PDF Renamer

Context and problem

Students' digital records arrive as scanned PDFs with names like scan_0412.pdf. For each one the routine was the same: open it, read it, figure out the type, type the standard name, pick the student's folder and save. A class has about 200 documents, so organizing a whole class took 3 days.

Time wasn't the only problem. Hand-typed names drift ("Ficha medica", "FICHA MÉDICA", "ficha_saude"), and a file saved in the wrong folder disappears from the student's record.

My role

Sole author: I spotted the bottleneck while working in the office myself, designed the flow, built the tool and followed its adoption. The team uses it every day.

Constraints

• No software installs: it had to run in the browser on the school's PCs.
• Local files: records live in Windows folders, not on a server.
• Scanned documents: many PDFs have no searchable text, just the image.
• Non-developer users: document types and rules vary by department, so adjusting them couldn't depend on me.

Technical decisions and trade-offs

Architecture

Implementation highlights

A classifier you can explain

The text is normalized (accents stripped, upper case) and each document type adds up the weights of the keywords it finds. Short terms must match a whole word, so "RG" (Brazil's ID card) isn't found inside another word. The winning type is only suggested above the threshold.

function scoreKeyword(text, kw) {
  // termos curtos exigem palavra inteira (evita "RG" casar dentro de outra palavra)
  if (kw.length <= 3) return new RegExp("\\b" + kw + "\\b").test(text);
  return text.includes(kw);
}
 
function classify(text) {
  const t = normalizeText(text);
  let best = null,
    bestScore = 0;
  for (const rule of CLASSIFY_RULES) {
    let score = 0;
    for (const [kw, w] of rule.kw) if (scoreKeyword(t, kw)) score += w;
    if (score > bestScore) {
      bestScore = score;
      best = rule.type;
    }
  }
  return bestScore >= CLASSIFY_THRESHOLD ? { type: best, score: bestScore } : null;
}

The built-in rules live alongside the ones the team creates in the UI: a department adds a keyword with a weight and it takes effect immediately, without me touching the code.

Built for daily use

• Duplicates: if the name already exists in the folder, the tool asks what to do and suggests a free numbered name.
• Keyboard: shortcuts to confirm, skip and move on, so hands never leave the keyboard mid-batch.
• Exportable CSV history of everything that was renamed.

Results and impact

• 3× more documents a day: from ~100 to ~300.
• A class of ~200 documents went from 3 days to 1 at most.
• Standard names and files in the right folder, which reduced filing errors across 545 students' records.

What I'd do differently and next steps

Links

• Code on GitHub
• Live demo