TRAINING - GIVE AGENT

How do I train an agent?

Right now, you can not: You will have to create well planned characters. Take answers you do not like, reword them, edit the character, add the new Q and A. Repeat. Add better text PDFs to the memory section. This is prompting and context expansion, but NOT training.

Future: you will be able to pay to fine-tune a model, and train it on data. Small data sets will not cost a lot. You could do X-Y Mb for $0.yy-$y.yy or up to a gigabyte for less than $Z.zz

Even these small data sets can greatly improve a models performance, and if using small models, you can get both speed, and accuracy. It is what we believe will help people to create really fast models, that are specialized in specific topics. In the past, we have referred to this as "Domain-specific".