GPT model version settings per AI feature

  • Its unclear what AI feature uses what GPT model.
  • There are already a number of AI featrures in Fibery, but no way to set the model they use. Not each AI featrure needs GPT-4. If I see my OpenAI API usage bill, the highest daily costs are GPT-4, and GPT-3 is just a fraction in comparison. Can these settings be added?

By the way, here is how I would use assign GPT models to Fibery AI features. Please let me know if this makes sense?

GPT-4 (deprecated) / GPT-4-Turbo (to be announced):

  1. AI Chatbot: Best for complex, in-depth conversations.
  2. AI Assistant: Ideal for detailed, context-rich responses.
  3. AI-Enhanced Search: Great for accurate, nuanced information retrieval.

GPT-3.5 / GPT-3.5-Turbo:

  1. Automations with AI: Good for summarizing texts and basic emotion analysis.
  2. Space Creation Automation: Adequate for straightforward database setup.
  3. Writing and Planning Tools: Useful for simple writing assistance and idea generation.
  4. Formula Generation: Efficient in converting written descriptions into formulas.

And… costs are related to the frequency of OpenAI calls.
I think that the following list is sorted according to most amount of calls?

  1. AI Chatbot: High-cost due to complex, ongoing interactions.
  2. AI Assistant: Moderate-cost for detailed entity queries.
  3. AI-Enhanced Search: Costs vary with the number of databases.
  4. Automations with AI: Cost depends on text complexity.
  5. Writing and Planning Tools: Variable cost based on task nature.
  6. Formula Generation via NLP: Generally lower cost for specific requests.
  7. Space Creation Automation: Lower cost, structured one-time process.