Very nicely written Krishna Chaitanya
May I share a thought on using the messages array with OpenAI's gpt-4 and gpt-3.5-turbo APIs. I use messages to help maintain conversational flow by recycling the chat history and adding an NLP analysis.
messages: [ { role: "system", content: topic },
{ role: "assistant", content: LastPrompt },
{ role: "assistant", content: LastResponse },
{ role: "user", content: Get("Keywords") },
{ role: "user", content: Get("Entities") },
{ role: "user", content: Get('Summary') },
{ role: "user", content: Get('GPT_Ideas') },
{ role: "user", content: prompt }, ]
Topic, LastPrompt, LastResponse, Keywords, Entities, Summary and new Ideas are ranges in the sheet and are used to prime the follow-up prompt. The NLP is generated by the Do_NLP_Analysis() executed prior to sending the prompt to the LLM.
function Do_NLP_Analysis() {
// generate an NLP analysis
Keywords = Do_Keywords() // uses fast-gpt-j
Summary = Do_Summary() // uses finetuned-gpt-neox-20b or Cohere Summarization
Entities = Do_Entities() // uses Google Cloud Natural Language
Ideas = Do_Ideas() // uses either gpt-4 or gpt-3.5-turbo
return; }
Get() is a simple library function that gets the value of a range in the sheet.
It works for me.