MedCORE Next – Design Handoff

Extraction done. Data lives here; the look is yours.

Corpus: 249 topics / 32 subjects, extracted verbatim from MedCORE Production/Source JSON/Premium/ into Data/.

Uniform topic record (Data// .json)

meta          topicName | topicCode (MED-DERM-01 style) | category | subcategory | identityColor | subject | sourceFile
layers.opener examPromise | nrePattern{howTested, disguise, discrimination} |
              classicVignette{patient, presentation, keyClue, bestNextMove} | recognitionTrigger |
              keyDiscriminator | fatalMiss
layers.spine  paragraphs[] (markdown-bold prose) | mechanismChain[] | keyNumbers[{param, value}] | extra?
layers.recognition  classicPresentation | disguisedPresentation | synonyms[] |
              triggerTable[[clue, nextAction], ...] | extra?
layers.discriminator  differentialTable{columns[], rows[]} | trapPairs[{correct, trap, discriminator}] |
              whyWrong[{option, whyTempting, whyWrong}] | extra?
layers.management    firstLine | unstableVsStable{unstable, stable} | contraindications[] |
              escalationLadder[] | emergencyException | drugDoses{}? | extra?
layers.mcq    howNREAsks | disguise | discrimination | commonWrongOptionLogic | futureAlert? |
              embeddedMCQs? (CHD only: stem, options{a..d}, correctAnswer, explanation,
              cognitiveTask, discriminator, trapType, futureAlert)
layers.recall recallPrompts[{prompt, answer}] | examTraps[{trap, why, futureAlert}] | errorLogFutureAlert
completeness  missing[] | nonstandard[]  (never hidden; renderer decides)

Renderer contract (for your new theme)

1. Walk layers in order; every layer may be null (render nothing, never a blank frame).
2. extra objects (only Cardiology CHD uses them) = titled sections at that layer’s position.
3. recognition.triggerTable rows are [clue, action] pairs -> two-column trigger table.
4. differentialTable.columns = header row; each rows[i] = one data row, row[0] is the feature axis.
5. keyNumbers -> stat chips (value big, param small).
6. trapPairs -> correct-vs-trap flip cards; whyWrong -> option autopsy list.
7. embeddedMCQs -> interactive quiz block (self-grading, options map).
8. Long prose fields (howTested etc.) contain numbered enumerations and **bold** markdown – render bold inline.
9. identityColor per-topic accent; your theme may unify instead.

Known gaps (flagged, not fixed – content is verbatim)

• Gastroenterology/Hepatitis_C_Other.json: only opener + spine + recognition.
• Cardiology/CHD.json: fully custom structure (surfaced via extra + embeddedMCQs).

Stats

• Topics per subject: 249 across 32 subjects (7-8 each; Supplementary 3).

Re-run: python3 Scripts/extract_premium_for_next.py (idempotent; regenerates Data/).