Skip to content

PointSav Documentation

The engineering library for the PointSav platform — operating systems and services for regulated businesses that own their data, their AI, and their record-keeping outright. Where the monorepo holds the code, this wiki holds the reasoning: architecture, services, security, and the governance commitments that bind future development.

PointSav Digital Systems

Revision as of · 289ecb40

Historical revision — this record as it stood on 18 August 2026, not the current version. View the current record →

AI routing and the linguistic air-lock

What this article used to claim, and what's actually true. Earlier versions described a "sanitize-outbound / rehydrate-inbound" mechanism: a local pass that would strip PII and location identifiers before any text left the customer's network, replace them with pseudonymous tokens, and reverse the substitution when the external model's response came back — a "linguistic air-lock." No part of that mechanism exists. The only real sanitization code, service-slm/crates/slm-doorman/src/redact.rs, is a plain regex-based secret/credential redactor — PEM private keys, AWS/GitHub/Slack tokens, generic bearer/API-key/secret/password patterns — and its own doc comment states it is "the only redaction surface in the apprenticeship pipeline," called exclusively when writing training-corpus tuples. It never runs on the path to an external model. A corpus-wide search for "PII," "pseudonym," "location identif," or any rehydration-table code returns zero hits anywhere in service-slm. Separately, the real Tier C client (crates/slm-doorman/src/tier/external.rs) documents "no live API calls in v0.1.x" and gates every call behind a compile-time allowlist that cannot be extended at runtime — so the routing this article describes as live today is not live at all yet, on top of never having had the sanitization step it claimed.

This is a compliance-relevant gap, not a cosmetic one. This article explicitly targets regulated-industry readers — real estate, financial advisory, clinics, law firms — evaluating the platform's actual privacy protections. The honest current answer: when Tier C does go live, raw customer text will reach an external model unless a real PII-scrubbing mechanism is built first — the credential redactor protects secrets, not customer data. What follows describes what's real today, not the retracted mechanism.

AI routing, as it actually exists, is the mechanism by which service-slm — the Doorman — mediates every request that involves a language model, across three real compute tiers. service-slm is confirmed as the sole boundary that holds external-model credentials and logs every call to the per-tenant audit ledger; no other component in the system holds external AI credentials or makes direct outbound calls to external language models. This part of the architecture is real and does what it claims — the gap is specifically the sanitization/rehydration layer, not the single-boundary design.

What's real today

  • Three compute tiers, confirmed in crates/slm-doorman/src/flow_policy.rs and lib.rs: Tier A (local), Tier B (Yo-Yo burst/GPU), Tier C (external API) as real routing targets (RouteTarget::TierALocal/TierBNode).
  • Model names: Tier A runs olmo-3-7b-instruct locally; Tier B (Yo-Yo) defaults to Olmo-3-1125-32B-Think.
  • Tier C is not live. tier/external.rs documents no live external API calls in this version; the client is wired only against a test mock, and every call is gated behind a fixed, compile-time allowlist.
  • Every proxied call is audit-logged — confirmed in audit_proxy.rs, lib.rs, and cost_ledger.rs — with cost and response recorded to a real audit ledger.
  • Every request carries a mandatory tenant tag (ModuleId in slm-core), confirmed real.
  • Not found in code: a "tenant's configured budget cap" determining routing decisions — cost tracking and pricing config exist, but no explicit per-tenant budget gate on the routing decision itself was located.

What is not real (retracted, not softened)

The sanitize-outbound / rehydrate-inbound pass; PII detection; location-identifier stripping; pseudonymous token substitution; a rehydration table of any kind. None of it exists in service-slm. Any future version of this article that reinstates language like "sanitizes sensitive information" or "masks PII" for the Tier C routing path needs a fresh code citation, not a restoration of this text.

Applications — reframed as what the platform can honestly promise today

  • Editorial workflow: TOPIC and GUIDE drafts that reach Tier C today do so through the fixed allowlist's narrow-precision tasks (see Learning Datagraph — SLM trajectory loop and apprenticeship queue for draft_generate's real behavior) — not through any sanitization step, because none exists for this path.
  • Financial advisory / real estate / clinic / law-firm use of Tier C: not yet a safe claim to make in this article's own terms, given the retraction above. Any of these use cases sending ledger records, client names, or property/owner records through Tier C today would reach the external model unredacted for PII. This should stay flagged until a real mechanism is built and independently verified, not implied as already solved.

See also

Important Information

Corporate structure. PointSav Digital Systems ("PointSav") is currently a trade name of Woodfine Capital Projects Inc. ("Woodfine"), planned to become a wholly-owned Woodfine subsidiary upon incorporation. PointSav does not itself offer, sell, or solicit any security. Any securities offering associated with Woodfine's real-property direct-hold solutions is made exclusively by Woodfine, and only by means of the applicable Private Placement Memorandum.

No investment advice. This wiki's content is provided for engineering, operational, research, and development purposes. Nothing on this wiki constitutes investment advice or a solicitation to invest in any Woodfine partnership or direct-hold solution.

Intellectual property. The PointSav name, trade name, wordmark, and marks, together with all current and future PointSav- and Totebox-branded products, services, and offerings — and the software, source code, documentation, design system, and all related materials — are proprietary to Woodfine and its affiliates, except for components identified as open source. No rights are granted except as expressly set out in a written license or agreement. The full trademark notice appears in the footer of every page on this site.

Open source components. Portions of the platform are made available under permissive open-source licenses identified in the accompanying repository. Use of those components is governed by their respective license terms.

No warranty; informational use. Content on this wiki is provided for general informational purposes only and does not constitute a representation, warranty, or commitment with respect to product functionality, availability, pricing, or roadmap. Some articles describe planned or intended features, capabilities, and milestones — language such as "planned," "intended," "targeted," "may," and "expected" marks this forward-looking content, which is subject to change and does not constitute a commitment regarding future performance.

Confidentiality. Where an article describes an operational or deployment detail that is not intended for public disclosure, that article is not published on this wiki. Content here is general-purpose engineering documentation, not customer-specific configuration.

Jurisdiction. Woodfine Capital Projects Inc. is organized in British Columbia, Canada. References to the Sovereign Data Foundation on this wiki describe a planned or intended initiative only, not a current equity holder or active governance body.

Changes to this notice. PointSav may update this notice from time to time; the version posted on this page governs.

Not a filing system. This wiki is not a securities filing system, an electronic disclosure repository, or a substitute for SEDAR+ or any other regulatory filing system. Formal securities filings are made through the applicable regulatory filing system, not through this wiki.

Full disclaimer. This notice supplements, and does not replace, the full Disclaimers article. In the event of any conflict, the full Disclaimers article governs.

Read the full disclaimer →