Authentication
Last updated
Last updated
OpenAI’s documentation provides thorough information about both the authentication process and how to set up calls.
External page: OpenAI API reference | Authentication
OpenAI uses a to . Imagine that you are entering a restricted area, where a username and password is needed to enter. A token is essentially a ticket that authorizes you to access that area without the username and password. The ticket verifies both your identity, and ensures that they have permission to access the requested resources.
Transferring this metaphor to the app world, the token can be used to send requests to the API, without having to send the OpenAI username and password each time. This not only makes it easier, but it makes the whole process more secure. After all, a token can have specific permissions, and those permissions can be changed. The token can be revoked completely and a new one generated if it somehow ends up in the wrong hands. Also, if someone were to get access to your token, they won’t be able to log in to your OpenAI account with it – it only gives access to API calls.
Phrase | Meaning |
---|---|
The bearer token is a common way to use secret API keys. You can confirm its use by checking the OpenAI API reference, in the Authentication section.
External page: Authentication | OpenAI Reference
On this page, we find the following. Take note of the Bearer in the code section at the bottom of the screenshot below:
Ok, so OpenAI API uses a bearer token for authentication, and the token is the API key we just generated on the OpenAI platform. By authenticating with that specific API key, OpenAI not only knows who we are, but can use the Permissions we set on that API key to know what resources we are authorized to access.
The bearer token (API key) is included in the HTTP header of an API call. The header can be seen as the envelope for the call. It carries important details about the message being sent, but not the message itself. Including the API key in the header is a secure way to identify and authorize the sender of the request without exposing sensitive information directly in the message body.
Let’s return to the screenshot above again, and see what OpenAI says about what the header should look like.
What can we learn from this?
OpenAI wants the authentication to be included in the header of the API call
The OpenAI will expect a , structured as:
Key: Authorization
Value:
Let’s see how that looks in the API Connector:
To set up authentication in the header, follow these steps:
First, add a new API to the API Connector, and give it a suitable name, such as OpenAI. Then:
In the Authentication dropdown, choose Private key in header.
In the Key name, make sure it says Authorization
.
In the Key value field, type in Bearer
After a space, insert the API key
your generated earlier (replacing API_KEY in the screenshot)
The short process above means the following:
In Step 1, we Instruct Bubble to include a private key token in the header of the call
Private means it’s encrypted, and not visible to your app’s users (even in the app’s JavaScript code)
In step 2 and 3, we set up the key-value pair that matches the structure OpenAI expects:
The Key is called Authorization
The Value is Bearer, [space]
and your API key
With that, we have translated OpenAI’s instructions into the language of the API Connector.
The natural next step is to test that the authentication works. But to do this, we first need to set up a call. Without it, we don’t have an endpoint to send the request to.
Let's jump into setting up our first Call.
Article: OpenAI Calls
Bearer token
The method used for authenticating
API key
The token itself