API Keys are one-shot on Win11 python

API Bugs
1

My API key can be used once and once only on my Win11 system but are OK on my Lambda’s. HELP!!! Much appreciated.

2

You don’t give a lot of information where information is needed, but we can go on an assumption: you are losing the API key setting somehow.

The best way to employ API keys is by the use of environment variables, set in the operating system. OpenAI libraries, and most example code, will have you set OPENAI_API_KEY.

With Windows 11, you should just be able to type ‘environment variables’ in the start menu (or win_key + pause/break), and it will take you to “Settings->About”, where “Advanced system settings” has an “Environment Variables” button.

Name a new one OPENAI_API_KEY, and place your key obtained from the platform site there.

That will ensure the API keys are always present, and will be used by default without needing to specify them with the OpenAI libraries for Python or node.js.

3

AH, yes, I realized that I had given little to no information (this issue has made me sad)

  1. I have hard-coded the API into the test code
  2. I can access the API once without issue
  3. next and subsequent attempts result in an
  4. openai.AuthenticationError: Error code: 401 - {‘error’: {‘message’: ‘Incorrect API key provided: [api-key] ar. You can find your API key at https://platform.openai.com/account/api-keys.’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘invalid_api_key’}}

error.

Many thanks

4

Are you running in concert with a platform like Github where you are exposing your keys to the public by sync?

Are they disappearing from the user interface on the platform site?

OpenAI automatically deactivates leaked and misused keys, scanning and in agreement with several cloud providers.

5

No, it’s in an IDE behind a firewall. It’s a private system. They are still on the platform site. That OpenAI could be deleting them is interesting, how would I know if that were the case?

6

That is different than you first described, where one might think you are running Python locally, for example.

Storing keys outside of code, in a key management solution by the provider to instill the environment variables if remote, is still the path to pursue even if you don’t understand what is going on.

I would refresh and watch the API key listings in the web pages if they can never be reused…

Another source of connection problems, that would generally result in a refusal, email warnings, and shutoff, is API connections being made to OpenAI from behind a prohibited country.

It is also possible that if you are using someone else’s software that is attempting to process or store keys, or by your own code foibles, that it doesn’t recognize the new and very long API key format, and/or is damaging the contents of where they are stored for reuse.

You can reproduce client.api_key values after running Python each time using the openai library, and see exactly what is being attempted.

Then find where you might be doing something like:

JavaScript

let apikey = "my-secret-key";

// User intends to check if `apikey` is an empty string, 
// but mistakenly assigns it with =.
if (apikey = "") {
    console.log("API key is empty, please set it.");
}
1 Like
7

The cause was the API Key being cached, from what I can see at OpenAI’s end. Tests:

  1. Failed in my python IDE.
  2. Same key failed on an EC2 proxy.
  3. Error message for both showed a different, deleted key that was nowhere in my codebase.

The Culprit:
I was using:
client = openai.Client()

which was picking an API Key out of… somewhere and ignoring the one I gave it by setting an env and setting it directly with .

SOLUTION:
I used this code:
openai.api_key = self.api_key
response = openai.chat.completions.create(
…

which used the correct key and, problem solved.

I would like to thank you for listening, helping and encouraging me with this “feature” as it had me a little spooked.