hh-shipments 0.3.3 → 0.3.5

package/package.json

@@ -18,7 +18,7 @@
     "test": "node test/run-tests.js && node test/run-carriers-tests.js"
   },
   "type": "commonjs",
-  "version": "0.3.3",
+  "version": "0.3.5",
   "bin": {
     "hh-shipments": "server.js"
   },

package/public/app.js

@@ -134,7 +134,15 @@ async function runQuick(text, targetEl, preset) {
       };
     });
     loadDashCounts();
-  } catch (e) { targetEl.innerHTML = `<p class="error">${esc(e.message)}</p>`; }
+    return r;
+  } catch (e) { targetEl.innerHTML = `<p class="error">${esc(e.message)}</p>`; return null; }
+}
+
+// When OCR finds no tracking number, show the raw read so the user can
+// spot the misread and type it in manually.
+function showOcrRead(targetEl, text) {
+  const t = (text || '').trim();
+  targetEl.innerHTML += `<details class="ocr-read"><summary>What I read from the picture</summary><pre>${esc(t || '(nothing readable)')}</pre></details>`;
 }
 async function loadDashCounts() { /* counts refresh on next dashboard visit */ }
 
@@ -168,7 +176,8 @@ async function handlePhotoFile(f) {
     fd.append('image', f);
     const up = await fetch('/api/intake/image', { method: 'POST', body: fd });
     if (!up.ok) throw new Error('upload failed');
-    await runQuick(text, $('photo-result'));
+    const r = await runQuick(text, $('photo-result'));
+    if (r && !r.added.length && !r.needs_review.length && !r.duplicates.length) showOcrRead($('photo-result'), text);
     loadPending();
   } catch (err) { $('ocr-status').textContent = ''; $('photo-result').innerHTML = `<p class="error">${esc(err.message)}</p>`; }
 }
@@ -203,7 +212,8 @@ async function handlePastedImageFile(f) {
     fd.append('image', f);
     const up = await fetch('/api/intake/image', { method: 'POST', body: fd });
     if (!up.ok) throw new Error('upload failed');
-    await runQuick(text, $('quick-result'));
+    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>`; }
 }
 

package/public/extract.js

@@ -81,6 +81,7 @@
       }
     });
     // 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;
@@ -95,6 +96,85 @@
     return false;
   }
 
+  function guessCarrier(num) {
+    var carriers = ['ups', 'usps', 'fedex', 'dhl'];
+    for (var i = 0; i < carriers.length; i++) if (isValidFormat(num, carriers[i])) return carriers[i];
+    return 'unknown';
+  }
+
+  // ---- OCR-tolerant pass ----
+  // Screenshots often mangle characters (0/O, 1/I/l, 5/S, 8/B) or split a
+  // number across spaces. Recover what the strict patterns miss.
+  var OCR_DIGIT_FIX = { O: '0', o: '0', I: '1', i: '1', l: '1', L: '1', S: '5', s: '5', B: '8', b: '8' };
+  function ocrDigitFix(token) {
+    return token.replace(/[OoIiLlSsBb]/g, function (c) { return OCR_DIGIT_FIX[c]; });
+  }
+
+  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}$/;
+    function push(num, carrier, confidence, extra) {
+      num = normalize(num);
+      if (!num || seen[num]) return;
+      seen[num] = true;
+      var c = { number: num, carrier: carrier, confidence: confidence, ocr_fixed: true };
+      if (extra) for (var k in extra) c[k] = extra[k];
+      found.push(c);
+    }
+    function repairUpsZ(ups) {
+      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+      // only when the rest has strict UPS structure AND the shipper segment
+      // contains a letter — real UPS shipper numbers do, random digit runs
+      // don't. Never invent numbers from thin air.
+      var cand = null;
+      if (/^12[0-9A-Z]{16}$/.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');
+      }
+    }
+    // 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) {
+      var fixed = normalize(ocrDigitFix(run[0]));
+      if (/^9[2-5]\d{18,20}$/.test(fixed)) push(fixed, 'usps', 'medium');
+      else if (/^96\d{20}$/.test(fixed)) push(fixed, 'fedex', 'medium');
+      else if (/^(\d{12}|\d{15}|\d{20})$/.test(fixed)) push(fixed, 'fedex', 'medium');
+      else if (/^\d{11}$/.test(fixed)) push(fixed, 'dhl', 'medium');
+      else if (/^\d{10}$/.test(fixed)) push(fixed, 'dhl', 'low');
+    }
+    // c) fallback: long alphanumeric tokens (must contain a digit) that matched
+    // nothing — surface the most plausible few as "verify the digits" candidates
+    var tokens = text.match(/\b(?=[A-Z0-9]*\d)[A-Z0-9]{12,30}\b/gi) || [];
+    tokens = tokens.map(normalize);
+    tokens = tokens.filter(function (t, i) {
+      if (seen[t] || seen[ocrDigitFix(t)]) return false; // already covered (possibly OCR-fixed)
+      return tokens.indexOf(t) === i;
+    });
+    tokens.sort(function (a, b) { return b.length - a.length; });
+    tokens.slice(0, 3).forEach(function (t) {
+      // UPS "1Z" misread as "12"/"IZ" in a joined (space-free) token
+      var fixed = repairUpsZ(t);
+      if (fixed) { push(fixed, 'ups', 'medium'); return; }
+      // OCR-fixed variant may match a carrier format (e.g. I->1 at the start)
+      var digitFixed = ocrDigitFix(t);
+      var gc = guessCarrier(digitFixed);
+      if (gc !== 'unknown') { push(digitFixed, gc, 'low', { possible: true }); return; }
+      push(t, guessCarrier(t), 'low', { possible: true });
+    });
+  }
+
   var api = { extractTrackingNumbers: extractTrackingNumbers, extractContext: extractContext, normalize: normalize, isValidFormat: isValidFormat };
   if (typeof module !== 'undefined' && module.exports) module.exports = api;
   globalThis.HHExtract = api;

package/public/styles.css

@@ -44,6 +44,8 @@ h2 { font-size: 17px; margin: 18px 0 8px; }
 .filebtn { display: block; text-align: center; background: #e5e5ea; border-radius: 12px; padding: 16px; font-weight: 600; cursor: pointer; }
 .pastebox { margin-top: 8px; border: 2px dashed #c7c7cc; border-radius: 12px; padding: 14px; text-align: center; outline: none; caret-color: transparent; min-height: 20px; }
 .pastebox:focus { border-color: #007aff; background: #f2f8ff; }
+.ocr-read { margin-top: 8px; }
+.ocr-read pre { white-space: pre-wrap; font-size: 12px; background: #f2f2f7; padding: 8px; border-radius: 8px; }
 .statusline { font-size: 17px; font-weight: 600; margin: 8px 0; }
 .timeline { border-left: 2px solid #d1d1d6; margin-left: 6px; padding-left: 14px; }
 .timeline .ev { margin-bottom: 12px; font-size: 14px; }

package/src/extract.js

@@ -81,6 +81,7 @@
       }
     });
     // 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;
@@ -95,6 +96,85 @@
     return false;
   }
 
+  function guessCarrier(num) {
+    var carriers = ['ups', 'usps', 'fedex', 'dhl'];
+    for (var i = 0; i < carriers.length; i++) if (isValidFormat(num, carriers[i])) return carriers[i];
+    return 'unknown';
+  }
+
+  // ---- OCR-tolerant pass ----
+  // Screenshots often mangle characters (0/O, 1/I/l, 5/S, 8/B) or split a
+  // number across spaces. Recover what the strict patterns miss.
+  var OCR_DIGIT_FIX = { O: '0', o: '0', I: '1', i: '1', l: '1', L: '1', S: '5', s: '5', B: '8', b: '8' };
+  function ocrDigitFix(token) {
+    return token.replace(/[OoIiLlSsBb]/g, function (c) { return OCR_DIGIT_FIX[c]; });
+  }
+
+  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}$/;
+    function push(num, carrier, confidence, extra) {
+      num = normalize(num);
+      if (!num || seen[num]) return;
+      seen[num] = true;
+      var c = { number: num, carrier: carrier, confidence: confidence, ocr_fixed: true };
+      if (extra) for (var k in extra) c[k] = extra[k];
+      found.push(c);
+    }
+    function repairUpsZ(ups) {
+      // Tesseract often reads the "1Z" anchor as "12" (or "IZ"). Repair it
+      // only when the rest has strict UPS structure AND the shipper segment
+      // contains a letter — real UPS shipper numbers do, random digit runs
+      // don't. Never invent numbers from thin air.
+      var cand = null;
+      if (/^12[0-9A-Z]{16}$/.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');
+      }
+    }
+    // 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) {
+      var fixed = normalize(ocrDigitFix(run[0]));
+      if (/^9[2-5]\d{18,20}$/.test(fixed)) push(fixed, 'usps', 'medium');
+      else if (/^96\d{20}$/.test(fixed)) push(fixed, 'fedex', 'medium');
+      else if (/^(\d{12}|\d{15}|\d{20})$/.test(fixed)) push(fixed, 'fedex', 'medium');
+      else if (/^\d{11}$/.test(fixed)) push(fixed, 'dhl', 'medium');
+      else if (/^\d{10}$/.test(fixed)) push(fixed, 'dhl', 'low');
+    }
+    // c) fallback: long alphanumeric tokens (must contain a digit) that matched
+    // nothing — surface the most plausible few as "verify the digits" candidates
+    var tokens = text.match(/\b(?=[A-Z0-9]*\d)[A-Z0-9]{12,30}\b/gi) || [];
+    tokens = tokens.map(normalize);
+    tokens = tokens.filter(function (t, i) {
+      if (seen[t] || seen[ocrDigitFix(t)]) return false; // already covered (possibly OCR-fixed)
+      return tokens.indexOf(t) === i;
+    });
+    tokens.sort(function (a, b) { return b.length - a.length; });
+    tokens.slice(0, 3).forEach(function (t) {
+      // UPS "1Z" misread as "12"/"IZ" in a joined (space-free) token
+      var fixed = repairUpsZ(t);
+      if (fixed) { push(fixed, 'ups', 'medium'); return; }
+      // OCR-fixed variant may match a carrier format (e.g. I->1 at the start)
+      var digitFixed = ocrDigitFix(t);
+      var gc = guessCarrier(digitFixed);
+      if (gc !== 'unknown') { push(digitFixed, gc, 'low', { possible: true }); return; }
+      push(t, guessCarrier(t), 'low', { possible: true });
+    });
+  }
+
   var api = { extractTrackingNumbers: extractTrackingNumbers, extractContext: extractContext, normalize: normalize, isValidFormat: isValidFormat };
   if (typeof module !== 'undefined' && module.exports) module.exports = api;
   globalThis.HHExtract = api;