ServiceNow with ChatGPT Integration

[Pavankumar_1]

It's time to integrate ServiceNow with ChatGPT. Hope it will help you.

Reference links:

•  Endpoint, headers and JSON link: https://platform.openai.com/docs/api-reference/making-requests

Refer below: 

curl https://api.openai.com/v1/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_KEY" \
-d '{"model": "text-davinci-003", "prompt": "Say this is a test", "temperature": 0, "max_tokens": 7}'

• API key Creation Link: Need to create API key using below link and use it on header.

         https://platform.openai.com/account/api-keys

Refer below:

Below are the 3 steps required to complete this integration.

1. Create a ChatGPT table with Question, Answer, Request, Response, Status code are string fields.

2. Create REST Message and Post HTTP method to update the data on ServiceNow.

3. Create Async business rule and use given script and give your names.

Step 1. Create ChatGPT table and below are the fields which I have created. Create your own fields.

Step 2: REST Message and Post HTTP method with No authentication on Authentication type.

Note: Authorization should be  Bearer YOUR_API_KEY

Rest Message:

POST HTTP Method:

Content:

{
"model": "${model}",
"prompt": "${prompt}",
"max_tokens": ${max_tokens},
"temperature": ${temperature}
}

Step 3. Async Business rule insert/update on the ChatGPT table and give specific condition as below example.

Ex: Question changes(condition)

Script:

(function executeRule(current, previous /*null when async*/ ) {
    try {
       
        var r = new sn_ws.RESTMessageV2('ChatGPT', 'POST');//give your rest mesage name and method name
        r.setStringParameterNoEscape('model', "text-davinci-003");
        r.setStringParameterNoEscape('prompt', current.u_question);//give your field name
        r.setStringParameterNoEscape('max_tokens', 100);
        r.setStringParameterNoEscape('temperature', 0);

        var response = r.execute();
        var responseBody = response.getBody();//it will give the response body
        var code = response.getStatusCode();//get status code

        var responseObj = JSON.parse(responseBody);
        current.u_request = r.getRequestBody();//it will give the request body
        current.u_response = responseBody;
        current.u_status_code = code;
        current.u_answer = responseObj.choices[0].text;//it will give the text from response
        current.setWorkflow(false);
        current.update();
        current.setWorkflow(true);

    } catch (ex) {
        var message = ex.message;
    }

})(current, previous);

Screenshot:

Results: when you give question and Save on ChatGPT table then we will get the response. 

Will get Response as below.

To make to easy and understandable to everyone, I have provided the all the details with screenshots.

[Mike322]

Maybe it's an idea to post the update set with the configuration and a little instruction on how to personalize it for those who deploy said update set (like pasting your own API key on the appropriate places)?

[Sahil Khan]

Please share technical document if you have

[johndoh]

https://www.servicenow.com/community/virtual-agent-nlu-forum/integrate-azure-openai/td-p/2600944

[sushma9]

Hi @Pavankumar_1 

 I have tried integrating chat gpt with service now  When i tested it i am getting below response as  error 

Please find the Rest Message and Http Method and Business rule screen shots below.

Rest message :

HTTP Method : 

BR:

[johndoh]

Hello @sushma9 ,

You need to populate the values for the variables you added in your HTTP Method screenshot:

Generate the variables and populate similar values as below:

Populate the variables with your requirements or anything for that matter as the BR you created will override but needs a sample:

[johndoh]

You can find which model is appropriate for you here Models - OpenAI API

[sushma9]

Hi @johndoh 

 I am going to use the model as  dynamic model and i am defining them in BR. Do i Need to again use these values  here and i have given those values  here and tried as well but i am getting below error message response.

[johndoh]

@sushma9 remove the quotes around 100 for the max tokens and the quotes around the 0 in temperature and test again. You should be successful then 🙂

[Pavankumar_1]

Hi @sushma9 ,

max_tokens and temperature are integers give as below it will work not as string.

{
"model": "text-davinci-003",
"prompt": "What is Chat GPT?",
"max_tokens": 100,
"temperature": 0
}

[niveditakumari]

Hi @Pavankumar_1, 

I have created API key with using https://platform.openai.com/account/api-keys. 

I'm creating ChatGPT table. 

Can you please tell about ChatGPT table and how to create that. Is this different table or same table. 

Please help me to start and achieve it. I'm really very interested to do this integration using ChatGPT. 

Regards, 

Nivedita

[Pavankumar_1]

Hi @niveditakumari ,

just create any test table and create those fields and use it.

[niveditakumari]

Hi @Pavankumar_1, 

Can we connect? 

Regards, 

Nivedita

[niveditakumari]

Hi @Pavankumar_1, 

It worked. 

Thank you so much for this useful article. 

Regards, 

Nivedita

[Gayathri Anba]

Is this possible that we can integrate this without global search/outbound calls here? Want to answer inside servicenow instance by using chatGPT

[RodGallegos]

Next step, plug VA into the mix! Nice post!

[Pavankumar_1]

Hi @Gayathri Anba ,

without rest call how ServiceNow will get the result?

[Pavankumar_1]

Hi @RodGallegos ,

Thanks @RodGallegos . I tried with VA as well.

[Gayathri Anba]

Hi @Pavankumar_1 , I tried with VA it is working with REST calls. But I need to use CHATGPT for Private calls(basic contextual search inside the private instance) not for Global search. Is that Possible ? 

[taofeek93]

Hi @Pavankumar_1 , I keep getting HTTP Status 400 when I tested it. 

 

 

[Pavankumar_1]

Hi @taofeek93 ,

you are not proving any test value on variable so try to give some values some do test.

please refer below screen and try again

you will get the response as below

[Fatima Zahra At]

@Pavankumar_1  

when I tested with the values i got this error message. can you help me please to solve it 

[Pavankumar_1]

Hi @Fatima Zahra At ,

you have exceeded you limit. Please check below link for limits

https://platform.openai.com/account/limits

You have to buy credits or try with different account.

[Danny Mortensen]

gpt-3.5-turbo

Thanks for this example Pavan. It is a great starting point for tinkering with Generative AI. This is my first integration attempt with ServiceNow .. I've not succeeded yet but your example is pretty clear.
I'm not getting a message to say I've exceeded my current quota but I assume that is possibly because I need to pay now or create a new account. I'll give that a go.

Just FYI .. InstructGPT model "text-davinci-003" is deprecated and replaced with "gpt-3.5-turbo-instruct". There might be more suitable models for your purpose so check them out here: https://platform.openai.com/docs/models

[senator013]

What is the business value of this integration?  It's not clear to me why I would do this.

[SylvainHauserN2]

For the ones whowould like to make it work with gpt-3.5-turbo, here is the updated script after having updated the endpoints to "https://api.openai.com/v1/chat/completions":

(function executeRule(current, previous /*null when async*/ ) {
    try {
       
        var r = new sn_ws.RESTMessageV2('ChatGPT', 'POST');//give your rest mesage name and method name
        r.setStringParameterNoEscape('model', "gpt-3.5-turbo"); //https://platform.openai.com/docs/guides/text-generation

		var messages = [
            { role: "user", content: current.u_question.toString() }   // Add the current question
        ];

		var requestBody = {
            model: "gpt-3.5-turbo",
            max_tokens: 100,
            temperature: 0,
            messages: messages // Adding the messages array to the request body
        };

		r.setRequestBody(JSON.stringify(requestBody));
		
		r.setStringParameterNoEscape('max_tokens', 100);
        r.setStringParameterNoEscape('temperature', 0);

        var response = r.execute();
        var responseBody = response.getBody();//it will give the response body

        var code = response.getStatusCode();//get status code

        var responseObj = JSON.parse(responseBody);
        current.u_request = r.getRequestBody();//it will give the request body

        current.u_response = responseBody;
        current.u_status_code = code;
        //current.u_answer = responseObj.choices[0].text;//it will give the text from response
		current.u_answer = responseObj.choices[0].message.content;
        current.setWorkflow(false);
        current.update();
        current.setWorkflow(true);

    } catch (ex) {
        var message = ex.message;
    }

})(current, previous);

[Souvick6917]

Hello @Pavankumar_1 

First , thank you for this article. But with everything configured as required I am getting below error which is somewhat new to me. Can you please help me with the issue:

{
"error": {
"message": "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.",
"type": "insufficient_quota",
"param": null,
"code": "insufficient_quota"
}
}

Screenshots are also added.

[fathahhkf]

@Pavankumar_1& everyone,

New update on the post.

Replace the below to make integration works,

End Point: https://api.openai.com/v1/chat/completions

Content:

{
    "model": "gpt-4o-mini",
    "messages": [
      {
        "role": "user",
        "content": "${content}"
      }
    ]
  }

Thanks,

Abdul Fathah

[MadhusaiK]

i have followed the above steps you mentioned and still unable to get response 
could you please help me with that 
i am attaching the configuration screenshots.

[MadhusaiK]

this error i am getting

[Aniket Shukla]

Hi @Pavankumar_1 , Needed help  i am getting this error.

 and also i have attached 4 scrennshot of rest ,http method with br .

Please Help @Pavankumar_1 

[Pavankumar_1]

@Aniket Shukla 

Maybe that is related to the license, and you reached your limit. Try with other account