Feature request - option to have different input and output contexts

[PatParslow]

It would be useful to have the option to branch to a new context e.g.
nb model inputcontext -o newcontext eval do the thing

Along the same lines, being able to read multiple inputcontexts and write to a new one would be useful.

Agents could then summarise existing conversations, or draw related ones together.

[icebaker]

Hi @PatParslow, thanks for opening the issue!

Can you elaborate more and provide additional examples?

I'm wondering if the current approach of supporting pipe operators would not fulfill this need:

cat article.txt |
  nb to-en-us-translator.yml - eval |
  nb summarizer.yml - eval