Skip to content

Home

Jump to bottom
Rafa Alvarez edited this page May 18, 2026 · 3 revisions

TekmerDB Wiki

The database that knows when it's wrong.

TekmerDB gives AI agents reliable memory — not just facts, but how confident to be in each one, where each came from, and when two sources disagree. It detects contradictions mechanically, tracks source reliability over time, and flags what it doesn't know. One air-gapped Rust binary. No cloud. No hallucinations presented as facts.

Wiki sections

Page What it covers Good for
Why TekmerDB Exists The problem AI is actually facing, why RAG fails in production, the design philosophy Business, investors, decision-makers
The PFO and Semantic Engine What a Probabilistic Fact Object is, how semantic similarity works, why embeddings beat keywords Architects, technical evaluators
Confidence and Trust How confidence is computed mechanically, why it's not an LLM opinion, source weight mechanics Everyone
Corroboration and Conflict What corroboration actually means, why contradictions are preserved, duplicate protection Compliance, regulated industries
Storage Architecture Hot/cold tier design, crash recovery, HNSW, why two storage layers Engineers, infrastructure teams
Domains and Compliance EU AI Act domain model, why the engine rejects certain facts Compliance officers, legal
Epistemic Design Why TekmerDB is a reliability engine not a truth machine, the core design principle Strategic, philosophical

At a glance

INSERT: "The drug was approved by the FDA"   source: medical_journal_a  confidence: 0.7
INSERT: "The drug was rejected by the FDA"   source: medical_journal_b  confidence: 0.7

ENGINE:
  PFO #1 → confidence: 0.52  CONFLICT FLAGGED  conflict_refs: [pfo#2]
  PFO #2 → confidence: 0.52  CONFLICT FLAGGED  conflict_refs: [pfo#1]

Both preserved. Neither resolved. No hallucination.

This is the entire philosophy of the project. In a terminal. In under a second.

Quick orientation

If you're evaluating TekmerDB for a product decision → start with Why TekmerDB Exists, then Confidence and Trust.

If you're a compliance officer → start with Domains and Compliance, then Corroboration and Conflict.

If you're integrating or building → start with The PFO and Semantic Engine, then Storage Architecture.

If you want to understand the design thinking → Epistemic Design is the deepest page.

Apache 2.0 · Rust · air-gapped · 198 integration test checks · 0 failures Enterprise features (audit exports, RBAC, EU AI Act compliance reporting) → commercial licence

Clone this wiki locally