How muhammadbinsalman191/mark-li-advanced works
Mark-LI Advanced is a local, voice-enabled JARVIS-style desktop assistant. Its main workflow is user interaction through the desktop or microphone, live-session reasoning, tool dispatch to assistant capabilities, and spoken or on-screen results. The README documents plugins, persistent memory, configuration, and a local HTTP dashboard as additional capabilities. The excerpts show session orchestration, memory and action behavior, but not all startup wiring; links for unsampled subsystems are therefore limited to documented relationships.
Main entry points: main.py Live Session · llm_client.py LLM Client · ui.py Desktop UI · stt.py Speech Input
Components 21
User Experience
- Desktop UI
ui.py
Assistant Session
- Live Session
main.py - Speech Input
core/stt.py - Speech Output
core/tts.py - LLM Client
core/llm_client.py
Assistant Capabilities
- Assistant Actions
actions - Web Search
actions/web_search.py - File Control
actions/file_controller.py - Vision and Screen
actions/screen_processor.py - Computer Control
actions/computer_control.py
Local State
- Assistant Memory
memory/memory_manager.py - Runtime Config
memory/config_manager.py - Config Files
config
Extensions and Remote
- Plugin Loader
core/plugin_loader.py - Plugins
plugins - HTTP Dashboard
dashboard/server.py
Other components
- User
- Gemini Service
- Local LLM Service
- User Files
- Remote Device User
Connections 22
- User → Desktop UI: interacts
- Desktop UI → Live Session: session input
- Live Session → Desktop UI: updates UI
- Live Session → Speech Input: captures speech
- Speech Input → Live Session: transcribes
- Live Session → LLM Client: requests response
- LLM Client → Live Session: returns response
- Live Session → Speech Output: speaks response
- Live Session → Assistant Actions: dispatches tools
- Live Session → Assistant Memory: loads and updates
- Live Session → Runtime Config: reads settings
- Runtime Config → Config Files: reads and writes
- Live Session → Plugin Loader: discovers plugins
- Plugin Loader → Plugins: loads skills
- Live Session → Web Search: dispatches search
- Live Session → File Control: dispatches file tools
- Live Session → Vision and Screen: dispatches vision tools
- Live Session → Computer Control: dispatches controls
- File Control → User Files: manages files
- Remote Device User → HTTP Dashboard: connects
- LLM Client → Gemini Service: uses API
- LLM Client → Local LLM Service: uses backend