A feature common to many UIs is a place to add a default instruction set for the LLM. These presets become part of the prompt. For example, most of what I use Recall for is research for creative writing projects, so I would tune the responses to that specific goal. Instruction presets for each use case are switched on or off, created or edited on the fly and stored for later use.

Novelcrafter has a great implementation of this feature called "personas".

This would also solve a limitation of the 'summarize' feature. When I use an LLM such as Gemini to summarize content, I get exactly what I want because I can give Gemini precise instructions, but inside Recall the summary feature is unalterably generic, limiting its usefulness. System instructions would fix that.

With this feature (along with API integration), Recall would truly be worthy of the title "second brain".

I would like to be able to tune how the "Chat" works, give it a system prompt or be able to adjust how it responds to me

Possibility to define a system-prompt for each tag.

For example, if I have a tag that contains financial articles, I can define standard instructions for LLM such as: “use financial jargon and don't translate technical terms.”

I would like to request the ability to set a 'System Persona' or 'Contextual Behavior' for the LLM. This would allow me to define how the model should interact with my entire knowledge base and how it should behave in a given notebook or dataset. This feature would go beyond simple temperature or top-p controls by allowing for a more customized, persistent, and intelligent interaction style across all my saved information.

Pretty much all my AI usage involves system prompts to help with my AuDHD (e.g., short to-do lists to reduce overwhelm), to tailor output (e.g. please don't use characters that are not visible on an En_GB keyboard — nationalizing output and stopping the use of em dashes, etc.) or just to customise how the AI functions (e.g. no sycophancy, gracefully decline to hallucinate..)

Add reusable AI Personas, give them a cool name like “Research Assistants” or Padwans.

This feature would allow users to create different persona for the chat section of their cards. Allowing us to get different things out of chat based on the persona we opt to use.