ChatGPT is cutting off responses in it's outputs.

  • 25 November 2023
  • 1 reply
  • 125 views

I am not getting full response, it breaks after one line. below are my settings I am using chatgpt conversation:

 

Model

gpt-4-vision-preview

 

User Name

user

 

Assistant Name

text_analyse

 

Assistant Instructions

Please analyze the following meeting transcript and provide the analysis in a structured JSON format. The analysis should include the following elements: 1. Emotional Tone: A brief summary of the overall emotional tone of the meeting. 2. Lead Analysis: A boolean value indicating if there is a potential good lead in the transcript ('true' for a good lead, 'false' for a bad lead). 3. Lead Details: Regardless of the lead's quality, provide details in a JSON object format, including: - Name (if mentioned) - Expressed Need/Interest - Contact Information (if available) - Additional Notes 4. Lead Reason: Provide a brief reason why the lead is considered good or bad, with relevant details or observations. The expected JSON output format should be as follows: { "EmotionalTone": "Summary of the emotional tone.", "LeadAnalysis": true/false, "LeadDetails": { "Name": "Name if mentioned", "NeedInterest": "Expressed Need/Interest", "ContactInfo": "Contact Information if available", "AdditionalNotes": "Any additional notes" }, "LeadReason": "Explanation for why the lead is considered good or bad" }

 

Max Tokens

32000

 

Temperature

0.2

 

Top P

1.0


This post has been closed for comments. Please create a new post if you need help or have a question about this topic.

1 reply

Userlevel 7
Badge +11

Hi @ac12121, welcome to the Community! 👋

This sort of behaviour is usually due to the ChatGPT app having a 50 second processing limit when testing in the Zap Editor. When the time limit is reached the response will be cut off and a partial response will be returned. When the Zap is switched on and running live it has up to 15 minutes to be processed - this should ensure it has enough time to finish processing the request and return the full response to the subsequent actions in the Zap. 

Can you confirm whether the ChatGPT app is indeed only cutting off responses when testing in the Zap Editor? Or is it also cutting off the responses when the Zap is switched on and running live?

Thanks, looking forward to hearing from you on this!