How vishwajitvm/omnishell works
OmniShell turns natural-language requests into answers or guarded host workflows, with scheduled and recurring execution as a persistent branch. The README documents the five-stage control plane, agent swarm, human approval, host execution, verification, and scheduled-task/log storage. Source excerpts show the NestJS page controller, backend logging, and host executor behavior, but omit much of backend wiring; internal relationships beyond those excerpts are therefore based on README-described architecture.
Main entry points: main.py Agentic backend · app.controller.ts Web control center · local_executor.py Host execution agent · views Dashboard views
Components 16
Frontend
- Web control center
frontend/src/app.controller.ts - Dashboard views
frontend/views
Request control
- Agentic backend
backend/main.py - Intent and planning agents
backend/main.py - Safety and policy gates
backend/main.py - Research and knowledge
backend/main.py - Capability routing
backend/main.py - Human approval
backend/main.py - Knowledge response
backend/main.py
Execution and scheduling
- Host execution agent
backend/local_executor.py - Native OS commands
- Execution verification
backend/local_executor.py - Scheduled workflows
backend/main.py
State and services
- Execution logs
backend/main.py - Redis cache
Other components
- User
Connections 16
- User → Web control center: submits request
README.md - Web control center → Dashboard views: renders pages
frontend/src/app.controller.ts - Web control center → Agentic backend: sends request
README.md - Agentic backend → Intent and planning agents: analyzes intent
README.md - Intent and planning agents → Safety and policy gates: checks policy
README.md - Safety and policy gates → Capability routing: routes request
README.md - Intent and planning agents → Research and knowledge: supports research
README.md - Capability routing → Knowledge response: returns knowledge
README.md - Capability routing → Human approval: requires approval
README.md - Human approval → Host execution agent: authorizes execution
README.md - Host execution agent → Native OS commands: runs commands
README.md - Native OS commands → Execution verification: checks outcome
README.md - Execution verification → Web control center: returns result
README.md - Capability routing → Scheduled workflows: queues workflow
README.md - Agentic backend → Execution logs: writes execution log
backend/main.py - Agentic backend → Redis cache: reads and writes cache
backend/main.py