hh-shipments 0.3.5 → 0.3.7

package/package.json

@@ -18,7 +18,7 @@
     "test": "node test/run-tests.js && node test/run-carriers-tests.js"
   },
   "type": "commonjs",
-  "version": "0.3.5",
+  "version": "0.3.7",
   "bin": {
     "hh-shipments": "server.js"
   },

package/public/app.js

@@ -169,7 +169,10 @@ async function ocrFile(file, statusEl) {
 }
 
 async function handlePhotoFile(f) {
+  const target = $('photo-result');
   try {
+    const url = URL.createObjectURL(f);
+    target.innerHTML = `<div class="review"><img src="${url}" style="max-width:100%;border-radius:10px"><p class="muted">📖 Reading screenshot…</p></div>`;
     const text = await ocrFile(f, $('ocr-status'));
     // Upload original for reference, then run the quick flow on OCR text
     const fd = new FormData();
@@ -203,25 +206,31 @@ function pastedImageFiles(e) {
   return files;
 }
 
-// Paste-text tab: pasting a screenshot runs OCR and extracts tracking numbers;
-// pasting plain text behaves normally
+// Paste-text tab: pasting a screenshot runs OCR and extracts automatically —
+// no "Extract" tap needed. Plain text paste still lands in the textarea.
 async function handlePastedImageFile(f) {
+  const target = $('quick-result');
   try {
+    // Instant feedback: show a thumbnail + progress the moment the image lands
+    const url = URL.createObjectURL(f);
+    target.innerHTML = `<div class="review"><img src="${url}" style="max-width:100%;border-radius:10px"><p class="muted" id="ocr-live">📖 Reading screenshot…</p></div>`;
     const text = await ocrFile(f, $('ocr-status-paste'));
+    const live = $('ocr-live');
+    if (live) live.textContent = '🔍 Finding tracking number…';
     const fd = new FormData();
     fd.append('image', f);
     const up = await fetch('/api/intake/image', { method: 'POST', body: fd });
     if (!up.ok) throw new Error('upload failed');
-    const r = await runQuick(text, $('quick-result'));
-    if (r && !r.added.length && !r.needs_review.length && !r.duplicates.length) showOcrRead($('quick-result'), text);
-  } catch (err) { $('ocr-status-paste').textContent = ''; $('quick-result').innerHTML = `<p class="error">${esc(err.message)}</p>`; }
+    const r = await runQuick(text, target);
+    if (r && !r.added.length && !r.needs_review.length && !r.duplicates.length) showOcrRead(target, text);
+  } catch (err) { $('ocr-status-paste').textContent = ''; target.innerHTML = `<p class="error">${esc(err.message)}</p>`; }
 }
 
 $('tab-paste').addEventListener('paste', async (e) => {
   const files = pastedImageFiles(e);
   if (!files.length) return; // text paste — let it land in the textarea
   e.preventDefault();
-  $('paste-box').innerHTML = '📋 <span class="muted">or tap here, then paste a screenshot</span>';
+  $('paste-box').innerHTML = '📋 <span class="muted">or tap here and paste a screenshot</span>';
   for (const f of files) await handlePastedImageFile(f);
 });
 
@@ -383,7 +392,11 @@ let searchT;
 $('search').addEventListener('input', () => { clearTimeout(searchT); searchT = setTimeout(loadDash, 400); });
 let searchAT;
 $('search-arch').addEventListener('input', () => { clearTimeout(searchAT); searchAT = setTimeout(loadArchive, 400); });
-$('btn-quick').onclick = () => runQuick($('paste-text').value, $('quick-result'));
+$('btn-quick').onclick = () => {
+  const t = $('paste-text').value.trim();
+  if (!t) { $('quick-result').innerHTML = '<p class="muted">Paste the seller\'s text above first — or tap <b>📋 Paste screenshot</b>, screenshots extract on their own.</p>'; return; }
+  runQuick(t, $('quick-result'));
+};
 $('btn-manual').onclick = async () => {
   $('manual-err').textContent = '';
   try {

package/public/extract.js

@@ -56,10 +56,85 @@
     return out;
   }
 
+  // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
+  var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
+  function repairUpsZ(ups) {
+    // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+    // only when the rest has strict UPS structure AND the shipper segment
+    // contains a letter — real UPS shipper numbers do, random digit runs
+    // don't. Never invent numbers from thin air.
+    var cand = null;
+    if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
+    else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
+    if (!cand || !UPS_STRICT.test(cand)) return null;
+    if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
+    return cand;
+  }
+
+  // UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
+  function spacedUpsIn(text, found, seen, conf) {
+    var m, re = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
+    while ((m = re.exec(text)) !== null) {
+      var ups = normalize(m[0]);
+      var num = /^1Z[0-9A-Z]{16}$/.test(ups) ? ups : repairUpsZ(ups);
+      if (num && !seen[num]) {
+        seen[num] = true;
+        found.push({ number: num, carrier: 'ups', confidence: conf, ocr_fixed: num !== ups });
+      }
+    }
+  }
+
   function extractTrackingNumbers(text) {
     var found = [];
     var seen = {};
     if (!text) return found;
+    // 0) ANCHOR PASS — on shipping labels the tracking number sits on or just
+    // under a line like "UPS GROUND" or "TRACKING #:". Scan those hot zones
+    // first so label structure decides, not blind whole-image regex. Anything
+    // found here outranks whole-image guesses (dedupe keeps the first hit).
+    anchorZones(text).forEach(function (z) {
+      runPatterns(z.text, found, seen, z.carrier);
+      spacedUpsIn(z.text, found, seen, 'high');
+    });
+    // 1) Whole-text strict patterns
+    runPatterns(text, found, seen, false);
+    // 2) OCR-tolerant recovery pass
+    extractOcrTolerant(text, found, seen);
+    var rank = { high: 0, medium: 1, low: 2 };
+    found.sort(function (a, b) { return rank[a.confidence] - rank[b.confidence]; });
+    return found;
+  }
+
+  // Shipping-label structure: the tracking number lives on or just under
+  // lines like "UPS GROUND" / "TRACKING #:". Returns hot-zone chunks with the
+  // carrier named on/near the anchor (null when the anchor is generic).
+  function anchorCarrier(line) {
+    var l = String(line).toLowerCase();
+    if (/\bups\b/.test(l)) return 'ups';
+    if (/\bfed\s?ex\b/.test(l)) return 'fedex';
+    if (/\busps\b/.test(l) || /priority mail/.test(l)) return 'usps';
+    if (/\bdhl\b/.test(l)) return 'dhl';
+    return null;
+  }
+  function isAnchorLine(line) {
+    if (/track/.test(String(line).toLowerCase())) return true;
+    return anchorCarrier(line) !== null;
+  }
+  function anchorZones(text) {
+    var lines = String(text).split(/\r?\n/);
+    var zones = [];
+    for (var i = 0; i < lines.length; i++) {
+      if (!isAnchorLine(lines[i])) continue;
+      // anchor line + the two lines under it (the number is printed below)
+      var slice = lines.slice(i, Math.min(lines.length, i + 3));
+      var carrier = null;
+      slice.forEach(function (sl) { carrier = carrier || anchorCarrier(sl); });
+      zones.push({ text: slice.join('\n'), carrier: carrier });
+    }
+    return zones;
+  }
+
+  function runPatterns(text, found, seen, boostCarrier) {
     PATTERNS.forEach(function (p) {
       p.re.lastIndex = 0;
       var m;
@@ -67,11 +142,13 @@
         var num = normalize(m[1]);
         if (seen[num]) continue;
         seen[num] = true;
-        // Context boost: carrier word near the number
+        // Context boost: carrier word near the number, or a hot-zone anchor
+        // naming this carrier (a label's "UPS GROUND" line means the number
+        // under it really is UPS).
         var idx = m.index;
         var window = text.slice(Math.max(0, idx - 60), idx + 60 + m[1].length);
         var confidence = p.confidence;
-        var boosted = false;
+        var boosted = !!(boostCarrier && p.carrier === boostCarrier);
         CARRIER_WORDS.forEach(function (cw) {
           if (cw.carrier === p.carrier && cw.re.test(window)) boosted = true;
         });
@@ -80,11 +157,6 @@
         found.push({ number: num, carrier: p.carrier, confidence: confidence });
       }
     });
-    // Sort: high confidence first, then by position
-    extractOcrTolerant(text, found, seen);
-    var rank = { high: 0, medium: 1, low: 2 };
-    found.sort(function (a, b) { return rank[a.confidence] - rank[b.confidence]; });
-    return found;
   }
 
   function isValidFormat(number, carrier) {
@@ -111,9 +183,6 @@
   }
 
   function extractOcrTolerant(text, found, seen) {
-    // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
-    // Used to validate "1Z"-as-"12" OCR repairs so we don't invent false positives.
-    var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
     function push(num, carrier, confidence, extra) {
       num = normalize(num);
       if (!num || seen[num]) return;
@@ -122,28 +191,9 @@
       if (extra) for (var k in extra) c[k] = extra[k];
       found.push(c);
     }
-    function repairUpsZ(ups) {
-      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
-      // only when the rest has strict UPS structure AND the shipper segment
-      // contains a letter — real UPS shipper numbers do, random digit runs
-      // don't. Never invent numbers from thin air.
-      var cand = null;
-      if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
-      else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
-      if (!cand || !UPS_STRICT.test(cand)) return null;
-      if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
-      return cand;
-    }
-    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
-    var m, spacedUps = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
-    while ((m = spacedUps.exec(text)) !== null) {
-      var ups = normalize(m[0]);
-      if (/^1Z[0-9A-Z]{16}$/.test(ups)) push(ups, 'ups', 'medium');
-      else {
-        var fixed = repairUpsZ(ups);
-        if (fixed) push(fixed, 'ups', 'medium');
-      }
-    }
+    // a) UPS number split by spaces/dashes (whole-text fallback; the anchor
+    // pass already tried the hot zones at high confidence)
+    spacedUpsIn(text, found, seen, 'medium');
     // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
     var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
     while ((run = runRe.exec(text)) !== null) {

package/public/index.html

@@ -61,11 +61,11 @@
       <button class="tab" data-tab="manual">Manual</button>
     </div>
     <div id="tab-paste" class="tabpanel">
-      <textarea id="paste-text" rows="6" placeholder="Paste the seller's message here…"></textarea>
-      <div id="paste-box" class="pastebox" contenteditable="true" spellcheck="false">📋 <span class="muted">or tap here, then paste a screenshot</span></div>
-      <div class="row" style="margin-top:8px"><button id="btn-clipboard" class="secondary">📋 Grab screenshot from clipboard</button></div>
+      <textarea id="paste-text" rows="4" placeholder="Paste the seller's message here…"></textarea>
+      <div class="row" style="margin:8px 0"><button id="btn-clipboard" class="primary">📋 Paste screenshot — extracts automatically</button></div>
+      <div id="paste-box" class="pastebox" contenteditable="true" spellcheck="false">📋 <span class="muted">or tap here and paste a screenshot</span></div>
       <div id="ocr-status-paste" class="muted"></div>
-      <button id="btn-quick" class="primary">Extract &amp; add</button>
+      <button id="btn-quick" class="secondary">Extract from text above</button>
       <div id="quick-result"></div>
     </div>
     <div id="tab-photo" class="tabpanel" hidden>

package/src/extract.js

@@ -56,10 +56,85 @@
     return out;
   }
 
+  // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
+  var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
+  function repairUpsZ(ups) {
+    // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+    // only when the rest has strict UPS structure AND the shipper segment
+    // contains a letter — real UPS shipper numbers do, random digit runs
+    // don't. Never invent numbers from thin air.
+    var cand = null;
+    if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
+    else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
+    if (!cand || !UPS_STRICT.test(cand)) return null;
+    if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
+    return cand;
+  }
+
+  // UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
+  function spacedUpsIn(text, found, seen, conf) {
+    var m, re = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
+    while ((m = re.exec(text)) !== null) {
+      var ups = normalize(m[0]);
+      var num = /^1Z[0-9A-Z]{16}$/.test(ups) ? ups : repairUpsZ(ups);
+      if (num && !seen[num]) {
+        seen[num] = true;
+        found.push({ number: num, carrier: 'ups', confidence: conf, ocr_fixed: num !== ups });
+      }
+    }
+  }
+
   function extractTrackingNumbers(text) {
     var found = [];
     var seen = {};
     if (!text) return found;
+    // 0) ANCHOR PASS — on shipping labels the tracking number sits on or just
+    // under a line like "UPS GROUND" or "TRACKING #:". Scan those hot zones
+    // first so label structure decides, not blind whole-image regex. Anything
+    // found here outranks whole-image guesses (dedupe keeps the first hit).
+    anchorZones(text).forEach(function (z) {
+      runPatterns(z.text, found, seen, z.carrier);
+      spacedUpsIn(z.text, found, seen, 'high');
+    });
+    // 1) Whole-text strict patterns
+    runPatterns(text, found, seen, false);
+    // 2) OCR-tolerant recovery pass
+    extractOcrTolerant(text, found, seen);
+    var rank = { high: 0, medium: 1, low: 2 };
+    found.sort(function (a, b) { return rank[a.confidence] - rank[b.confidence]; });
+    return found;
+  }
+
+  // Shipping-label structure: the tracking number lives on or just under
+  // lines like "UPS GROUND" / "TRACKING #:". Returns hot-zone chunks with the
+  // carrier named on/near the anchor (null when the anchor is generic).
+  function anchorCarrier(line) {
+    var l = String(line).toLowerCase();
+    if (/\bups\b/.test(l)) return 'ups';
+    if (/\bfed\s?ex\b/.test(l)) return 'fedex';
+    if (/\busps\b/.test(l) || /priority mail/.test(l)) return 'usps';
+    if (/\bdhl\b/.test(l)) return 'dhl';
+    return null;
+  }
+  function isAnchorLine(line) {
+    if (/track/.test(String(line).toLowerCase())) return true;
+    return anchorCarrier(line) !== null;
+  }
+  function anchorZones(text) {
+    var lines = String(text).split(/\r?\n/);
+    var zones = [];
+    for (var i = 0; i < lines.length; i++) {
+      if (!isAnchorLine(lines[i])) continue;
+      // anchor line + the two lines under it (the number is printed below)
+      var slice = lines.slice(i, Math.min(lines.length, i + 3));
+      var carrier = null;
+      slice.forEach(function (sl) { carrier = carrier || anchorCarrier(sl); });
+      zones.push({ text: slice.join('\n'), carrier: carrier });
+    }
+    return zones;
+  }
+
+  function runPatterns(text, found, seen, boostCarrier) {
     PATTERNS.forEach(function (p) {
       p.re.lastIndex = 0;
       var m;
@@ -67,11 +142,13 @@
         var num = normalize(m[1]);
         if (seen[num]) continue;
         seen[num] = true;
-        // Context boost: carrier word near the number
+        // Context boost: carrier word near the number, or a hot-zone anchor
+        // naming this carrier (a label's "UPS GROUND" line means the number
+        // under it really is UPS).
         var idx = m.index;
         var window = text.slice(Math.max(0, idx - 60), idx + 60 + m[1].length);
         var confidence = p.confidence;
-        var boosted = false;
+        var boosted = !!(boostCarrier && p.carrier === boostCarrier);
         CARRIER_WORDS.forEach(function (cw) {
           if (cw.carrier === p.carrier && cw.re.test(window)) boosted = true;
         });
@@ -80,11 +157,6 @@
         found.push({ number: num, carrier: p.carrier, confidence: confidence });
       }
     });
-    // Sort: high confidence first, then by position
-    extractOcrTolerant(text, found, seen);
-    var rank = { high: 0, medium: 1, low: 2 };
-    found.sort(function (a, b) { return rank[a.confidence] - rank[b.confidence]; });
-    return found;
   }
 
   function isValidFormat(number, carrier) {
@@ -111,9 +183,6 @@
   }
 
   function extractOcrTolerant(text, found, seen) {
-    // Strict UPS structure: 1Z + 6-char shipper number + 2-digit service + 8-digit serial.
-    // Used to validate "1Z"-as-"12" OCR repairs so we don't invent false positives.
-    var UPS_STRICT = /^1Z[0-9A-Z]{6}[0-9]{2}[0-9]{8}$/;
     function push(num, carrier, confidence, extra) {
       num = normalize(num);
       if (!num || seen[num]) return;
@@ -122,28 +191,9 @@
       if (extra) for (var k in extra) c[k] = extra[k];
       found.push(c);
     }
-    function repairUpsZ(ups) {
-      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
-      // only when the rest has strict UPS structure AND the shipper segment
-      // contains a letter — real UPS shipper numbers do, random digit runs
-      // don't. Never invent numbers from thin air.
-      var cand = null;
-      if (/^12[0-9A-Z]{16}$/.test(ups)) cand = '1Z' + ups.slice(2);
-      else if (/^IZ[0-9A-Z]{16}$/i.test(ups)) cand = '1Z' + ups.slice(2);
-      if (!cand || !UPS_STRICT.test(cand)) return null;
-      if (!/[A-Z]/.test(cand.slice(2, 8))) return null;
-      return cand;
-    }
-    // a) UPS number split by spaces/dashes: "1Z 9999 99 ..." (or "12 ..."/"IZ ..." misreads)
-    var m, spacedUps = /\b(?:1Z|12|IZ)[\s-]?(?:[0-9A-Z][\s-]?){16}\b/gi;
-    while ((m = spacedUps.exec(text)) !== null) {
-      var ups = normalize(m[0]);
-      if (/^1Z[0-9A-Z]{16}$/.test(ups)) push(ups, 'ups', 'medium');
-      else {
-        var fixed = repairUpsZ(ups);
-        if (fixed) push(fixed, 'ups', 'medium');
-      }
-    }
+    // a) UPS number split by spaces/dashes (whole-text fallback; the anchor
+    // pass already tried the hot zones at high confidence)
+    spacedUpsIn(text, found, seen, 'medium');
     // b) long digit-like runs with OCR confusions -> fix -> re-match digit formats
     var runRe = /[0-9OILSBoilszb]{10,24}/g, run;
     while ((run = runRe.exec(text)) !== null) {