Design Controls
iDetail Explorer
i⬡
Select a Node or Edge
Click any graph element to inspect it.
Click any graph element to inspect it.
v0.5.5 · April 2026
Set 9 parameters including Design Intent. A deterministic engine generates constraints, tensions, derived properties, and compound interactions from your choices. Intent actively shapes the graph — a compliance-oriented system produces different structural consequences than one optimized for speed. The Configuration Summary shows system profile classification, severity-ranked tensions, operational impacts, top risks, improvement directions, and the highest-impact variable change.
Add Humans, AI Assistants, Orchestrators, and Equipment AI. The system generates interactions shaped by your configuration state — edge conditions reflect governance and oversight (approval gates, periodic review, autonomous operation), coordination style affects handoff patterns, and monitoring coverage depends on observability level. At Layered resolution, orchestrators mediate between thinking and execution layers. Issues such as bottlenecks, coupling risks, orchestration gaps, and access friction are detected deterministically.
When resolution is set to Layered or Detailed, an orchestration layer appears between AI Assistants and Equipment AI. Orchestrators manage workflow scheduling, task sequencing, and state tracking. Their presence changes the system from direct AI-to-equipment control to mediated flow. Missing orchestrators in a full system is a diagnostic signal.
Use Scope (Full/Thinking/Equipment) to test system completeness by removing layers. Use Resolution (Abstract/Layered/Detailed) to control orchestration visibility. Diagnostic scenarios combine scope and resolution with specific parameters to produce known structural outcomes.
Filter the graph by concern: Governance, Data Flow, Collaboration, or Risk & Tension. The full graph is preserved underneath.
Local: deterministic analysis of graph state. Connected: uses your API key for broader questions. AI can propose setting changes but never applies them without confirmation.
A PI-facing interpretation layer derived from configuration and simulation. Shows system profile classification (e.g., High-Control System, Reactive System), a compositional narrative built from governance, oversight, intent, and coordination style, severity-ranked risk summary, improvement directions, scenario delta, activity feed with config-aware events, and lab/equipment/AI status blocks.
Every tension, constraint, and simulation issue is assigned a severity level (Critical, High, Moderate, Low). The system identifies the top risks, generates directional improvement guidance tied to specific tensions, and computes the single variable change that would most reduce structural friction. All outputs are deterministic and traceable to configuration choices.
Real-time design exploration for AI research infrastructure. Visualize dependencies, tensions, and operational implications — test system completeness and orchestration diagnostics — then see what matters most and what to change.
Not a single best design — every configuration has tradeoffs. But AIR does identify your top risks by severity, generate directional improvement guidance tied to specific tensions, and compute the single variable change that would most reduce structural friction. It helps you decide, not just analyze.
Structural conflicts needing resolution — not errors. They signal where design choices create opposing requirements. Each tension is assigned a severity level so you can focus on what matters most.
AIR can generate formatted reports from your configuration. The Basic Report is a shareable PDF of what the tool already shows you. The Expanded Report adds deeper interpretation and recommendations. See the Reports button in the top bar for options.
Pre-configured system states designed to produce known structural outcomes. Unlike presets, scenarios are validation tools — they define expected diagnostics and narrative insights for a given input. Use them to verify AIR's diagnostic behavior or as onboarding demonstrations.
The mediation layer between AI Assistants (thinking) and Equipment AI (execution). Includes workflow schedulers and agentic middleware. Visible at Layered and Detailed resolution. Its absence at Layered resolution produces a 'missing orchestration' diagnostic.
Yes. The tool runs entirely in your browser. Connected AI sends data only to your configured provider.
No. Core features — all parameters, graph visualization, reasoning engine, decision layer, and local AI analysis — work without one. An API key is only needed for Connected AI mode.
Watch how a configuration flows through AIR's five layers: Design → Logic → Behavior → Interpretation → Decision. Choose a scenario to see distinct system profiles and outputs.
Connected AI is optional. The tool runs locally without an API key.
Export regularly to preserve design history.
Optional. Use comma-separated labels. Leave blank to use smart defaults.
Resets entire tool — clears configuration, simulation, and saved session data. Cannot be undone.
Get a structured analysis of your AI research infrastructure design. Each report option builds on the deterministic logic AIR already generates.