🕸️ SemanticEmbed — AI Agent Topology Risk Analyzer
Pick a mode → click Analyze. Get a 6D structural encoding plus risk findings — single points of failure, amplification cascades, convergence sinks — from topology alone.
Each mode loads a starter example into the code box. Edit it, paste your own file over it, or just hit Analyze on the example to see what the output looks like.
Designed for AI agent pipelines where vendor concentration, gateway bottlenecks, and guardrail SPOFs hide in the orchestration graph.
PyPI · GitHub · Demo dashboard · Validation methodology
Encoding runs server-side. The Space sends only the edge list — your file content stays on this machine.
Topology graph
Node size and color encode criticality (bigger and redder = more structural risk). Risk-flagged nodes get a colored ring. Hover for the full 6D vector. Use the Inspect node picker below for the full breakdown of any node.
Pick a node to see its full 6D vector and any risks. Sorted by criticality.
Run Analyze first to load a topology, then pick a node to inspect.
6D structural encoding (top 20 nodes by criticality)
Structural risks
Drift comparison
Paste your before code on the left and the after version on the right (same mode for both). The union graph shows nodes added (teal +), removed (gray ×), and Δ criticality for nodes in both. Useful for architecture review: what did this refactor actually change about structural risk?
Pick a node to see its 6D vector on each side and the Δ. Sorted by |Δ criticality|, biggest swings first.
Run Analyze drift first, then pick a node to compare its 6D vector before vs after.
Per-node delta (top 20 by |Δ criticality|)
Built by Jeff Murray · GitHub @jmurray10 · Patent pending · Application #63/994,075