Inference Runner

Execute Python inference scripts and watch the 8-step reasoning process in real time

Select Agent

Inference Steps

Step 1: Canonicalise Atoms

Load JSON, dedupe, normalize families

Step 2: Infer Elements

Partition atoms into evidence bundles by family

Step 3: Load Primitives

Define agent design primitives with required families

Step 4: Infer Constructs

Map primitives to frameworks/gates with validation rules

Step 5: Infer Compounds

Create executable decision units from constructs

Step 6: Compute Key Atoms

Select load-bearing atoms per construct

Step 7: Validate Operator Set

Check family coverage, orphans, collisions, compositionality

Step 8: Emit Outputs

Write JSON + Markdown inventory reports

Event Log