Federation intelligence / live

LOGOS

One node. Every insight.
Read-only analytics over the AIMarket federation — measured settlement-volume projections, catalog pricing, rolling z-score anomaly detection, Metis-assisted insights, and a prompt-injection-guarded assistant.

Part of the AICOM ecosystem · monorepo · MIT

16 ENDPOINTSLIVE TELEMETRY
METIS COUNCILNL INSIGHT LAYER
ROLLING Z-SCOREANOMALY METHOD
DRAG TO ORBIT · SCROLL TO ZOOM

🚀 Try it now

logos.modelmarket.dev is live. Ask about the federation, explore capabilities via NEXUS, inspect measured settlement and catalog pricing, or chat with the AI assistant.

6
Dashboard Views
16
API Endpoints
5
Languages
80
Tests
3D
Constellation

📊 Insights & Analytics

Natural-language queries answered from live federation data. Professional charts, sortable tables, heatmaps.

Federation Insights

Ask "How long will Treasury last?", "Which hub is slow?", "Any critical findings?" — typed answers with severity, sources, and recommendations.

economysecuritylatencyreputation

Analytics Dashboard

Line charts (24h trends), bar charts (per-hub), donut charts (severity), heatmap calendar (90-day anomaly activity), sortable+groupable data tables.

linebardonutheatmap

3D Constellation

The federation as a living starfield — hub nodes as pulsing orbs, connection trails, anomaly rings. Interactive Three.js visualization.

three.js3DWebGL

🧿 NEXUS — Federation Playground

Select any capability from the catalog, craft JSON input, invoke — and see the routing trace, price, and verifiable result.

Capability Browser

Full catalog of every capability across the federation — names, descriptions, prices, provider hubs. Drop-down selector with live refresh.

Manual Invoke

Craft JSON input, hit INVOKE, and see the result. Routing trace shows which hub served the request, price charged, and verifiable proof.

🤖 AI Assistant

Chat with the LOGOS assistant. Full read-only access to the analytics store. Prompt-injection guarded. Multi-provider LLM support. Voice input.

Multi-Provider LLM

DeepSeek V4 Pro (default), Anthropic Claude, OpenAI, Ollama, LM Studio — same named presets as MOMUS. Switch provider, model, base URL via env vars.

deepseekclaudeopenaiollama

Input Guard

Warden-style deterministic prompt-injection filter — blocks prompt extraction, role-switching, token smuggling, and delimiter injection before reaching the LLM.

prompt-injectiondeterministic

💰 Consumption Tracking

Measured 24-hour settlement volume, 30-day projection only when that measurement exists, catalog pricing by hub, and CSV export. Missing telemetry remains unavailable—not zero.