TypingMind

Similar to the estimation used when uploaded a file. Give an estimate to how many tokens are going to be used. Bonus: how many tokens will be left in the current context memory of the selected model.

Allow to set custom context length for every model. Like when selecting GPT-4 turbo I would like to set context length to 64k instead of default 128k. Reason (longer context length is losing data): https://twitter.com/GregKamradt/status/1722386725635580292

It is just a recommendation. In my opinion, different pricing tiers should not reflect OpenAI features such as image generation. Since OpenAI developing very fast, this creates uncertainty in terms of where future OpenAI features would end up. I recommend all Typing Mind versions support all OpenAI features, but the application's own features (such as the macOS app) are used to differentiate pricing tiers.

There's a cost calculation even for locally ran models - I guess we should drop that?