[FEAT]: Built-in Macro or Agent inside AnythingLLM to Automate repeated inference tasks

### What would you like to see? Can AnythingLLM automate the following form of repeated inference task? For example, using a built-in macro or agent? I have a very long text, arranged as small chunks (paragraphs). I want to translate it not as a whole, but as blocks of text, to ensure that the context does not interfere with the results of the translation. I want to first select the LLM, then feed the whole translation text (as JSON, XLSX, CSV, etc) into AnythingLLM, so that it processes one chunk (within the first row), get the result and store it in the thread and ALSO in an output CSV or JSON. Then AnythingLLM should automatically open a new thread with the same LLM but with a clean context, and move to processing the next row from the input JSON / CSV, with the same system prompt (translate the text). I guess this is possible with llama.cpp / transformer / torch etc. coding. But it would be great if AnythingLLM could have a built-in macro or agent that could do that with a few GUI clicks.