Begoña Manzano Garrido 

Nice to see you trying this one out... I think you may be the first to provide any feedback!

An observation at this stage would be that you don't seem to have selected anything from the Search dropdown ad it's blank.  I'm not sure why this would be as it's always populated for me and I have choose between source or target.  Perhaps you can try selecting something first?

You show this for example:

I have this:

However, I have also noticed that the OpenAI server is so slow at the moment that multiple calls seem to result in nothing at all.  Can you also test this with the default prompt that only asks for a single translation?  That works for me.

We'll do some more testing but I have a feeling this could be a problem on the OpenAI end as they have been inundated in the last months with an unexpected demand and they need to do a lot of work to improve their ability to support the demand.

I got the same Bad Request error, but I fixed it by adding my credit card data to my account. However the "Not found" error pops up as soon as I select the GPT-4 Model.

I couldn't even add my debit card... strange!

We will investigate these errors, and interpret better the messages that are returned from the OpenAI API.  As you already noted, this is most likely related to a time-out as the gpt-3.5-turbo and especially gpt-4 models don't seem to be provisioned appropriately for public consumption at the scale they are used.

Also Consider that you need a paid account to access both gpt-3.5-turbo and gpt-4.  Additionally, some countries are not allowed access to the API entirely, like Italy.

The model text-davinci-003 (e.g. gpt 3.0) is available for free for the first 3 months and is resonably fast, so might be worth confirming if this working, just to confirm you are able to communicate with the OpenAI API.  It will help to narrow down the cause.

Hi Patrick, 

Thanks for all the details. Do you know if Spain is one of these countries not allowed to access the API? 

I get the "Bad request" error when selecting the text-davinci-003 model too. 

Let me know if you need any further details.

Thanks!

Countries such as North Korea, Iran, China, Cuba, and Syria are among the countries where ChatGPT is not accessible.

Italy banned ChatGPT over Data Privacy Concerns and talks that Germany and France will soon follow. I'm reading now that Spain opened an offical investigation into OpenAI's ChatGPT over a potential data breach in the last days...  not sure what that means exactly with general accessiblity.

One way to confirm definitively if you can access the OpenAI API from where you are located in the world, is to load the website in a browser: https://platform.openai.com/ and login

Thanks, Patrick. I could create the account and the API key so I do have access to the platform

touché :-)
let me mull over it for a bit; trying to idenfity better ways of tackling these types of problems; possibly including more information in the log that might help with more context to idnetify the root cause etc...

Patrick Andrew Hartnett 

Hi, Patrick. I'm having the same problem as Begoña. I got either "Bad request" or "Not found". Then, after I entered my credit card information in OpenAI, I got the same messages or "Unauthorized". I can log into the OpenAi platform without a problem, but not activate my API key for Trados 2022. I created several API keys and tried with all of them, and with several segments from a project too. I also filled out the "Search from" field.  

Patrick Andrew Hartnett 

Hi, Patrick. I'm having the same problem as Begoña. I got either "Bad request" or "Not found". Then, after I entered my credit card information in OpenAI, I got the same messages or "Unauthorized". I can log into the OpenAi platform without a problem, but not activate my API key for Trados 2022. I created several API keys and tried with all of them, and with several segments from a project too. I also filled out the "Search from" field.  

Hi Rocío Valero ,

There are several reasons why you are receiving this error, incorrect API key, rate limiting, account suspension, but also could be related to system requiring a proxy connection.

I've used ChatGPT to help me articulate the response, underneath. Can you review these responses and lets see if we can narrorow down the cause?

Q: Explain why a user would receive an unauthorized http response from a plugin integration with OpenAI API?

A: There are several reasons why a user might receive an "unauthorized" response when attempting to integrate with the OpenAI API. Here are some possible explanations:

1. Incorrect API Key: The most common reason for an "unauthorized" response is an incorrect API key. OpenAI requires an API key for authentication purposes, and if the key is incorrect, the API will reject the request. Developers should ensure that they are using the correct API key for their integration.

2. Missing Permissions: OpenAI's API is divided into several endpoints, each with its own set of permissions. Developers should ensure that their API key has the necessary permissions to access the endpoint they are trying to use.

3. Rate Limiting: OpenAI limits the number of requests that can be made to its API within a certain timeframe. If a developer exceeds this limit, the API will return an "unauthorized" response. Developers should ensure that they are not making too many requests within a short period of time.

4. Account Suspension: In rare cases, OpenAI may suspend a developer's account for violating its terms of service or engaging in abusive behavior. If a developer's account is suspended, they will receive an "unauthorized" response when attempting to access the API.

Q: If the user's computer requires a proxy configuration to connect to the internet, could this also be problematic when connecting to OpenAI API?

Hi, Patrick. Thank you for your reply. I'm not a developer, but a translator, and I don't understand everything you say. I don't think Spain has banned OpenAI or anything. I think the problem may be that my API keys are assigned to a different organization rather than Trados, or something like that. I would need some professional help (that's what ChatGPT suggested:). Can you suggest something in that sense? I would be grateful if you could send a DM

Hello,

from the amounts of users having same issues there's probably something not working as expected.

Is there a debug feature in the plugin that produce a log of the requests/answers?

PS: i have the same "bed request" issue. The same prompt via curl works as expected.

Hi Rocío Valero ,  I'm currently working on a new release that will enable us to track a bit more information related connectivity, with the objective of providing more insight to what is causing these types of issues.

I will circle back here once we have a build available.  It might also help me if I could follow up with one of you directly to diagnose on a computer that reproduces this issue

Thank you to everyone for your feedback; much appreciated.

Hi, Patrick. Thanks a lot for that. I'm looking forward to the new release, and I would be happy if you wish to diagnose my computer. If you wish to contact me, my email address is rociovalerol@gmail.com (with an "L" after "valero").  

Hi Rocío Valero , thank you very much for spending time with me this afternoon, in a debug session to identify the root cause of the issues reported.  I've been able to identify what is causing the underlying exception from the log file you provided.

I will work on a new release in these days to resolve this issue definitvely and circle back to you.

Amazing support, thx much (y)

Hi, Patrick Andrew Hartnett. Thank you for your patience in spite of my clumsiness in the videocall and for helping me successfully install the new release of the plug-in. I am glad that you could identify the issue. The plug-in is working well for me now. I am sure it will be very useful.