Skip to content
Odyssey

Start here

Introduction

What Odyssey does, and how a request travels through it.

Odyssey is one HTTP API in front of the major model families. You send a request to https://odysseyapi.tech/v1 with one key, and Odyssey runs the model you asked for, exactly that model and nothing else, and bills every request to one account.

What Odyssey is

  • Every major shape. OpenAI chat completions and Responses, Anthropic Messages, and image generation, so existing SDKs and coding agents work with a changed base URL.
  • One credential. A single Odyssey key replaces per-provider keys, and you can limit what each key may call and spend.
  • No surprises. The model you name is the model that answers. Odyssey never swaps it for another one behind your back.
  • One record of what happened. Every request, its latency, tokens and cost end up in the same request log and usage reports.

How a request travels

  1. Your key is checked, along with the models and spend limits attached to it.
  2. Your balance or subscription, and your rate limit, are checked.
  3. The request is sent to the model. If the model fails or is rate limited, you get that error straight away, with a status that says whether to retry.
  4. Tokens stream back to you as they are produced, or the whole response is returned at once.
  5. The request, its timing, tokens and cost are written to the request log and drawn from your credit balance.

Only output that was delivered is billed. Failed requests appear in the log with their status so you can see what happened, but they don't cost anything.

What Odyssey doesn't do

  • Odyssey doesn't train models, and it never substitutes a different model for the one you requested.
  • It doesn't rewrite your prompts, inject system messages or change sampling parameters.
  • It doesn't store the content of your prompts or responses.

Where to go next