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

Compounding Doorman

← All revisions

190b1cbb · PointSav Digital Systems ·

editorial(substrate): fix compounding-doorman fabricated sanitise/rehydrate claim (Track-B) — confirmed against redact.rs: the mechanism only strips credentials for the apprenticeship-corpus write path, no per-request outbound-tokenize/inbound-rehydrate code exists anywhere (matches the already-established finding on documentation/ai/sovereign-ai-routing.md); resolved the audit-ledger path discrepancy directly (SLM_AUDIT_DIR, daily <YYYY-MM-DD>.jsonl files, no tenant/month structure) rather than leaving it an open flag; register-clean EN+ES

View the full record as of this revision →

@@ -24,7 +24,7 @@ In the PointSav platform, `service-slm` is the Compounding Doorman. It is the so

The Doorman enforces four disciplines simultaneously on every call:

**Sanitise outbound, rehydrate inbound.** [[architecture-decisions|SYS-ADR-07]] prohibits structured data from routing through AI. Before any request leaves the Doorman boundary, identifiers, foreign keys, and structured fields are replaced with opaque tokens. After the response returns, the tokens are resolved back to their original values. The AI model sees only prose and opaque tokens; it never receives or returns structured records. This is what makes the audit-routing path reversible and auditable (see [[disclosure-substrate]]).
**Redaction on the apprenticeship path.** [[architecture-decisions|SYS-ADR-07]] prohibits structured data from routing through AI. The Doorman's redaction logic strips credentials and identifiers from every brief, attempt, and verdict before it's captured into the apprenticeship corpus — the path that produces training data. A general sanitise-outbound/rehydrate-inbound tokenization on every Ring 3 call is the design intent, not yet built beyond that corpus-capture path.

**Three-tier compute routing.** The Doorman selects among three compute tiers per request shape and budget policy (see [[four-tier-slm-substrate]]):
- **Tier A — local**: OLMo 3 7B on the customer's own hardware. Zero marginal cost, full data locality. Default for most operations.
@@ -53,9 +53,8 @@ The compounding loop closes because the audit ledger that records what the Doorm
The metaphor maps precisely to the operational discipline:

- A doorman greets every guest — every request enters through this boundary.
- A doorman checks identification — the Doorman validates `moduleId` and sanitises inputs.
- A doorman checks identification — the Doorman validates `moduleId`.
- A doorman directs guests to the right destination — routing to local, GPU burst, or external API per policy.
- A doorman carries bags through the door — rehydrating the response with full identifiers stripped at outbound.
- A doorman keeps a log of who came and went — the audit ledger.
- A doorman never enters the rooms themselves — the Doorman never writes to the authoritative knowledge graph; it only proposes, per [[architecture-decisions|SYS-ADR-07]].

@@ -69,17 +68,7 @@ This is what allows a deployment to operate Ring 3 at any trust level — from f

## Implementation state

The Doorman (`service-slm`) is operational on the workspace VM. It binds at `127.0.0.1:9080`, routes Tier A to the local model service, and enforces sanitise-outbound and rehydrate-inbound discipline on every call. The audit ledger writes to `data/audit-ledger/<tenant>/<YYYY-MM>.jsonl`.

**Correction (2026-07-18):** the bind address checks out (`127.0.0.1:9080`, confirmed
against `SLM_BIND_ADDR` in the live deploy config), but the audit ledger path does not.
The real env var is `SLM_AUDIT_DIR` (`slm-doorman-server/src/main.rs`), with the
workspace dogfood deployment explicitly setting it to `/var/lib/local-doorman/audit/`
(default when unset: `$HOME/.service-slm/audit/`) — not `data/audit-ledger/<tenant>/
<YYYY-MM>.jsonl` as this article states. Whether the per-tenant/per-month filename
structure inside that directory matches this article's `<tenant>/<YYYY-MM>.jsonl` shape
was not independently re-verified here. **Flagged, not silently rewritten** — needs
project-totebox confirmation of the real internal layout before this line is corrected.
The Doorman (`service-slm`) is operational on the workspace VM. It binds at `127.0.0.1:9080` and routes Tier A to the local model service. The audit ledger writes daily files (`<YYYY-MM-DD>.jsonl`) to a directory set by `SLM_AUDIT_DIR` — `/var/lib/local-doorman/audit/` on the workspace deployment, `$HOME/.service-slm/audit/` by default when unset.

Tier B integration (Yo-Yo GPU burst) is in progress. Tier C external API routing is available for narrow precision tasks where the operator has explicitly configured an allowlist.
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 →