SPINE → Interactive Demo

OODA Loop Explorer

Interactive visualization of SPINE's 5-phase agent decision cycle

The OODA Cycle

Phase 1 Observe
Phase 2 Orient
Phase 3 Decide
Phase 4 Act
Phase 5 Reflect
Click a phase in the diagram to explore it

Phase ↔ Component Mapping

Phase SPINE Components
Observe ContextDiscoveryEngine, EphemeralMemory, WorkTracker
Orient MemoryFacade, PatternStore, GapTracker
Decide CapabilityRegistry, TaskTypeRouter, TaskDecomposer
Act SubagentExecutor + SafeAccessGate + ToolEnvelope
Reflect StructuredCritique, FeedbackLoop, VerdictRouter, OscillationTracker

Cycle Simulator

0
Cycles
--
Phase Time
0ms
Total Time
Ready
Status
Observe
Waiting...
Orient
Waiting...
Decide
Waiting...
Act
Waiting...
Reflect
Waiting...

Architecture

Key Classes

OODALoop Main orchestrator
OODAConfig Phase configuration
LoopContext Shared state across phases
WorldState Environment snapshot
Outcome Cycle result

System Capabilities

6
Embedding Providers
7
Memory Tiers
7
Executor Types
5
OODA Phases

How the OODA Loop Works

OODA (Observe-Orient-Decide-Act) is a decision cycle originally developed by military strategist John Boyd. SPINE extends it with a fifth phase — Reflect — to enable continuous learning.

loop = OODALoop(config)
outcome = await loop.run(goal, world_state)
# Cycles through all 5 phases, returns Outcome

Each phase is handled by dedicated SPINE components. The loop runs continuously until the goal is achieved, a maximum cycle count is reached, or the Reflect phase determines no further progress is possible.