Changelog

v0.2.1 — 2026-05-15

Added

• redactPii option — redacts personally identifiable information from the resume text before sending it to the LLM, then reinjects the real values into the structured output after extraction.

  Redacted categories: email addresses, phone numbers, physical addresses, and URLs. Each is replaced with a deterministic placeholder (__PII_EMAIL_0__, __PII_URL_1__, etc.).

  const result = await parseResume(buffer, {
    model,
    redactPii: true,
  })
  // The LLM sees "__PII_EMAIL_0__" instead of "john.doe@gmail.com"
  // result.data.basics.email is still "john.doe@gmail.com"

  Works with both parseResume() and streamResume().

• confidenceScore in sectionResults — each section now includes a confidence score (0.0–1.0) based on retry count and field completeness.

  for (const s of result.meta.sectionResults ?? []) {
    console.log(`${s.section}: ${s.confidenceScore}`)
    // basics: 0.87  work: 1.00  skills: 1.00
  }

• redactPii, reinjectPii, describePiiRedaction exports — utility functions for custom redaction pipelines.

Fixed

• Zod schema accepts PII placeholders — email and url fields now accept __PII_*__ placeholders during extraction when redactPii is active. Previously, z.string().email() would reject placeholders and cause the basics section to fail after all retries.

---

v0.2.0 — 2026-05-07

Added

• streamResume() function — AsyncGenerator that yields events as each section is extracted.

  for await (const event of streamResume(buffer, { model })) {
    if (event.type === 'section') updateUI(event.section, event.data)
    if (event.type === 'done') setResult(event.result)
  }

• StreamResumeEvent type export

Fixed

• "2025-01" → "2025" (month-only date padding stripped)
• volunteer: [], interests: [] → omitted from output
• Skills with no keywords → filtered out
• Empty strings in highlights arrays → removed

---

v0.1.3 — 2026-05-01

Added

• Output normalizer (date padding, truncated URLs, invalid profiles, "Present" endDate)
• maxConcurrency option
• onProgress callback
• ocrLanguage option
• normalizeDate and cleanUrl exports

Fixed

• Work section missing personal project entries
• Scan detection false positives on skill-heavy CVs

---

v0.1.2-hotfix.1 — 2026-04-30

Fixed

• disableOcr option — prevents Tesseract WASM crash on Vercel/Lambda

---

v0.1.2 — 2026-04-30

Added

• 15 sections, sections option, outputSchema option, CLI, serverless worker fix

---

v0.1.1-beta.1 — 2026-04-28

Added

• Per-section maxTokens, temperature: 0, OCR text cleaning, per-section retry, deduplication, sectionResults

---

v0.1.0 — 2026-04-27

Initial release.