Adding a new context to the prompt when generating text

Hello, I’m trying to understand how I can use an LLM to add a specific context during generation. Generally I use several LLMs but at the moment I use mostly Mistral-7B-Instruct. Honestly, the question is complicated and I’m not quite sure how to phrase it.

I have a character created in text-generation-webui, and a Twitter account created on his behalf, and there are several variations of information that need to be generated. For example, when someone tweets a private message I need the LLM to understand that they are replying to a private message, and if someone writes a comment I need the LLM to understand that it is a reply to a comment. I would also like to be able to specify many other contexts and intentions. Say if a comment is not very friendly I would like the LLM to be more bold.

Is there any way to pass additional generation context? I know you can prescribe answers in the character editor, but what if a question comes up now that I haven’t prescribed before and I need to answer something specific. Let’s say a person in the comments under my post wrote “What food do you like?” and if I just directly ask the LLM to answer that question, he’ll answer some random food like “I like fast food” and I want the character to answer that he likes “steak” specifically, but the answer would use that character’s narrative style that I prescribed for him.

Can I write something like “In the comments to the question “What food do you like?” have the answer be “steak””, is there a similar syntax?