Five systems.
One discipline.

The platform beneath Tektaris. A governed object model wrapped around the D.A.E. engine — it does not reimplement the engine, it governs, secures, multi-tenants, and serves it. Every conclusion traces to first principles; every decision is auditable end to end.

• MLDR — input boundary, rule-based parsers before any fallback
• Topos — append-only data layer, policy attached at ingestion
• Nomos — reasoning layer, structural confidence and Tri-Con refs
• Immutable hash-chained audit log, separately encrypted

Domain-configurable operational intelligence. Resolves which records describe the same real-world thing, fuses every source onto one object without silently reconciling conflicts, hands the fused entity to Stratum to reason, and lets operators act on an append-only audit. Tektaris never reasons — Stratum reasons.

• Deterministic entity resolution — ambiguity is a stable terminal state, never a guess
• Conflicting values are both retained, each tagged with its source
• Findings returned unaltered; confidence is never touched downstream
• Launch configuration: City v1 — Toronto municipal intelligence

The baseline engine. Drop one dataset in, get findings out — each with a confidence band, a traceback to the rules that produced it, and an audit trail. Embeddable as a library or called as a hosted multi-tenant service.

• Pure engine, no I/O — embeddable in-process or over HTTP
• Per-result confidence, traceback, and bias detection
• Determinism and golden tests run against every build

The next layer. Runs the engine across two linked datasets — partitions, compares, and correlates each independently, then surfaces which findings appear in both. Same audit and runHash guarantee as TAL.

• Two-dataset ingest with independent analysis per side
• Cross-dataset finding comparison, not a merged blob
• runHash guarantee — same input, same output, always

The full deterministic matrix. Heterogeneous sources — CSV, JSON, or pasted data — normalized into a single auditable run under one runHash. The enterprise tier of the family.

• Mixed-format ingest normalized to one canonical run
• One audit trail across every source in the matrix
• Same determinism contract as TAL and POP