How fatihmakes/mark-lvii works
Mark LVII is a cross-platform voice assistant whose main workflow is a user request entering the conversation runtime, being handled through discovered tools or long-running tasks, and returning as speech or a HUD update. Separate branches provide persistent background research and configurable routines. The graph also preserves the alternate STT/LLM/TTS engine, provider settings, memory, and the command-center UI. Unsampled capabilities are included only where README or source-index evidence supports them; some internal wiring is therefore omitted.
Main entry points: main.py Assistant Runtime · window.py Assistant Window · engine.py Task Scheduler · brain.py Task Planning
Components 27
Conversation Runtime
- Assistant Runtime
main.py - Gemini Live
core/gemini.py
Tasks and Routines
- Task Commands
actions/background_tasks.py - Task Host
core/tasks/desktop.py - Task Scheduler
core/tasks/engine.py - Task Planning
core/tasks/brain.py - Deep Research
core/tasks/research.py - Task Checkpoints
core/tasks/store.py - Routine Commands
core/routines/desktop.py - Routine Scheduler
core/routines/engine.py - Routine Rules
core/routines/rules.py
Actions and Memory
- Action Discovery
core/action_loader.py - Bundled Actions
actions - Web Search
actions/web_search.py - Verified Outcomes
core/outcome.py - Recallable Memory
memory/memory_manager.py
Speech Engines
- Speech Pipeline
providers/engine_pipeline.py - Provider Registry
providers/registry.py - Compatible LLM
providers/llm_openai.py - Speech Providers
providers
HUD and Settings
- Assistant Settings
memory/config_manager.py - Engine Settings
gui/providers_panel.py - Assistant Window
gui/window.py - Command-Center HUD
gui/hud.py - HUD Panels
gui/overlays.py - Remote Dashboard
dashboard/server.py
Other components
- User
Connections 36
- User → Assistant Runtime: speaks to
- Assistant Runtime → Gemini Live: runs conversation
main.py - Assistant Runtime → Action Discovery: loads actions
main.py - Action Discovery → Bundled Actions: discovers skills
- Assistant Runtime → Task Commands: dispatches commands
main.py - Task Commands → Task Host: requests task
actions/background_tasks.py - Task Host → Task Scheduler: schedules work
core/tasks/desktop.py - Task Scheduler → Task Planning: plans rounds
core/tasks/engine.py - Task Scheduler → Deep Research: runs research
core/tasks/engine.py - Task Scheduler → Task Checkpoints: checkpoints state
core/tasks/engine.py - Deep Research → Task Planning: requests planning
core/tasks/research.py - Deep Research → Task Checkpoints: persists progress
core/tasks/research.py - Task Planning → Action Discovery: gets tool definitions
core/tasks/brain.py - Task Planning → Gemini Live: uses model calls
core/tasks/brain.py - Task Commands → Verified Outcomes: formats status
actions/background_tasks.py - Assistant Runtime → Routine Commands: sets up routines
main.py - Routine Commands → Routine Scheduler: manages schedules
core/routines/desktop.py - Routine Scheduler → Routine Rules: evaluates triggers
core/routines/engine.py - Routine Scheduler → Task Checkpoints: starts tasks
core/routines/engine.py - Assistant Runtime → Recallable Memory: uses memory
main.py - Web Search → Gemini Live: uses search fallback
actions/web_search.py - Assistant Runtime → Assistant Settings: reads settings
main.py - Assistant Window → Assistant Runtime: starts assistant
gui/window.py - Assistant Window → Command-Center HUD: presents HUD
gui/window.py - Assistant Window → HUD Panels: opens panels
gui/window.py - Assistant Window → Engine Settings: opens engine settings
gui/window.py - Assistant Window → Routine Commands: configures routines
gui/window.py - Assistant Window → Task Host: shows task controls
gui/window.py - HUD Panels → Recallable Memory: shows stored facts
gui/overlays.py - HUD Panels → Remote Dashboard: connects remote panel
gui/overlays.py - Engine Settings → Provider Registry: lists providers
gui/providers_panel.py - Assistant Runtime → Speech Pipeline: selects alternate engine
main.py - Speech Pipeline → Provider Registry: creates providers
providers/engine_pipeline.py - Provider Registry → Compatible LLM: loads compatible model
providers/registry.py - Speech Pipeline → Speech Providers: uses speech components
providers/engine_pipeline.py - Engine Settings → Assistant Settings: saves engine choices
gui/providers_panel.py