#GPT-3.5 TURBO Completion (not Chat)

5 messages · Page 1 of 1 (latest)

brisk spire
#

I've just created an APP to get prompts from Davinci, but the output is not great. GPT4.0 blows it out of the water, but I'm still on the waiting list.

Is there a way to use prompts (like completion) for GPT3.5Turbo API, or does it only work as Chat?

I just don't understand how I can use the following chat format:
model: "gpt-3.5-turbo",
messages: [{role: "user", content: "Hello world"}],

to be more like this:
"model": "text-davinci-003",
"prompt": "Respond to this email." + emailContent,

I guess:
model: "gpt-3.5-turbo",
messages: [{ role: "user", content: "Respond to this email." + emailContent }],

rotund blaze
#

Include the system prompt that describes the overall behavior. For example, "You're an autoresponder for email that generates a response for the supplied email"

#

Also, I think I remember that the options are system, assistant, and user. And anything in the messages field is conversation history. Look up the documentation for chat system prompts and it should give you what you need.

# 
model: "gpt-3.5-turbo",
messages: [
  {"role": "system", "content": "You are an assistant that generates an email reply given an email."},
  {"role": "user", "content": "Respond to this email: " + emailContent },
]

// *GPT generates a response*
// *GPT stores the response in the conversation history / context*

// And now the next time you look at messages you will see the response from GPT-3
// which is how it stores conversation history / context
messages: [
  {"role": "system", "content": "You are an assistant that generates an email reply given an email."},
  {"role": "user", "content": "Respond to this email: " + emailContent },
  {"role": "assistant", "content": "{The content of the auto-generated email response}"},
]
compact laurel