Skip to content

Archived Brief

This brief has been completed and is retained as a build record.

BRIEF: B-ENRICH-KN-01

Knowledge Navigator — Pilot Enrichment + UI Scaffold

Date: 20 March 2026
Operator: Tim + Pace
Status: Ready to drop to Alex — after backup confirmed complete


SURFACE

rtopacks.com.au — unit drawer UI + rtopacks-db enrichment
CLOUDFLARE ACCOUNT: e5a9830215a8d88961dc6c80a8c7442a

DO NOT TOUCH

  • engine-db, ops-db, mcp-db
  • Any FTS5 virtual tables (rtos_fts, qualifications_fts, units_fts and their shadow tables)
  • Any existing unit columns
  • Do not run wrangler d1 export on rtopacks-db under any circumstances

BACKGROUND

Knowledge Navigator (KN) is a new RTOpacks product layer that sits inside the unit drawer. It provides plain-English interpretations of TGA performance criteria — assistive, not regulatory.

This brief covers three things in one pass: 1. Two new DB columns for KN data 2. Grok enrichment of 10 pilot units at two model tiers (fast vs reasoning) 3. UI scaffold to render both interpretations in the drawer so Tim and Pace can evaluate quality in situ on the live site

Service mark: Knowledge Navigator℠ is a service mark of RTOpacks, an UCCA Inc. platform.


PART 1 — DATABASE

Add two new columns to the units table:

ALTER TABLE units ADD COLUMN kn_criteria_fast TEXT;
ALTER TABLE units ADD COLUMN kn_criteria_reasoning TEXT;

Both store JSON. Schema per column:

{
  "elements": [
    {
      "element_number": 1,
      "element_title": "Plan training",
      "element_summary": "One sentence summary of what this element is actually asking a person to do.",
      "criteria": [
        {
          "criterion_id": "1.1",
          "criterion_text": "Confirm trainee readiness...",
          "kn_interpretation": "Plain English interpretation here."
        }
      ]
    }
  ]
}

PART 2 — GROK ENRICHMENT SCRIPT

Script: scripts/enrich-grok-kn.js

Run against these 10 unit codes only — two passes, one per model:

Code Title Complexity
CUAACD101 Use basic drawing techniques Simple
MSMWHS110 Follow emergency response procedures Simple
CHCFAM002 Work with a child-focused approach Simple-mid
HLTAID014 Provide Advanced First Aid Medium
TAEASS412 Assess competence Medium
TAEASS514 Develop and implement plans for RPL Medium
CHCCCS044 Follow established person-centred behaviour supports Complex
UEEEL0039 Design, install and verify compliance of general electrical installations Complex
MARB039 Maintain and repair shipboard machinery Complex
AVIM0003 Conduct aeronautical knowledge training and flight training Complex

Pass 1 — fast model: grok-4-1-fast-non-reasoning → writes to kn_criteria_fast

Pass 2 — reasoning model: current best reasoning model on xAI API → writes to kn_criteria_reasoning

Run sequentially, not in parallel. One unit at a time. Log each unit code and model on completion.

Prompt — per criterion

You are an expert interpreter of Australian vocational education and training (VET) competency standards.

Below is a performance criterion from the TGA unit "${unit_code}: ${unit_title}".

Element: ${element_number}. ${element_title}
Criterion ${criterion_id}: "${criterion_text}"

Write a plain-English interpretation (2-4 sentences) that explains what a trainer or 
learner needs to actually understand or do to meet this criterion.
- Be specific and practical, not generic
- Do not restate the criterion verbatim
- Do not use phrases like "this means you need to..." — just say it directly
- If the criterion references a standard, regulation or framework, briefly note what 
  that is in plain terms

Prompt — element summary (after all criteria in an element processed)

In one sentence, summarise what Element ${element_number} "${element_title}" requires 
a person to demonstrate overall.

Safety guard

Only write to kn_criteria_fast / kn_criteria_reasoning where that column IS NULL for the target unit codes. Never overwrite existing data.


PART 3 — UI SCAFFOLD

In the unit drawer, after the existing PE/KE sections, add a Knowledge Navigator toggle section.

Behaviour

  • Default state: collapsed, showing button "Knowledge Navigator ℠" with a toggle slider
  • On toggle: expands inline, rendering KN content under each performance criterion
  • Both model outputs rendered side by side or stacked, labelled "Interpretation A" and "Interpretation B" — do NOT label as fast/reasoning, keep it blind for the evaluation
  • Element summary renders at the top of each element block in KN mode
  • Colour: different background tint from TGA verbatim content — subtle warm tone, nothing jarring
  • Footer of KN section (always visible when expanded):

Knowledge Navigator interpretations are RTOpacks' editorial perspective on TGA content. They are assistive only — not directive, not regulatory. © 2026 UCCA Inc.

Inviolable rule

Existing TGA verbatim content must remain fully intact and visible above KN content at all times. KN never replaces or obscures TGA text.


PART 4 — QA HANDBACK

When complete, Alex to provide: - Direct links to all 10 unit pages on rtopacks.com.au - Confirmation of which column is A and which is B (private note to Tim only, not on site) - Confirmation both Grok passes completed without errors

Tim and Pace will review all 10 units on the live site. No further DB work until review is complete and winner column is confirmed.


DEFINITION OF DONE

  • Two new columns added to units table (kn_criteria_fast, kn_criteria_reasoning)
  • 10 units enriched in kn_criteria_fast (Pass 1 complete)
  • 10 units enriched in kn_criteria_reasoning (Pass 2 complete)
  • UI toggle rendering both interpretations in drawer on live site
  • 10 unit page links provided to Tim
  • Grok superseded data untouched and intact

WHAT COMES NEXT (not this brief)

  • Tim and Pace review all 10 units in situ
  • Winning model confirmed → full 15,200 unit KN run
  • Losing column dropped from schema
  • kn_criteria becomes the canonical column name
  • USPTO Class 41 filing for Knowledge Navigator℠ at RTOpacks launch
  • B-OPS-RESTORE-01: rtopacks-db HTTP API restore path (low priority, next session)