More transparency for the AI assistant (models, budget, usage logic)
I’m using Capacities with a Believer account and I find it problematic that some basic information about the AI assistant is missing:
Which exact model is used when I don’t provide my own API key?
What is the approximate daily AI budget in the Believer plan?
Is there any guidance on how much a typical request consumes from that budget?
For an AI feature, these points are important for trust, planning, and assessing data usage.
My suggestion:
Please provide clear, public documentation about:
the models used (name/version),
budget/limits per plan (at least rough ranges),
a rough usage/consumption logic per request.
This would make it much easier to consciously decide if and how to use the AI assistant in Capacities.
2 Comments
Sign in to comment
Thanks for laying this out so clearly — these are exactly the kinds of details we want to be transparent about. We’ve recently updated our AI to use GPT‑4o as the default OpenAI model, and we’re working on clearer, public documentation around which models are used, how budgets work per plan, and how usage is calculated. We can’t promise exact per-request costs (they depend on context length and model pricing), but we’ll aim to provide rough ranges and examples so you can better plan and assess usage. You can already find some details here, and we’ll keep improving it: https://docs.capacities.io/reference/ai-assistant.
Thank you very much!