How dev-pradhuman/aegis-jarvis works
JARVIS Galaxy is a local assistant console backed by a durable service. Its main user-to-result path is chat or a workspace action into the service, where requests can be routed to models or canonical tools and tracked as runs; approvals can pause execution for user action. The README also describes browser automation, workflows, reminders, capability discovery, and system telemetry. Some backend wiring is inferred only from module names and documentation because implementation excerpts are partial; uncertain links are omitted.
Main entry points: router.js API routes · orchestrator.js Request orchestration · toolRuntime.js Tool runtime · workflowEngine.js Workflow execution
Components 24
Galaxy interface
- Galaxy home
src/components/GalaxyHome.jsx - Workspace windows
src/galaxy/WindowManager.jsx - Chat interface
src/components/Chat.jsx - Workspace views
src/components/WorkspaceViews.jsx
Service boundary
- HTTP service
server/index.js - API routes
server/router.js - Request orchestration
server/orchestrator.js - Run lifecycle
server/runEngine.js - Workflow execution
server/workflowEngine.js - Approval handling
server/approvalContinuation.js - System metrics
server/systemMetrics.js
Assistant intelligence
- Brain sources
server/brainService.js - Model routing
server/modelRouterService.js - Model providers
server/providerClient.js
Execution capabilities
- Tool runtime
server/toolRuntime.js - Browser automation
server/browser/BrowserManager.js - Desktop capabilities
server/platform/index.js - Reminder scheduler
server/scheduler.js - Research workflow
server/researchWorkflow.js - Media generation
server/mediaGeneration.js
Durable state
- Persistent state
server/store.js
Other components
- User
- Model services
- Host desktop
Connections 28
- User → Galaxy home: uses
- Galaxy home → Workspace windows: opens workspaces
- Workspace windows → Chat interface: shows chat
- Workspace windows → Workspace views: shows modules
- Chat interface → HTTP service: sends chat
- Workspace views → HTTP service: requests workspace data
- HTTP service → API routes: dispatches requests
- API routes → Request orchestration: routes commands
- Request orchestration → Brain sources: gets intelligence
- Brain sources → Model routing: selects models
- Model routing → Model providers: dispatches inference
- Model providers → Model services: requests inference
- Request orchestration → Run lifecycle: tracks execution
- Request orchestration → Tool runtime: invokes tools
- Tool runtime → Browser automation: dispatches browser actions
- Tool runtime → Desktop capabilities: dispatches desktop actions
- Desktop capabilities → Host desktop: controls desktop
- Workflow execution → Tool runtime: invokes canonical tools
- Workflow execution → Run lifecycle: synchronizes run state
- Approval handling → Workflow execution: resumes approved work
- Reminder scheduler → Tool runtime: attempts notifications
- API routes → Research workflow: serves research
- API routes → Media generation: serves media jobs
- API routes → System metrics: serves metrics
- Request orchestration → Persistent state: persists state
- Workflow execution → Persistent state: updates workflow state
- Reminder scheduler → Persistent state: claims reminders
- API routes → Persistent state: reads and writes state