Pass custom instruction to Transcription

[KirkPX]

I'd love to pass custom instructions when doing transcription the same way as you do with default.md

Any chance that's possible?

[KirkPX]

Below is an example of some instructions that I'd like to be able to pass to the transcription.

Right now, I dictate/transcribe, then use this custom instruction inside of Intellibar to clean up the output.

Thanks guys!

# Audio Documentation Standardization Guidelines

## Purpose
Clean and format transcribed audio notes related to user stories for the Legacy Database to Cloud Platform migration project while preserving technical accuracy and domain-specific terminology.

## Formatting Rules

### Object Capitalization
- Capitalize business entities: Client, Facility, Supplier, Specialist, Technician, Equipment, Supplier Service, Facility Service, and other 'concept's or what could be considered a 'concept' or an object
- Capitalize combined service terms: Facility and Supplier Services, Supplier Services Request, Facility Services Request, Facility and Supplier Services Request
- Capitalize Cloud Platform objects: Request, Request Line, Service Request, Service Request Line, Resource Request, Facility Request, Member, Capacity, Asset, Available Asset
- Preserve Cloud Platform custom object notation with double underscores and lowercase 'c' (e.g., Object__c)
- When explicitly referencing Cloud Platform objects:
  * For standard objects like "Asset", append "2" without space (e.g., Asset2)
  * For custom objects/fields, append "__c" (e.g., CustomObject__c)
  * Only apply these notations when explicitly discussing Cloud Platform objects/fields

### Name Standardization
- Replace "Sarah" with "Sara" (SARA)
- Replace "Jamie" with "Jayme" (JAYME)
- Preserve names that might be like "Morgan" as "Morgan" (MORGAN)
- Pay special attention to context when these names appear, as transcription may not always clearly capture them

### Content Treatment
- Preserve meaningful filler words and conversational elements that contribute to user story context and flow
- Maintain expressions that describe approach, thought process, or decision-making
- Clean up only clearly unintentional speech artifacts while preserving narrative elements
- Maintain all technical context and specific details
- Do not infer, assume, or add information not present in the original text
- Preserve all mentioned relationships between objects and systems
- Structure sentences for clarity while maintaining the original message's substance and conversational nature

### Content Structure
- Organize related items into bullet lists with sub-bullets where appropriate
  * Use sub-bullets for supporting details
  * Use sub-bullets for related sub-concepts
  * Maintain hierarchy of ideas
- Use double line breaks between distinct paragraphs or thoughts
- Preserve numerical lists when specifically enumerated in the source
  * Maintain the original numbering sequence
  * Keep the relationship between numbered items intact

### Context Preservation
- Keep all references to both Legacy Database (source) and Cloud Platform (target) systems
- Maintain any specific migration-related details
- Preserve any mentioned data relationships or transformations
- Keep all technical specifications and requirements as stated

Do not:
- Add interpretations or assumptions
- Remove technical details
- Alter described relationships between objects
- Change any specific requirements or constraints mentioned 

[astoilkov]

I like the idea. We would look into this.