irudnyts / openai Goto Github PK

First of all, thanks for you effort in doing this library! Great work.

I'm trying to repeat the keyword example. Like it does in the sandbox, it should return a list of keywords extracted from the prompt. Note that in my code below, I've heavily truncated the prompt to save space here.

kw_res <- create_completion(
  model = "text-davinci-003",
  max_tokens = 60,
  temperature = 0.5,
  top_p = 1,
  frequency_penalty = 0.8,
  presence_penalty = 0,
  prompt = "Extract keywords from this text:\n\nKun näitä ilmastosuunnitelmia työnnetään kunnille ja kun kaikkien kuntien — ei nyt ihan kaikkien, mutta suurimman osan kunnista — rahat ovat aivan loppu"
)

The function runs without error, but the returned list is somehow corrupted. The keyword list is only a stub, and it is preceded with text that does not seem to be in the right place.

> kw_res$choices$text
[1] "ksen taholta tehdään jotain, niin se ei ole pelkkä näennäistoimintaa vaan se on sellaista, mikä todella vaikuttaa.\n\nKeywords: ilmast"

In the sandbox, with the same argument values, the function returns a dozen keywords or so.

I wonder if the issue is somehow caused by the OpenAI API itself?

See details here.

Hi! I have been using this package, and I must say I really liked it and greatly appreciate it.

Is it possible for you to return a return code with the response indicating if the returned item is a real response or an error? After ChatGPT release the server started to give error messages instead of replies occasionally. It would be awesome if it was handled in openai package.

Thanks!

What should I do for this？
Error in curl::curl_fetch_memory(url, handle = handle) :
Timeout was reached: [api.openai.com] Connection timeout after 10007 ms

See https://platform.openai.com/docs/api-reference/audio for details.

I've been trying to loop create_chat_completion() over different message lists, and I got the following error on the third entry on the list:

Error in paste("OpenAI API probably has been changed. If you see this, please",  : 
  OpenAI API probably has been changed. If you see this, please rise an issue at: https://github.com/irudnyts/openai/issues

Per the instruction, I'm rising an issue here. Any help would be appreciated!

EDIT: I've found a workaround using the retry() function from the retry package, where I substitute the following for the function call in the loop:

retry(
    expr = create_chat_completion(text),
    when = "502|API",
    max_tries = 3,
    interval = runif(1, 6, 10)
  )

Still not entirely sure what the source of the error is, but this seems to work. It might be worth making a more informative error for the future.

See details here.

See details here.

See details here.

I get this error when I use the example command:
Error in dyn.load(file, DLLpath = DLLpath, ...) :
unable to load shared object '/Users/myname/Lokaal/R/packages/curl/libs/curl.so':
dlopen(/Users/myname/Lokaal/R/packages/curl/libs/curl.so, 0x0006): tried: '/Users/myname/Lokaal/R/packages/curl/libs/curl.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64')), '/System/Volumes/Preboot/Cryptexes/OS/Users/myname/Lokaal/R/packages/curl/libs/curl.so' (no such file), '/Users/myname/Lokaal/R/packages/curl/libs/curl.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64'))

Not compatible with M1 macs yet?

Any chance this will be added in the near future?

See details here.

Hello!

Is there any plan to implement the "function calling" feature of the API?
https://platform.openai.com/docs/guides/gpt/function-calling

See https://platform.openai.com/docs/api-reference/chat for details.

Hi- thanks so much for the package. It seems like one of the URLs called in the create_completion() function may need to be updated for gpt-3.5-turbo. When I request it as a model option, I get the following error message:

`Error: OpenAI API request failed [404]:

This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?`

If you have the bandwidth to take this on, I'd be very grateful!

As it turns out the OpenAI API frequently returns very unhelpful error messages like this one:

The server had an error while processing your request. Sorry about that!

There are also several threads on this issue, for instance in the OpenAI Develper forum.

I think some parameter like max_retries could be really useful to handle such transient errors that arise while applying functions like create_chat_completion().

The httr2 package provides the very useful req_retry() function that allows to define the maximum number of retries, which errors are transient, and how long to wait between retries. Is there any specific reason that you sticked httr for the openai package?

See details here.

Currently it does not seem to work for GPT-4 api.

See branch full-doc for more details.

See the list of models here: https://platform.openai.com/docs/models/model-endpoint-compatibility

One can use this example

See details here.

Is it necessary?

Installing CRAN version on 4.1.1

Warning in install.packages :
  package ‘openai’ is not available for this version of R

Installing Github version on 4.1.1

ERROR: this R is version 4.1.1, package 'openai' requires R >= 4.2

I have set the open API key as follows:

Sys.setenv(OPENAI_API_KEY = 'XXXX')

And proceed to try this out but got the following error:

library(openai)

create_completion(

• model = "babbage-002",
  

• prompt = "Generate a question and an answer"
  

• )
  Error: OpenAI API request failed [429]:

You exceeded your current quota, please check your plan and billing details. For more information on this error, read the docs: https://platform.openai.com/docs/guides/error-codes/api-errors.

I just logged on and check my quota and I am not over the quota at all? Is there another step that I need to do to try this out?

My API calls return an error that either says:

Error in paste("OpenAI API probably has been changed. If you see this, please",  : 
  OpenAI API probably has been changed. If you see this, please rise an issue at: https://github.com/irudnyts/openai/issues

or

Error: OpenAI API request failed [503]:

That model is currently overloaded with other requests. You can retry your request, or contact us through our help center at help.openai.com if the error persists. 

I'm sometimes able to process 3-5 calls at a time through the API, but usually, it stops after a single one. Last night I was able to make about 1,000 calls with no issues.

It seems that just need to change the base_url to a function parameter that can be passed in.
https://github.com/search?q=repo%3Airudnyts%2Fopenai%20https%3A%2F%2Fapi.openai.com%2Fv1&type=code

Reference: https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?tabs=powershell&pivots=rest-api

The headers of the response contain information about the remaining requests and tokens before limits are exceeded.

Returning the response headers along with the content (or just the entire response as is) could give access to users which need to slow down their requests.

How can I see the image created by create_image?

See details here.

Hi, great package.
For the sake of reproducibility and for ease of continuing a chat, could the create_chat_completion() returned result object include the inputted settings (in particular the messages argument)?

I am not sure wether it is a bug or not:
When using:
response <- create_chat_completion(
model = "gpt-3.5-turbo",
......

we set the model to "gpt-3.5-turbo",
but when we see the results:

response
$id
[1] "chatcmpl-6tat7xkifZsnPUi3t1C0SePRpZQ6A"

$object
[1] "chat.completion"

$created
[1] 1678707857

$model
[1] "gpt-3.5-turbo-0301"

Here it shows the model "gpt-3.5-turbo-0301"

I am not sure if it is correct.

I am trying to use openai but when I include my API keys it cannot be recognized.

Sys.setenv(
    SPOTIFY_CLIENT_ID = 'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx'
)

create_completion(
    engine_id = "ada",
    prompt = "Generate a question and an answer"
)

Error: OpenAI API request failed [401]:

You didn't provide an API key. You need to provide your API key in an Authorization header using Bearer auth (i.e. Authorization: Bearer YOUR_KEY), or as the password field (with blank username) if you're accessing the API from your browser and are prompted for a username and password. You can obtain an API key from https://beta.openai.com.

Can I help on getting the stream functionality in?

See details here.

API was recently updated to return logprobs for chat completions but the change appears to be only present in the python version and not the R version. Would really appreciate everything up to date and ported over!

This is the error I get

Error in fetch(key) :
lazy-load database '/Library/Frameworks/R.framework/Versions/4.3-arm64/Resources/library/openai/help/openai.rdb' is corrupt

You can see the screenshot also. I removed and reinstalled the library a couple of time, but this keeps happening.

Thanks

It would be great to use JSON mode, yet even with the development version, I don't see how to pass that parameter. Could that maybe be added?

https://platform.openai.com/docs/guides/text-generation/json-mode

See details here.

A number of open source models like LLaMa 2 can run in local environments where there's a webserver (like LM Studio, Koboldcpp, etc.) that has identical endpoints to OpenAI. Can we have a flag/option that allows for specifying a different OpenAI-compatible endpoint?

I was running a test to compare whisper to google transcript. This was for an audio file in Spanish so called like this:

transcription <- create_transcription(file = "sound-data/29320.mp3",
                                   language = "es",
                                   model = "whisper-1"
                                   )
})

What I get back is:

"Más temprano que tarde. Cuestión Política. Daniel Ciaschetti, doctor en Ciencia Política, docente e investigador de la Universidad de la República. Daniel Ciaschetti, doctor en Ciencia Política, doctor en Ciencia Política, doctor en Ciencia Política, doctor en Ciencia Política, doctor en Ciencia Política,

and so on for a couple of hundred lines. It seems it transcribes the first Sentence, and then got stuck, just repeating the last part of it for the rest of the transcript. I am not sure if this is a problem with the R-package or a bug in the API.