply. This is the request most integrations build on. - [Embeddings](https://docs.orvyx-ai.com/examples/embeddings) — POST /v1/embeddings turns text into vectors you can compare, cluster or search. It follows the OpenAI embeddings contract. - [Examples](https://docs.orvyx-ai.com/examples/index) — Working, copy-paste recipes for the ORVYX API. Every snippet runs against the same base URL and the same key, so you can move between them freely. - [Streaming Responses](https://docs.orvyx-ai.com/examples/streaming) — Set "stream": true to receive the answer token by token over Server-Sent Events instead of waiting for one JSON body. Use it whenever a human is watching the output. - [Structured Output](https://docs.orvyx-ai.com/examples/structured-output) — Ask a model for strict JSON instead of prose by setting response_format. This is the reliable way to feed model output into code. - [Tool Calling](https://docs.orvyx-ai.com/examples/tool-calling) — Let the model call functions you define. You declare tools, the model asks to call one, you run it and feed the result back. This is how you connect a model to your own data and actions. - [API Keys](https://docs.orvyx-ai.com/guide/api-keys) — An API key is how your application authenticates to ORVYX. Keys are created in the dashboard under API Keys and shown once at creation. ORVYX cannot reveal the plaintext again, so store it immediately. - [Chat Completions](https://docs.orvyx-ai.com/guide/chat-completions) — POST /v1/chat/completions is the primary endpoint. It is OpenAI-compatible and supports both streaming and non-streaming responses, vision inputs and tool calling. - [Configuration](https://docs.orvyx-ai.com/guide/configuration) — This page covers everything you configure as an ORVYX user. There is very little of it: a base URL, a key, and optionally a couple of headers. - [Embeddings](https://docs.orvyx-ai.com/guide/embeddings) — POST /v1/embeddings turns text into vector representations. It follows the OpenAI embeddings contract. - [Error Handling](https://docs.orvyx-ai.com/guide/error-handling) — Every ORVYX error uses the same OpenAI-compatible envelope, carries a stable error code and includes the request id for support. - [Installation](https://docs.orvyx-ai.com/guide/installation) — ORVYX is OpenAI-compatible, so installing ORVYX means installing an OpenAI client and pointing it at ORVYX. There is no ORVYX-specific SDK to learn. - [Introduction](https://docs.orvyx-ai.com/guide/introduction) — ORVYX is an AI gateway: one OpenAI-compatible API in front of a large catalogue of AI models. You integrate once, and ORVYX handles model availability, authentication, billing, rate limiting and usage analytics for you. - [Key Concepts](https://docs.orvyx-ai.com/guide/key-concepts) — A short vocabulary. Skim it before reading the API reference,a handful of terms carry most of the meaning. - [Quickstart](https://docs.orvyx-ai.com/guide/quickstart) — This page gets you from zero to a first response. It assumes you already have an ORVYX API key (oryx_…); if not, create one in the dashboard under API Keys. - [Streaming](https://docs.orvyx-ai.com/guide/streaming) — ORVYX supports streaming responses on POST /v1/chat/completions using Server-Sent Events (SSE), matching the OpenAI streaming contract. - [Tool Calling](https://docs.orvyx-ai.com/guide/tool-calling) — ORVYX supports OpenAI-style tool (function) calling. You declare tools, the model may ask to call one, you execute it and feed the result back. - [FAQ](https://docs.orvyx-ai.com/resources/faq) — Yes. The endpoints under https://api.orvyx-ai.com/v1 implement the OpenAI contract, so the official OpenAI SDKs work by changing only the baseURL and the API key. See Migrating from OpenAI. - [Migrating from OpenAI](https://docs.orvyx-ai.com/resources/migrating-from-openai) — ORVYX speaks the OpenAI contract, so migrating is usually a two-line change: point the client at the ORVYX base URL and use an ORVYX API key. - [Troubleshooting](https://docs.orvyx-ai.com/resources/troubleshooting) — Common problems and how to diagnose them. Every response carries an x-request-id; quote it when contacting support, as it ties the failure you saw to the exact request on the server. ## Machine-readable - Every page as Markdown: append `.md` to its URL, e.g. https://docs.orvyx-ai.com/guide/quickstart.md - The whole documentation in one file: https://docs.orvyx-ai.com/llms-full.txt - Machine-readable index of every page: https://docs.orvyx-ai.com/llms-manifest.json