OntoGuard

Evidence and benchmark scope

Release-reported results, controlled proof and independent integration reviews are not interchangeable.

Current reference releases

ProductReferenceEvidence statusLimit
OntoGuard Semantic Intelligence Runtime OEMv1.89.332Exact v1.89.332 qualification records inspected; owner publication approval pendingScoped three-source release campaign; residual sampled errors; no universal semantic perfection
OntoGuard Headless Decision Authorization Runtimev1.5.2Release reference in requirements; public PRs separately availableExternal production log, Python 3.12 qualification and signed OCI publication not established by supplied record

The exact-release reports, corpus identifiers, host configuration, baseline, denominators and approval must accompany any numerical runtime performance or semantic accuracy claims. Figures below are release-reported measurements; no customer-host estimate or unconditional guarantee is asserted.

Public historical integration results

PR #201 · TRACE adapter

Independently reviewed bounded integration

Merged September 20, 2026. Independent maintainer reproduction: 27 adapter tests. Authorized $250,000 path committed once with TRACE; altered $260,000 path refused, pending, zero commits and no TRACE.

Read PR #201 and review

PR #236 · Pre-commit enforcement

Independently reviewed bounded integration

Merged October 6, 2026; October 3 review by Carlos (carloshvp): 36 integration tests, 97 negative wrapper/direct-executor probes refused with zero commits and no protected effect, four positive probes executed the validated exact action. Imran Siddique approved and merged.

Read PR #236 and review

Bounded adapter and software harness, not independent validation of the complete runtime, SI, customer deployment or universal non-bypassability. Source: linked PRs; launch recheck required.

The 27 and 36 tests, 97 negative and four positive probes refer to separate bounded integration work, not current full release test counts.

Evaluation methodology

SI: measure constituted population, recall and sampled correctness with exact denominators, evidence coverage, export parity, elapsed time and hardware. DA: measure exact action refusal/execution, false block/escalation, protected effect observations, latency distribution, throughput and trust configuration. Compare like workloads and define boundaries.

Semantic Intelligence — exact release-reported evidence

v1.89.332: the signed package reports release approval and commercial production readiness under its tiered policy. The selected campaign reports 1,244 source and 1,244 installed test passes; the supplied exact three-source acceptance reports 1,362 constituted concepts, 3,434 relationships, 17/17 exports and zero XLSX/JSONL core-field mismatches. The release engineer scored against source cells, not as an independent external reviewer.

MetricObserved release-reported result
Established Definition Precision279/281 (99.3%)
Candidate Definition Precision37/38 (97.4%)
Domain Precision334/334 (100.0%)
Category Accuracy333/333 (100.0%)
Pdf Recall13/14 (92.9%)
Structured recall119/121 (98.3%)

Same-host, same-session queue-to-ontology-ready replay: 186.6s parallel and 194.2s serial versus 234.2s and 235.4s respectively for v1.89.302. Hardware detail not released here; no customer-host estimate or universal performance promise. Residual semantic errors and historical non-selected test failures/skips remain outside the selected campaign. Publisher identity is not independently established by the bundled signing key.

Inspect separate synthetic runtime run · Download scoped facts

Decision Authorization — exact release-reported evidence

v1.5.2 reports 463 selected source and 463 installed-wheel tests on Python 3.13.5. Reference-log anchor-added p95 23.1 ms; group commit 3,714 authorizations/s/tenant at a 20 ms window. These are reference-environment measurements, not end-to-end production throughput or latency. External independently operated production log, Python 3.12 qualification and signed OCI publication remain unestablished.

Inspect fresh v1.5.2 reference-vector verification

All numerical release facts on this staging build require owner disclosure approval before publication.

Evidence and buyer questions

What are SI qualification results?

The v1.89.332 release reports 20/20 commercial checks, 1,244 selected source and 1,244 installed tests, 1,362 concepts, 3,434 relationships, 17/17 exports and zero core XLSX/JSONL mismatches. Established definition precision is 279/281. These are scoped release-engineer results with residual errors, not independent certification or all historical tests passing.

Inspect source or product evidence

What are DA measured results?

The v1.5.2 release reports 463 selected source and 463 installed tests on Python 3.13.5. Reference-log anchor-added p95 is 23.1 ms and throughput is approximately 3,714 authorizations/s/tenant at a 20 ms window. Not end-to-end production latency; external production log, Python 3.12 campaign and signed OCI publication remain outstanding.

Inspect source or product evidence