---
type: source-summary
title: 'Source: Tier 3 reference batch (~28 docs)'
modified: 2026-04-11
tags:
  - reference
  - tier3
  - evaluation
  - tutorials
  - adrs
  - source
  - preston-test
---

# Source: Tier 3 reference batch (~28 docs)

## What this source is

~28 reference documents from the Preston-Test repo cached to `raw/preston-test/` for Squire's evaluation. These are the deeper-reference layer: ADRs (architecture decision records), tutorials, feature implementation guides, integration docs, the CHANGELOG, and developer setup guides. They support the Tier 1 and Tier 2 evaluation by providing the "how do I learn to use this?" and "what decisions were made and why?" layers.

All files cached to `raw/preston-test/` as verbatim copies from the repo. Claim-by-claim extraction not performed for Tier 3 — these are reference material with high per-document specificity but low cross-wiki impact. When a specific page needs detail from a Tier 3 source, the raw file can be read and the relevant claims extracted at that time.

## Document inventory by category

### ADRs (4 docs — architectural decision records)
- **ADR-001**: Server-side secrets only (all API keys server-side, never client-exposed)
- **ADR-002**: Unified ProviderRegistry (single registry for all AI provider routing)
- **ADR-003**: Unified telemetry schema (single schema across all agents and connectors)
- **ADR-005**: Demo connector decorator pattern (wraps real connectors for demo mode without forking)

### Architecture (2 docs)
- **AI-SYSTEMS-COMPARISON.md**: Side-by-side of AINaturalLanguageService (action) vs Help Chat RAG (retrieval) — two separate AI subsystems
- **ADR-001-NETSUITE-MCP-INTEGRATION.md**: MCP-to-NetSuite integration ADR (OAuth 1.0a, SuiteScript REST, tool registration, policy enforcement)

### Tutorials (10 docs)
- MDM tutorial, NetSuite setup, encrypted credential storage, AI provider configuration, AI agents comprehensive guide, comprehensive AI configuration, workflow integration, advanced field mapping AI features, enterprise features guide (already in Tier 1), RAG knowledge base guide (Phase 4, 92-97% → 95-99% accuracy), AI NLQ cross-module intelligence

### Features (3 docs)
- Advanced field mapping editor AI implementation, feature showcase (full feature inventory), help chat system (RAG vector store + embedding pipeline)

### Governance (1 doc)
- Governance quick-reference card (route-by-route status, policy flags, error codes)

### Developer (2 docs)
- Developer setup (prerequisites, clone/install, Husky hooks, env vars, Docker option), troubleshooting guide (common errors, AI failures, connector auth, test debugging)

### Integrations (4 docs)
- NetSuite-BusinessCentral dual-platform strategy, connector ecosystem strategy (6 production vs 200+ planned), mapping center module deep-dive, embedding guide (Context Sidecar deployment pattern)

### Top-level (2 docs)
- CHANGELOG (semver changelog with DLP commit notes, architectural changes), MCP-USER-SETTINGS-GUIDE (per-user MCP feature flags)

## How to use this batch

These files are cached in `raw/preston-test/` and can be pulled into wiki pages on-demand when a reviewer asks a question that requires deeper detail. Example flows:

- "How does the encrypted credential storage work?" → Read `raw/preston-test/encrypted-credential-storage-guide.md`
- "What decisions led to the dual-AI design?" → Read `raw/preston-test/adr-004-dual-ai-system-design.md` (already in Tier 1)
- "How do I set up the RAG knowledge base?" → Read `raw/preston-test/22-rag-knowledge-base-guide.md`
- "What's in the CHANGELOG?" → Read `raw/preston-test/changelog.md`

No wiki pages created from this batch — the files are cached for on-demand reference.

## Pages updated

**Updated** (1 existing page):
- [[pages/concepts/suitecentral-2-overview]] — added a "Reference documentation" section noting that ~28 Tier 3 docs are cached in `raw/preston-test/` for on-demand access

## Notes

- **68 total raw files** are now cached in `raw/preston-test/` across all three tiers. This is the most comprehensive raw-source cache in the wiki — essentially a snapshot of the Preston-Test repo's `docs/` directory (minus archive, sessions, and planning noise).
- The **CHANGELOG** is useful for verifying when specific features were added — it has semver entries with dates that can corroborate timeline claims from other sources.
- The **RAG knowledge base guide** documents a Phase 4 accuracy improvement from 92-97% to 95-99% — which corroborates the 95-99% field mapping accuracy claim from [[sources/04-technical-proof]].
- ADRs 001-005 together document the five most important architectural decisions: secrets management, provider routing, telemetry, dual-AI system, and demo connector patterns. A CTO can walk these in ~30 minutes for a deep understanding of the engineering philosophy.