leaden harness
Hey!
I've been reading the docs and was wondering about function calling. If functions are injected into the system message and count against the model's context limit (being billed as input tokens), does this mean that in the grand scheme of things we actually save tokens by using them? Or is it more like we're just spending a similar amount of tokens, just in a different manner? Would appreciate any insights on this.
Cheers! 🚀🤔
- Marcos