Seed is a snowflake type operation when it comes to writing. The idea is to start with a asset (character or setting) and then add descriptors. A descriptor can be shared between assets but are destinct from assets.
A seed grows much like a plant, and is meant to foster ideas in an organic fashion. This should be taylored for the experience, but for now, the seed's rules follow that of the Fibonacci sequence.
The idea is to take the assets and descriptors and use them as a seed that get's passed into an AI model. The AI can then be asked specific prompts about an asset and/or provide a summary for the asset based on the input.
The concept of Leveling Up is similar to an RPG. For the initial implementation, the Leveling Up algorithm makes use of the Fibonacci sequence. This is meant to facilitate organic growth.
TODO: Leveling Up should be abstracted so that other numerical sequences can be used.
This holds the global state of assets and descriptors. When the number of assets or descriptors reach a fibonacci number, the global asset or global descriptor will be set to a Level Up status. The next fibonacci number is calculated to determine how many assets/descriptors are needed for the next Level up.
An asset levels up when it's number of linked descriptors is equal to a Fibonacci number.
A descriptor levels up when it's number of descriptions is equal to a Fibonacci number. NOTE: Descriptions themselves must have a word length equal to a Fibonacci number.
len("blue".split())
len("blue hair".split())
len("blue haired woman".split())
len("blue haired woman in bar".split())
len("blue haired woman in bar drinking a beer".split())
Examples of invalid descriptions:
len("blue haired emo woman".split())
len("blue haired emo woman in bar".split())
len("blue haired emo woman in bar drinking an alcoholic cocktail".split())
These are the next planned features.
- TODO: Export to JSON
- TODO: Build a simple GTK GUI to capture Assets and Descriptors
- TODO: Setup pytorch and interaction with a model.
- TODO: Add the ability to pass in the lengths of the global objects as a part of the calculations
- TODO: Plugin to support descriptors as social media hashtags, and the description as any content associated with a post
- TODO: Automate how a model is added/downloaded into the project
- TODO: Add 'type' category for the cacluation function used to determine leveling up
- TODO: Abstract out calculations for Assets and descriptors allowing for much easier plugin/alternate level up function calculations.