[Feature]: Simplify scenario creation
- Dominant language
- Python
- Stars
- 0
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
### Is your feature request related to a problem? Please describe.
Yes, the current ScenarioSpec-based architecture is overly complex and rigid. The JSON schema with actions, preconditions, effects, random events, and JSONLogic conditions creates a high barrier for both LLMs to generate valid specs and users to understand/modify scenarios. The Monte Carlo simulations add unnecessary computational overhead and complexity without clear benefits for creative storytelling. This makes the system difficult to use and limits creative freedom.
### Describe the solution you'd like
I'd like to completely redesign the scenario generation and engine architecture around a simpler, more narrative-driven approach:
Remove ScenarioSpec entirely and replace with:
World Background Generation: Generate rich, descriptive world lore and setting from the user's scenario description
Character Memory System: Create detailed backstories, personalities, and memories for each character/entity
Dynamic Narrative Engine: Use LLM-powered turn-by-turn storytelling without rigid rule systems
Key Changes:
Simplified Generation: Instead of generating complex JSON specs, generate:
World background text
Character profiles with memories and traits
Basic scenario metadata (name, themes, difficulty hints)
Remove Monte Carlo Simulations: No need for statistical validation - let creative scenarios exist without artificial difficulty constraints
Flexible Engine: The orchestrator would use the world background and character memories as context for LLM narrative generation, allowing more organic storytelling
Backward Compatibility: Keep the same API endpoints but change the internal data structures. This might have to also have frontend changes in the api response so I need to make sure that still works.
### Describe alternatives you've considered
_No response_
### Additional context
_No response_
### Would you like to contribute this feature?
- [ ] Yes, I would like to implement this feature
Contributor guide
Assessment
This issue has not been assessed yet.