Guys, if you're following this video and get an error: "Action Output: answer: Empty output value cannot be blank on publish," it's because in the script include, he put "AskChatGPT" as the second parameter to his request message when it should be "askChatGPT" because that's the name of the method in the REST outbound message
Olá Filipe, excelente demo. Uma pergunta, será que daria para o ChatGPT manter o rumo da conversa? Porque ele apenas lê a mensagem inserida no momento, ou seja, "esquece" as mensagens escritas anteriormente. Seria possível ele ler todas as mensagens inseridas e dadas antes de dar uma nova mensagem? Obrigado
Hi Filipe, this is a very cool demo, thanks for sharing. Is it (yet) possible to connect ChatGPT to a ServiceNow database in order to process the internal information and reply to customer questions with responses that take that data into account? e.g. imagine ChatGPT could search resolutions of incidents that are specific to an organisation, perhaps then it could provide helpful answers to complicated application-based questions...
Hi @josephskewes ! Yes, you can do it. Please follow the link for more details on how to train your GPT-3 model with custom data: openai.com/blog/customized-gpt-3/#example Hope this helps!
Hi Filipe,Everything is looking good but how we will send attachment to the chatgpt ,The attachment icon(paper clip )also not coming in the chat when we connected to chatgpt
hi thanks for sharing this. It really helps. But when i show my client the same POV in my company. My client asked - Why should i connect it with ServiceNow VA if i can directly question the ChatGPT bot?? I had no answer. Actually people are looking for real use cases of how this bot will help in ServiceNow environment and add value to ServiceNow client?
Hello i am new to this and this was a real great video can you help me how can i set up the authorisation token where can i get the access token in Servicenow or how can that be fetched ?
HI Felipe, Thanks for the video. I'm stuck at the flow designer step where you defined the output variables. Can you tell us the exact steps you took to define the output variables? I got the output variables to populate but not in the order (answer first then action status) and I can't publish it that way and it gives me this error "Please correct the following errors before save or publish: Action Output: answer: Empty output value cannot be blank on publish."
When i test your flow i get an error. I follwed all your steps in the video and i cant get it tork. there must be something missing that wasnt shown on the video
Hello, thank you for sharing this video. When in the askchatgpt http method, when i test i get this message in the response section: error": { "message": "We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly. The OpenAI API expects a JSON payload, but what was sent was not valid JSON.)." I made sure content type is : application/json. Do you know how i fix this? thank you!
Hi @gabrielebernini1440! Go to the "askChatGPT" HTTP method, select the "HTTP Request" tab, bellow the "HTTP Query Parameters" you will find the "content" field. Make sure you have a content like the following: { "model":"${model}", "prompt":"${prompt}", "temperature":${temperature}, "max_tokens":${max_tokens} } Then, in order to test, create variable substitutions like the following: model --> text-davinci-003 max_tokens --> 1024 temperature --> 0.5 prompt --> Let me know if this way you can get a valid answer! Best Regards, Filipe
@@CeRe4L unfortunately still not working. I get the following messages when testing : HTTP status: 400 Error message : Method failed: (/v1/completions) with code: 400 Error Code: 1 i'm quite sure my http headers ( authorization and content type) are correct. Maybe is mething related to the script include? thank you a lot for your answer.
Ok, so I had the same publishing error message "Action Output: answer: Empty output value cannot be blank on publish," but eventually got this to work! 1) There are two places for the "answer" variable, one within the Inputs Script Step (I had to scroll down to see it) and one in the Outputs 2) When it gives you the error, you go back and set the default value of answer by dragging the "answer" pill from the Script Step. 3) I did have a typo! In the (Action Script Step) - the second part of this line, after the (). outputs.answer = new ChatGPT().askChatGPT(inputs.question); must exactly match the first part of this line, before the colon askChatGPT: function(message) of this line which in the (Script Include) 4) Sometimes a black arrow would show up to the left of the Action Status in Output Variables (as shown in this video). Once that appeared, it started giving me errors about Action Status "feature cannot be blank" etc. I finally decided to start from scratch, recreating the entire Action with a clean start. In my final working result, in Output Variables - I had "Action Status" on top, with no arrow on the left, and "answer" below that. This fixed my issues, gave the correct answer when testing the Action, and allowed me to Publish! Hope this helps!
I'm getting the following response while testing the REST message,i gave model parameter as text-davinci-003 .?Can you help me with this "error": { "message": "you must provide a model parameter", "type": "invalid_request_error", "param": null, "code": null
Go back and make sure you have the information for the REST message in the form correctly. seems you didn't include "text-davinci-003" as a variable for "${model}" parameter
Guys, if you're following this video and get an error: "Action Output: answer: Empty output value cannot be blank on publish," it's because in the script include, he put "AskChatGPT" as the second parameter to his request message when it should be "askChatGPT" because that's the name of the method in the REST outbound message
Olá Filipe, excelente demo.
Uma pergunta, será que daria para o ChatGPT manter o rumo da conversa? Porque ele apenas lê a mensagem inserida no momento, ou seja, "esquece" as mensagens escritas anteriormente. Seria possível ele ler todas as mensagens inseridas e dadas antes de dar uma nova mensagem?
Obrigado
Hi Filipe, this is a very cool demo, thanks for sharing. Is it (yet) possible to connect ChatGPT to a ServiceNow database in order to process the internal information and reply to customer questions with responses that take that data into account? e.g. imagine ChatGPT could search resolutions of incidents that are specific to an organisation, perhaps then it could provide helpful answers to complicated application-based questions...
Hi @josephskewes !
Yes, you can do it. Please follow the link for more details on how to train your GPT-3 model with custom data:
openai.com/blog/customized-gpt-3/#example
Hope this helps!
Hi Filipe,Everything is looking good but how we will send attachment to the chatgpt ,The attachment icon(paper clip )also not coming in the chat when we connected to chatgpt
hi thanks for sharing this. It really helps. But when i show my client the same POV in my company. My client asked - Why should i connect it with ServiceNow VA if i can directly question the ChatGPT bot?? I had no answer. Actually people are looking for real use cases of how this bot will help in ServiceNow environment and add value to ServiceNow client?
How did you configure Topic picker to show ChatGPT?
Hello i am new to this and this was a real great video can you help me how can i set up the authorisation token where can i get the access token in Servicenow or how can that be fetched ?
What did you put on Send topic script action greetings chat gpt topic? what is the script?
there resolution of your video is poor, can you please copy the code your using on this post. That would help me a lot , thanks
HI Felipe, Thanks for the video. I'm stuck at the flow designer step where you defined the output variables. Can you tell us the exact steps you took to define the output variables? I got the output variables to populate but not in the order (answer first then action status) and I can't publish it that way and it gives me this error
"Please correct the following errors before save or publish:
Action Output: answer: Empty output value cannot be blank on publish."
I am not getting answer variable in the Output variables, please help me out
@@085raveendranathhosamani2 Were you able to fix this?
Hi , I am facing issue in flow designer. I am unable to find answer variable in the Output variables under Action Status.
When i test your flow i get an error. I follwed all your steps in the video and i cant get it tork. there must be something missing that wasnt shown on the video
When I test it, it says the bot runs into technical issues and on checking the logs, it says cannot convert null to an object. Type error
Hello, thank you for sharing this video. When in the askchatgpt http method, when i test i get this message in the response section: error": {
"message": "We could not parse the JSON body of your request. (HINT: This likely means you aren't using your HTTP library correctly. The OpenAI API expects a JSON payload, but what was sent was not valid JSON.)." I made sure content type is : application/json. Do you know how i fix this?
thank you!
Hi @gabrielebernini1440!
Go to the "askChatGPT" HTTP method, select the "HTTP Request" tab, bellow the "HTTP Query Parameters" you will find the "content" field.
Make sure you have a content like the following:
{
"model":"${model}",
"prompt":"${prompt}",
"temperature":${temperature},
"max_tokens":${max_tokens}
}
Then, in order to test, create variable substitutions like the following:
model --> text-davinci-003
max_tokens --> 1024
temperature --> 0.5
prompt -->
Let me know if this way you can get a valid answer!
Best Regards,
Filipe
@@CeRe4L unfortunately still not working. I get the following messages when testing :
HTTP status: 400
Error message : Method failed: (/v1/completions) with code: 400
Error Code: 1
i'm quite sure my http headers ( authorization and content type) are correct. Maybe is mething related to the script include? thank you a lot for your answer.
@@gabriberna902were you able to generate a successful API key?
@@gabriberna902 check the content body properly follow the (" ") , I had the same issue but then I rectified the content body.
Hi @majorlazer45
Make sure to pass "Content-Type" as "application/json" in the Http headers.
Ok, so I had the same publishing error message "Action Output: answer: Empty output value cannot be blank on publish," but eventually got this to work!
1) There are two places for the "answer" variable, one within the Inputs Script Step (I had to scroll down to see it) and one in the Outputs
2) When it gives you the error, you go back and set the default value of answer by dragging the "answer" pill from the Script Step.
3) I did have a typo! In the (Action Script Step) - the second part of this line, after the (). outputs.answer = new ChatGPT().askChatGPT(inputs.question); must exactly match the first part of this line, before the colon askChatGPT: function(message) of this line which in the (Script Include)
4) Sometimes a black arrow would show up to the left of the Action Status in Output Variables (as shown in this video). Once that appeared, it started giving me errors about Action Status "feature cannot be blank" etc. I finally decided to start from scratch, recreating the entire Action with a clean start. In my final working result, in Output Variables - I had "Action Status" on top, with no arrow on the left, and "answer" below that.
This fixed my issues, gave the correct answer when testing the Action, and allowed me to Publish! Hope this helps!
I'm getting the following response while testing the REST message,i gave model parameter as text-davinci-003 .?Can you help me with this
"error": {
"message": "you must provide a model parameter",
"type": "invalid_request_error",
"param": null,
"code": null
Go back and make sure you have the information for the REST message in the form correctly. seems you didn't include "text-davinci-003" as a variable for "${model}" parameter