It would be helpful if we could provide specific preferences individually when they come up. For instance, if we always want "epiretinal membrane" to be abbreviated "ERM" in the A/P section, we could tell it this preference for all future notes. Or we could make more complex "rules," such as always commenting on the higher risk of intraoperative complications in the A/P for a cataract evaluation whenever the cataract is described as a "mature white cataract" in the exam section. Often times, we notice these things when they come up, and it would be nice to be able to give it a specific instruction that sticks with our preferences ongoing. These preference could be organized in an ongoing list with entires that can be revised/deleted individually in the app preferences (similar to "Rules" in email apps).

When creating referrals the default is for the document to have lots of bullet points. This isn't a standard referral template most clinicians in our field use. Whilst we can create our own template stating to do it in prose or ask Heidi to do so, this is often an extra step. Currently there isn't an efficient way of using our own templates to create a referral for a specific problem, one consultation often has multiple problems so I have to either use Heidi's recommendation for a referral and "ask Heidi" to change to prose or choose my template and "ask Heidi" to focus on a specific problem only. Eliminating a step would be beneficial. I find I'm always asking Heidi to change to prose

I wanted to suggest a possible feature that I think would help with consistency. I suppose the core idea is to have a system of customisable, persistent rules, inspired by the Rules feature in Cursor.

Cursor allows coders to create a text file containing global or context-specific instructions for the AI. These provide persistent context for each generation, ensuring the AI e.g. adheres to specific coding standards, uses certain libraries, or maintains a particular style. It's a good way to automate consistency.

Currently, you have to include writing styles/rules in each template, and update individual templates if wanting to make wider reaching changes. Heidi could use a similar system allowing clinicians to define a set of persistent rules that guide the AI's output across notes and documents. This is flexible The UI could be an evolution of the current 'Replacements' feature, with a similarly simple interface.

Here’s how it could be structured with multiple levels of control:

Examples:

Use UK English spelling. Maintain a formal tone.

Refer to our local Paediatrics service as 'The Children's Centre'.

Always use the phrase 'Attend A&E if...'.

Category-Specific Rules: Separate instructions for different document types, recognising that notes and letters have different requirements.

There are certain words that I do NOT want Heidi to use EVERY - such as therapy and therapeutic. Would it be possible to have global defaults - so that this does not need to be added to every template.

Could we set up general custom instructions that apply to all templates, similar to the replace function under the memory tab? This way, any instructions I like in one template won't get missed when copying it over to others.

It would be good to be able to issue global instructions that are followed for all notes / documents, including those made with direct instructions (ie ad hoc notes / letters)