Prepare scanner branch for merge

src/lib/artifactOcrParser.ts

@@ -82,6 +82,7 @@ export function parseArtifactCandidate(capture: CaptureResult | null): ParsedArt
   const parsedLevel = levelField.value ? Number.parseInt(levelField.value, 10) : null;
   const level = parsedLevel ?? 0;
   let mainStatField = inferMainStat(slotField.value, mainText);
+  mainStatField = promoteSlotPercentMainStat(slotField.value, mainStatField);
   let mainValueField = findMainValue(mainText, mainStatField.value, slotField.value, parsedLevel);
   if (!mainStatField.value && mainValueField.value) {
     const inferredFromValue = inferMainStatFromValue(slotField.value, mainValueField.value, mainText, parsedLevel);
@@ -227,6 +228,8 @@ function parseSetName(setText: string, artifactName: ParsedField): ParsedField {
   const setFromText = fuzzyFindKnown(`${directLine ?? ""}\n${setText}`, knownSets, 0.64);
   if (setFromText && (!setFromPiece || setFromText.score >= 0.78)) return field(setFromText.value, Math.round(setFromText.score * 100), setFromText.score >= 0.95 ? "ocr" : "fallback");
   if (setFromPiece) return field(setFromPiece, derivedConfidence(artifactName, 92), "derived");
+  const partialSetFromPiece = deriveSetFromPartialPieceName(artifactName.value);
+  if (partialSetFromPiece) return field(partialSetFromPiece, 72, "derived");
   return setFromText ? field(setFromText.value, Math.round(setFromText.score * 100), "fallback") : field("", 0, "missing");
 }
 
@@ -275,6 +278,22 @@ function firstUsefulLine(text: string, rejectIncludes: string[]) {
     .find((line) => line.length > 5 && !rejectIncludes.some((reject) => simplifyForMatch(line).includes(simplifyForMatch(reject)))) ?? "";
 }
 
+function deriveSetFromPartialPieceName(text: string) {
+  const words = cleanupOcrLabel(text)
+    .split(/\s+/)
+    .map((word) => simplifyForMatch(word))
+    .filter((word) => word.length >= 5);
+  if (words.length < 2) return "";
+
+  const candidates = knownPieceNames.filter((piece) => {
+    const normalizedPiece = simplifyForMatch(piece);
+    const hits = words.filter((word) => normalizedPiece.includes(word)).length;
+    return hits >= 2;
+  });
+  const sets = [...new Set(candidates.map((piece) => pieceToSet.get(piece)).filter((set): set is string => Boolean(set)))];
+  return sets.length === 1 ? sets[0] : "";
+}
+
 function findMainValue(text: string, mainStat: string, slot: string, level: number | null): ParsedField {
   const cleaned = text.replace(/\b20\b/g, " ").replace(/[Oo]/g, "0");
   const percentValue = extractPercentValue(cleaned);
@@ -504,7 +523,7 @@ function parseEquippedCharacter(text: string): ParsedField {
   const wholeTextMatch = fallbackSearch ? fuzzyFindKnown(fallbackSearch, knownCharacters, 0.88) : null;
   if (wholeTextMatch) return field(wholeTextMatch.value, Math.round(wholeTextMatch.score * 100), "fallback");
 
-  return afterLabel ? field(afterLabel, 50, "fallback") : field("Not detected", 45, "missing");
+  return field("Not detected", 45, "missing");
 }
 
 function field(value: string, confidence: number, source: ParsedField["source"]): ParsedField {
@@ -520,6 +539,20 @@ function promotePercentVariant(stat: string, text: string) {
   return stat;
 }
 
+function promoteSlotPercentMainStat(slot: string, mainStat: ParsedField): ParsedField {
+  if (!["ATK", "HP", "DEF"].includes(mainStat.value)) return mainStat;
+  const references = getSlotMainStatValueReferences(slot);
+  const hasFlat = references.some((candidate) => candidate.stat === mainStat.value);
+  const hasPercent = references.some((candidate) => candidate.stat === `${mainStat.value}%`);
+  if (hasFlat || !hasPercent) return mainStat;
+  return {
+    ...mainStat,
+    value: `${mainStat.value}%`,
+    confidence: Math.max(mainStat.confidence, 90),
+    source: "derived",
+  };
+}
+
 function getSlotMainStatValueReferences(slot: string): MainStatValueReference[] {
   const valueReferences = mainStatValueReferences[slot];
   return Array.isArray(valueReferences) ? valueReferences as MainStatValueReference[] : [];
@@ -607,6 +640,7 @@ function cleanupOcrLabel(line: string) {
 function cleanupCharacterNoise(text: string) {
   return text
     .replace(/^.*?equipped\s*:?\s*/i, "")
+    .replace(/^(?:by|to)\s+/i, "")
     .replace(/[^A-Za-z'\-\s]/g, " ")
     .replace(/\s+/g, " ")
     .trim();