PF research library · 3 min read
PF Simulation r34 — ISAES next-phase readiness assessment
r34 documentation · isaes-readiness.md · Formatted reading copy
On this page
This dated record retains its original release identity and evidence scope. For the latest r34 package, see current release validation, the cumulative ledger and the hosted trial guide. Earlier hashes and test counts below continue to identify the work originally recorded.
Ready for controlled next-phase evaluation on the existing architecture. The SDK correction and reporting-specific assistant integration are implemented. This maintenance release closes the documentation gap with an illustrated ISAES chapter and working Documentation navigation. Full original-scope acceptance and broad production activation remain pending; the optional report assistant is disabled by default.
Evaluated maintenance archive: 1987acf5712a7d2311744089bfb350885501bd925e18a2c31ee43447329d6a74. Fresh documentation/navigation execution: 2026-09-28T05:22:15.473246+00:00. Reporting/ISAES 2.2.0 and SDK 1.0.0 are unchanged.
Resolved findings
| Earlier concern | Current result |
|---|---|
| Missing native SDK prerequisites could pass unchecked | The preceding feature release corrected required/null metadata enforcement and dependent-work blocking. |
| No reporting-specific assistant connection | Owner- and report-bound discussion uses bounded evidence/proposal tools and existing provider/BYOK infrastructure. |
| Incomplete reviewed proposal application | Review, idempotent application, queued child revision and separate execution are implemented. |
| Dense saved-view restoration had little margin | The preceding release measured 438.17 ms p95 against 2000 ms; that baseline is retained. |
| Primary manual and document navigation lagged ISAES | Edition 1.18.0 (236 pages) adds Chapter 32; six current guides are linked in standalone and hosted Documentation. |
The shared core supplies scientific contracts, dependency planning, provenance, extension registration and presentation rules. The implemented local scientific and visualization integrations support the bounded catalog. The available evidence does not indicate a need for another architectural rewrite or a mandatory remote-service layer. Hardware and specialist integrations remain workload-specific.
Evidence and limits
Fresh maintenance validation passed 94 recorded Python cases, without failure, error or skip, plus 1 JavaScript command. It covers documentation/navigation, real Chromium interactions, integrity and reader/startup compatibility. The new archive was read back completely and verified against its member hashes. Engine files are unchanged from the incoming feature release.
The 1436-case/four-JavaScript-command feature acceptance and all scientific/performance receipts remain byte-identical under validation_r34_isaes_ai. They are baseline evidence for archive 870b7de102f37d83f0988fbbfac91186030b534c265caf934784f3be09b70eb0, not a fresh full rerun for this maintenance package. Dense 100,000-point restoration p95 was 438.17 ms, first view 1675.53 ms and linked selection 385.73 ms across three repetitions. Analysis and standard-export p95 were 1.273 s and 2.272 s. These small-sample local results are not production SLAs.
Software checks do not establish scientific truth for arbitrary private code or generated prose. Evidence membership checks do not certify semantic entailment or free-text citation correctness. Approved interpretations remain attributed commentary. The classical reporting path continues to work without an AI provider.
Next phase
Use the updated manual for a controlled researcher evaluation. Have representative users complete the defined reporting tasks without coaching, record failures and scientific misunderstandings, and review the observations independently. Qualify native Windows with actual supported-runtime execution.
For the optional assistant, explicitly enable PF_ISAES_AI_ENABLED=1, select a supported provider/model, and use the ordinary credential and admission controls. Evaluate actual responses against realistic retained reports, including failed or insufficient scientific checks. Assess citation accuracy, unsupported interpretations and useful proposals; retain invocation evidence and review outcomes. Review, application and explicit execution remain separate user actions. Cooperative pause may await an in-flight request.
Before broader deployment, complete the applicable production-isolation and live-BYOS checks, or document a justified deployment-specific out-of-scope decision. Native Windows, at least five representative researchers with at least 80% unassisted task completion and zero critical scientific misinterpretations, applicable production/BYOS qualification, and live-provider/model scientific evaluation remain open. Updating documentation does not close those gates.
The implementation report gives current scoped counts, baseline limits, archive identity and verification instructions. Use docs/ISAES_ACCEPTANCE.md inside the package for external qualification procedures and manual Chapter 32 for the user workflow.