The MVP is the foundation, not a reduced final product. These capabilities must remain on the roadmap.
- FastAPI backend
- SQLite persistence
- local vector-store abstraction
- L1/L2/L3 memory boundary
- Python emotion engine
- developer runtime APIs
- plugin manifest scanning
- Flutter macOS UI
- chat screen
- developer dashboard
- local backend launcher
- Application Support data management
- OpenAI adapter
- Claude adapter
- local model adapter
- prompt/context builder
- streaming output
- cost/token accounting
Current code already contains ModelGateway, ModelProvider, and a MockModelProvider.
- Chroma or LanceDB adapter
- automatic REM cycles
- memory conflict detection
- memory merge and decay
- backup and restore
- manual memory editor
- Rust PAD engine
- Rust energy engine
- PyO3 or sidecar process bridge
- Python fallback remains available
- memory_plugin
- emotion_plugin
- model_plugin
- perception_plugin
- action_plugin
- ui_plugin
- plugin sandbox
- plugin permissions
- file perception
- web perception
- image input
- voice input/output
- tool execution
- task automation
- virtual avatar
- speech synthesis
- motion and gesture system
- more free-form interaction beyond chat
- multiple AI entities
- shared memory graph
- relationship/bond metrics
- Synaptic Lounge
- token economy experiments
- strategy A/B testing
- log-driven improvement
- offline fine-tuning pipeline
- supervised online adaptation
- rollback, audit, and governance