How pranav-sketch-cc/kayal1 works
Mark LVII is a desktop voice assistant with a graphical HUD. The main workflow runs from user input through the application runtime and model/tool handling to an action or spoken response. Its broader product includes configurable model and speech providers, persistent memory, background tasks and scheduled routines, plus desktop controls and web research. The source index and excerpts support the main runtime and several subsystem relationships; unsampled capabilities are represented as components where the tree supports them, without guessing their call paths.
Main entry points: main.py App runtime · engine.py Task scheduler · window.py Assistant window · brain.py Task brain
Components 23
Assistant runtime
- App runtime
Mark-LVII-main/main.py - Action loader
Mark-LVII-main/core/action_loader.py
Conversation and tools
- Task brain
Mark-LVII-main/core/tasks/brain.py - Gemini calls
Mark-LVII-main/core/gemini.py - Web search
Mark-LVII-main/actions/web_search.py - Tool outcomes
Mark-LVII-main/core/outcome.py
Tasks and routines
- Task scheduler
Mark-LVII-main/core/tasks/engine.py - Task store
Mark-LVII-main/core/tasks/store.py - Task research
Mark-LVII-main/core/tasks/research.py - Routine scheduler
Mark-LVII-main/core/routines/engine.py - Routine rules
Mark-LVII-main/core/routines/rules.py
Providers and state
- Speech pipeline
Mark-LVII-main/providers/engine_pipeline.py - Provider registry
Mark-LVII-main/providers/registry.py - OpenAI-compatible
Mark-LVII-main/providers/llm_openai.py - Configuration
Mark-LVII-main/memory/config_manager.py - Assistant memory
Mark-LVII-main/memory/memory_manager.py
Desktop interface
- UI entry
Mark-LVII-main/ui.py - Assistant window
Mark-LVII-main/gui/window.py - HUD overlays
Mark-LVII-main/gui/overlays.py - Dashboard server
Mark-LVII-main/dashboard/server.py
Other components
- User
- Model service
- Desktop and apps
Connections 27
- User → App runtime: initiates
Mark-LVII-main/main.py - App runtime → Action loader: loads actions
Mark-LVII-main/main.py - App runtime → UI entry: starts UI
Mark-LVII-main/main.py - UI entry → Assistant window: builds window
Mark-LVII-main/ui.py - UI entry → HUD overlays: uses overlays
Mark-LVII-main/ui.py - Assistant window → HUD overlays: opens panels
Mark-LVII-main/gui/window.py - Assistant window → Task scheduler: uses tasks
Mark-LVII-main/gui/window.py - Assistant window → Routine scheduler: uses routines
Mark-LVII-main/gui/window.py - Assistant window → Gemini calls: uses calls
Mark-LVII-main/gui/window.py - Task brain → Gemini calls: calls model
Mark-LVII-main/core/tasks/brain.py - Task brain → Action loader: loads tools
Mark-LVII-main/core/tasks/brain.py - Task brain → Provider registry: selects provider
Mark-LVII-main/core/tasks/brain.py - Task scheduler → Task brain: runs rounds
Mark-LVII-main/core/tasks/engine.py - Task scheduler → Task store: persists tasks
Mark-LVII-main/core/tasks/engine.py - Task scheduler → Task research: uses research
Mark-LVII-main/core/tasks/engine.py - Routine scheduler → Routine rules: checks triggers
Mark-LVII-main/core/routines/engine.py - Web search → Gemini calls: fallback search
Mark-LVII-main/actions/web_search.py - Web search → Tool outcomes: formats result
Mark-LVII-main/actions/web_search.py - Provider registry → OpenAI-compatible: builds provider
Mark-LVII-main/providers/registry.py - Speech pipeline → Provider registry: selects providers
Mark-LVII-main/providers/engine_pipeline.py - Speech pipeline → Model service: sends requests
- Gemini calls → Configuration: reads key
Mark-LVII-main/core/gemini.py - Task brain → Configuration: reads settings
Mark-LVII-main/core/tasks/brain.py - Task scheduler → Assistant memory: uses memory
Mark-LVII-main/core/tasks/engine.py - App runtime → Assistant memory: uses memory
Mark-LVII-main/main.py - App runtime → Dashboard server: starts service
Mark-LVII-main/main.py - Action loader → Desktop and apps: dispatches actions