Architecture

Nanokit is organised into four horizontal layers. Each page below covers one of them in depth.

┌──────────────────────────────────────────────┐
│ CLI layer (nkapp)                            │  parses commands, resolves
│                                              │  project context, orchestrates
├──────────────────────────────────────────────┤
│ Core engine                                  │  config parsing, deep merge,
│                                              │  secrets, state, reconcile
├──────────────────────────────────────────────┤
│ Runtime layer                                │  Docker, Caddy gateway, bridge
├──────────────────────────────────────────────┤
│ Infrastructure layer (providers)             │  Docker/VPS, AWS, Azure, GCP, DO
└──────────────────────────────────────────────┘

Pages

Page	What it covers
Core Engine	Config parsing, deep merge, secret resolution, state management
Resource Reconciler	The observe → diff → plan → execute → report loop and drift hashing
Infrastructure Providers	The Provider Abstraction Layer and the supported provider set
Lifecycle & Hooks	How hooks fit into the orchestration pipeline
State & Locking	Local state files, infra locks, and how to clear a stale one
UI Bridge (WebSockets)	The telemetry/control bridge behind the Hub
Term & Hub Separation	Why the web terminal and the Hub are separate services

Related

• The Declarative Contract — the reconcile loop from an agent's point of view.
• nanokit.yml reference — the input to all of the above.
• CLI reference — the commands that drive it.