How thekaifansari01/jarvis-os-agent works
JARVIS OS Agent is a Windows-focused desktop and cross-device action agent. Its main workflow takes a voice, keyboard, or remote command through command handling and brain selection, executes tools or MCP actions, then returns a response while updating memory. Separate branches provide voice output and input, proactive outreach, retrieval/indexing, and desktop/mobile interfaces. The README documents several architectural links; sampled source confirms selected runtime wiring, but unsampled modules mean some internal calls are not shown.
Main entry points: CommandHandler.py Command Handler · main.py Application Startup · Memory.py Context Memory · AgenticBrain.py Agentic Brain
Components 23
Command Runtime
- Application Startup
main.py - Command Handler
core/main/CommandHandler.py - Command Processor
core/brain/Processor/Processor.py - Agentic Brain
core/brain/Processor/AgenticBrain.py - Fast Brain
core/brain/Processor/FastBrain.py - Action Executor
core/brain/executor.py
Memory and Retrieval
- Context Memory
core/brain/Memory/Memory.py - Lifetime Memory
core/brain/Memory/LifetimeMemory.py - Behavior Patterns
core/brain/Memory/BehavioralEngine.py - Hybrid RAG
core/brain/RagEngine.py - Knowledge Indexer
core/brain/Indexer
Voice and Interfaces
- Voice Input
core/voice/stt.py - Keyboard Input
core/main/HotKeyManager.py - Telegram Remote
core/main/TelegramRemoteBot.py - Speech Output
core/voice/tts.py - Desktop Interfaces
core/ui
Tools and Integrations
- Native Tools
tools - MCP Manager
core/mcp/MCPManager.py - Mobile Bridge
tools/MobileBridge/server.py
Proactive Services
- Proactive Agent
Proactive/proactive_agent.py - Event Queue
Proactive/event_queue.py - Proactive Channels
Proactive
Other components
- User
Connections 29
- User → Voice Input: speaks to
- User → Keyboard Input: types via
- User → Telegram Remote: commands via
- Application Startup → Command Handler: starts
main.py - Application Startup → Voice Input: starts
main.py - Voice Input → Command Handler: submits command
- Keyboard Input → Command Handler: dispatches command
core/main/HotKeyManager.py - Telegram Remote → Command Handler: routes command
core/main/TelegramRemoteBot.py - Command Handler → Command Processor: processes command
core/main/CommandHandler.py - Command Processor → Agentic Brain: selects complex tasks
- Command Processor → Fast Brain: selects simple tasks
- Command Handler → Action Executor: executes actions
core/main/CommandHandler.py - Command Handler → Context Memory: records exchange
core/main/CommandHandler.py - Agentic Brain → Context Memory: uses memory
- Fast Brain → Context Memory: uses memory
- Context Memory → Lifetime Memory: uses lifetime store
core/brain/Memory/Memory.py - Context Memory → Behavior Patterns: uses patterns
core/brain/Memory/Memory.py - Agentic Brain → Native Tools: invokes tools
- Fast Brain → Native Tools: invokes tools
- Agentic Brain → MCP Manager: uses MCP tools
- Application Startup → Hybrid RAG: starts service
- Hybrid RAG → Knowledge Indexer: loads index settings
core/brain/RagEngine.py - Application Startup → Proactive Agent: starts
main.py - Proactive Agent → Event Queue: queues events
Proactive/proactive_agent.py - Proactive Agent → Proactive Channels: dispatches outreach
Proactive/proactive_agent.py - Application Startup → MCP Manager: starts service
- Native Tools → Mobile Bridge: includes bridge
- Command Handler → Speech Output: speaks response
- Application Startup → Desktop Interfaces: starts interfaces
main.py