Error 422: Unprocessable Entity

[Mortezanavidi]

Bug description i get this response from the docker image or g4f library whenever i send a chatcompletion request to the api? this is the error:

openai.InternalServerError: Error code: 500 - {'error': {'message': 'ChatCompletionsError: RequestsError: HTTP Error 422: Unprocessable Entity'}, 'model': 'deepinfra/airoboros-70b', 'provider': 'DeepInfra'}]

even with cURL i get the below error:

{
    "error": {
        "message": "ChatCompletionsError: RequestsError: HTTP Error 422: Unprocessable Entity"
    },
    "model": "deepinfra/airoboros-70b",
    "provider": "DeepInfra"
}

this is not related to deepinfra, this happens with any model beside gpt-*

Environment

• python version
• location USA

[Mortezanavidi]

weirdly enough, openai library and cURL give error but chatbox works perfectly fine with the models

[hlohaus]

Update to version 0.2.1.9. The last update have fix for this issue.

[Mortezanavidi]

Update to version 0.2.1.9. The last update have fix for this issue.

i am already using the latest version which is hlohaus789/g4f:0.2.1.9

[hlohaus]

Maybe your request is incomplete. Model and messages are required. If you forget them, is the error valid.

[Mortezanavidi]

Maybe your request is incomplete. Model and messages are required. If you forget them, is the error valid.

this is the cURL i am using:

curl -XPOST -H 'Authorization: Bearer sk-ANYTHING' -H "Content-type: application/json" -d '{
  "model": "airoboros-70b",
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant."
    },
    {
      "role": "user",
      "content": "Who won the world series in 2020?"
    },
    {
      "role": "assistant",
      "content": "The Los Angeles Dodgers won the World Series in 2020."
    },
    {
      "role": "user",
      "content": "Where was it played?"
    }
  ]
}' 'http://127.0.0.1:7500/v1/chat/completions'

[Mortezanavidi]

Maybe your request is incomplete. Model and messages are required. If you forget them, is the error valid.

this is the cURL i am using:

curl -XPOST -H 'Authorization: Bearer sk-ANYTHING' -H "Content-type: application/json" -d '{
  "model": "airoboros-70b",
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant."
    },
    {
      "role": "user",
      "content": "Who won the world series in 2020?"
    },
    {
      "role": "assistant",
      "content": "The Los Angeles Dodgers won the World Series in 2020."
    },
    {
      "role": "user",
      "content": "Where was it played?"
    }
  ]
}' 'http://127.0.0.1:7500/v1/chat/completions'

which if i just change airoboros-70b to gpt-4, everything works as expected but anything beside gpt-*, gives the stated error, on g4f:0.2.1.9

[hlohaus]

Ok, only not GPT models. I will have a look.

Maybe you have to upgrade some packages. I can use the API with all models on my phone.

[hlohaus]

And can you look in the logs? You can do it with this command:

docker logs

Or this command:

docker-compose logs

[github-actions[bot]]

Bumping this issue because it has been open for 7 days with no activity. Closing automatically in 7 days unless it becomes active again.

[github-actions[bot]]

Closing due to inactivity.

[IAkumaI]

I have this response almost randomly (but about 80% of requests) to gpt-3.5-turbo. Checked with many providers. These response 422 Unprocessable Entity most often: FreeGpt ChatgptX ChatgptAi Chatgpt4Online ChatForAi

Tried with/without proxy. No matter.

Any solution?

[IAkumaI]

Tried the latest v0.2.8.0 - didn't help

[IAkumaI]

Logs in docker: INFO: 172.17.0.1:59932 - "POST /v1/chat/completions HTTP/1.1" 422 Unprocessable Entity (yes, that's all)

NodeJS openai lib error message: UnprocessableEntityError: 422 status code (no body)

[IAkumaI]

@xtekky not sure if you will see a closed issue message

[hlohaus]

Do you have you a example for a request body to the API?

[IAkumaI]

I think i found the problem. I get the error when I send multiple user-roled messages.

For example:

const openai = new OpenAI({
    baseURL: 'http://my-server:1337/v1',
    apiKey: 'rubbish string',
});

// This works fine
const chatCompletion = await openai.chat.completions.create({
        model: 'gpt-3.5-turbo',
        provider: 'FreeGpt',
        temperature: 0.5,
        messages: [
            {
                role: 'user',
                content: 'Say hello',
            },
        ],
    });

// This throws error
const chatCompletion = await openai.chat.completions.create({
        model: 'gpt-3.5-turbo',
        provider: 'FreeGpt',
        temperature: 0.5,
        messages: [
            {
                role: 'user',
                content: 'Say hello again',
            },
            {
                role: 'user',
                content: 'Say hello',
            },
        ],
    });

InternalServerError: 500 RequestsError: HTTP Error 400: 
    at Function.generate (/Users/akuma/projects/gpt-test/node_modules/openai/src/error.ts:95:14)
    at OpenAI.makeStatusError (/Users/akuma/projects/gpt-test/node_modules/openai/src/core.ts:383:21)
    at OpenAI.makeRequest (/Users/akuma/projects/gpt-test/node_modules/openai/src/core.ts:446:24)
    at processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async main (/Users/akuma/projects/gpt-test/index-test.ts:14:28) {
  status: 500,
  headers: {
    'content-length': '106',
    'content-type': 'application/json',
    date: 'Sat, 06 Apr 2024 08:24:46 GMT',
    server: 'uvicorn'
  },
  error: { message: 'RequestsError: HTTP Error 400: ' },
  code: undefined,
  param: undefined,
  type: undefined
}

This time in the console: https://dropover.cloud/5c6e69 (Sorry, can't copy it as a text)

[IAkumaI]

Now this is a 400 code, not 422. Have no idea why. Yesterday all evening I was getting 422 from almost every provider.

[IAkumaI]

Oh my god, I'm so sorry. I found the problem.

Accidentally, undefined-s got into my message array. This is what is causing the problem.

const chatCompletion = await openai.chat.completions.create({
        model: 'gpt-3.5-turbo',
        provider: 'FreeGpt',
        temperature: 0.5,
        messages: [
            {
                role: 'user',
                content: 'Say hello again',
            },
            undefined,
            {
                role: 'user',
                content: 'Say hello',
            },
        ],
    });

And we will get UnprocessableEntityError: 422 status code (no body) error.

In fact, this is a user flaw, but that's would be good if you can fix it on the script side, because official openapi does not show any errors, that's why I hadn't found it earlier.

[hlohaus]

Hello, I'll try to make the UnprocessableEntityError response better.

[hlohaus]

I made the error message better.