Roadmap

TDA n-gram point cloud, Betti numbers β₀/β₁ via persistent homology. <2ms.

Depth-3 recursive: Base64, hex, ROT13, Caesar, UUencode, homoglyphs.

15 regex patterns + Shannon entropy scan for unknown secrets.

420 rules, compound risk escalation: 3× MEDIUM → HIGH.

MiniLM ONNX + Poincaré ball: 70% cosine + 30% hyperbolic blend.

5-node Bayesian DAG, Pearl do-calculus, backdoor correction. <3ms.

URL phishing detection + semantic social-engineering classifier.

Redis sliding window ERS. Score ≥0.75 → Shadow Ban (gaslight/delay/standard).

Claude Opus auto-generates examples from BLOCK events. Hot-reload without restart.

8 jurisdictions, MASQUE_H3 tunnels, TOFU TLS pinning, cross-border enforcement.

HybridSigner (Ed25519+ML-DSA-65), HybridKEM (X25519+ML-KEM-768). FIPS 203/204.

18-provider fingerprint DB. Async /24 subnet probe, DNS telemetry classifier.

9-stage DAG, causal attribution, counterfactuals, HTML + PDF reports.

4 sub-agents (SOVAOperator, ThreatHunter, ForensicsAgent, ComplianceAgent). HMAC tokens.

Claude Opus 4.6 autonomous operator. 30 tools, 7 ARQ cron jobs, Redis memory.

UECIID business identity, STIX 2.1 audit chain, Knock-and-Verify invitations.

Multi-cloud vault connectors. SQLite inventory, risk scoring, rotation scheduling.

Scan, share, XAI pipeline visualizer, Dataview dashboard, offline publish queue.

Per-layer spans across all 9 filter stages. Jaeger UI at :16686.

Centralized metric contracts, dimension mappings, access rules. LLM → deterministic SQL. OSI 1.0 export/import.

Unified settings: Agents, Notifications, Agentic Commerce, Semantic Layer. Portal + Streamlit + SOC.

9 built-in semantic models (incidents, vendors, commerce, sovereignty, compliance, AI spend). Redis cache 10 min TTL. SOVA tools: semantic_query + list_semantic_models.

Semantic Layer–backed budget control. Every AP2 payment pre-checked against actual MTD spend (ai_spend model). Per-transaction + monthly limits + approval gate + Slack alert.

Tenants register custom semantic models via /models/catalog. Persisted to SQLite, hot-loaded. CRUD: POST/PUT/DELETE /models/catalog/{id}.

Pre-merge security gate: scans commit message + per-file diff via 9-layer pipeline. Composite action, PR comment risk table, 90-day audit artifact, pre-commit hook mode.

Real-time posture across SOC 2 / GDPR / ISO 27001 / HIPAA / NIS2. 168-snapshot ring buffer, 30s auto-refresh, SVG score gauge, Streamlit + SOC dashboard pages.

All 93 Annex A controls mapped across 4 themes (Org/People/Physical/Tech) with evidence strings. Per-theme coverage %, HTML report, Streamlit + SOC dashboard drilldown.

DID registration + protocol schema discovery (GET /protocol/schema/{action}), pgvector semantic search (Layer 3), Brand Agent 4-gate filter (deny-list/TrustRank/rate-limit/capability), ClearingEngine dual-write SQLite+PostgreSQL.

AgentHandoffMemory (Layer 2) ~61% token savings for multi-agent handoffs. pgvector Layer 3 semantic search with SQLite fallback. SOVA tools #70-74: write_handoff_memory, read_handoff_memory, semantic_listing_search, get_protocol_schema, send_order_proposal.

Buyers must evaluate minimum 3 alternatives before purchase. Best offer ranked by price x (1 - rep_score) utility. Confused Deputy protection: analytics SQL scoped to caller DID.

Threshold triggers from Semantic Layer metrics. Slack/PagerDuty integration.

Drag-and-drop chart builder from QueryObjects. Embed in Portal.

Cross-tenant metric aggregation with per-tenant access isolation.