Skip to main content
Poixe AI uses an API Key to identify the caller, and to determine the available quota and optional restrictions (such as rate limits and model scope). In the console, an API Key is also often called an API Token or simply a token. They all refer to the same concept. If you haven’t created an API Key yet, go to the console to create one first:

Create a Token (Console)

Create, view, and manage your API Keys (Tokens) in the console.

​
Key balance and account balance

When you create an API Key, you can set a Key balance for it (think of it as the key’s “budget cap / available quota”).
  • The Key balance is customizable: you can set it to a very large value, even one that looks “unlimited”.
  • Actual charges are still based on your account balance: every request is constrained by both the account balance and the Key balance.

​
Two rules (important)

  1. Insufficient account balance
    Even if the Key balance is sufficient, the request will still fail (insufficient balance error).
  2. Insufficient Key balance
    When the account balance is sufficient but the Key balance is not, the key is automatically disabled to prevent it from being called further.
Recommendation: in production, split keys by service/environment and set a reasonable budget and limits for each key, making them easier to control and track.

​
How to use it in requests

Authentication fields differ slightly between protocols (header names and field formats may vary). We recommend filling in authentication details according to the documentation of the protocol you are using:

Protocols and Authentication Overview

Quickly compare the authentication methods and differences of the 4 protocols.

Errors and Troubleshooting

Start here when you run into authentication failures, insufficient balance, and similar issues.
  • Split keys by purpose: create separate keys for production, testing, and scripts to make tracking easier and isolate risk
  • Rotate and revoke promptly: disable or delete keys you no longer use
  • Don’t hard-code them: inject them via environment variables or a secrets management system (CI/CD secrets, Vault, etc.)