Resume Book Generator

Generate a resume book as a single PDF with a terminal-style cover page, table of contents (grouped by graduation date), section title pages between resume groups, and merged resume PDFs. TOC names link to resume pages; section headings (e.g., "Dec 2024") link to their section title pages; LinkedIn and GitHub appear as icons after each name when present. All resume text remains selectable. Uses US Letter paper, GreyHat branding (logo, gray/cyan colors), binary matrix background, and Font Awesome icons.

Requirements

• Python 3.10+
• Typst CLI (typst compile)
• Ghostscript (for PDF merge; optional—falls back to PyMuPDF)
• LibreOffice (for DOCX → PDF conversion; scripts/convert_docx_to_pdf.py)
• Typst package: @preview/fontawesome:0.5.0 (auto-fetched for LinkedIn/GitHub icons)
• assets/grey_hat_hat_small_400x400.png (club logo; copied to output when generating)

Installation

cd resume-book-generator
uv sync
# or: pip install -e .

Usage

1. Place resume PDFs and DOCX files in input-raw/
2. Prepare input/ (copies PDFs, converts DOCX to PDF, skips duplicates):

uv run python scripts/prepare_input.py --raw input-raw --output input

Or convert DOCX only to a separate folder:

uv run python scripts/convert_docx_to_pdf.py --input input-raw --output input-pdf

3. Generate a template CSV:

uv run python scripts/generate_csv_from_resumes.py --input input --output candidates.csv

scripts/generate_csv_from_resumes.py is only a filename/name scaffold. Do not use repo scripts to extract candidate fields from resume text.

4. Fill in candidates.csv by manually reviewing the resumes with AI assistance. The expected workflow is:
   • use the template CSV only for filename and an initial name
   • read each resume and manually confirm name, major, grad_date, email, linkedin, github, phone, and tags
   • keep the reviewed result in candidates.csv For the current GreyHat workflow, the canonical local file is candidates.csv, and output/manual_review/candidates_manual.csv is the preserved manual-review copy.
5. Generate the resume book:

nix-shell -p typst ghostscript libreoffice python313Packages.pandas python313Packages.pymupdf python313Packages.pypdf --run "python scripts/generate_book.py --csv candidates.csv --input input --output output/resume_book.pdf"

The local .venv currently fails to import pymupdf on this machine due to a missing libstdc++.so.6, so prefer the Nix Python command above for final builds.

CSV Schema

Column	Required	Description
filename	yes	PDF filename (e.g. john.pdf)
name	yes	Full name
major	no	Degree/major
grad_date	yes	Graduation (e.g. May 2025)
email	no	Email address
linkedin	no	LinkedIn URL (shown as icon after name)
github	no	GitHub URL (shown as icon after name)
phone	no	Phone number
tags	no	Comma-separated tags (e.g. Leadership, Mentors, CTF Competitor) shown as badges

Candidates are sorted by grad_date then name. Use "Unknown" for unknown graduation dates.

Current local normalization rules:

• Names: title case, never all caps
• Bachelor of Science -> B.S.
• Master of Science in -> M.S. in
• LinkedIn URLs -> https://www.linkedin.com/...
• GitHub URLs -> https://github.com/...
• Phone numbers -> 123-456-7890 or +1 123-456-7890
• If the resume does not clearly state grad_date, keep the old sheet value instead of blanking it
• Preserve sheet tags; they are not resume-derived

Tag sections config

Tags that get a dedicated TOC section (Leadership, Mentor, CTF Competitor, etc.) are configured in resume_book.json:

{
  "tag_sections": ["Leadership", "Mentor", "CTF Competitor"],
  "hidden_tags": ["Leadership"]
}

• tag_sections: Tags that get a dedicated TOC section. Matching is case-insensitive.
• hidden_tags: Tags to hide from badge display (e.g. Leadership when already in the Leadership section).

Override via CLI: --tag-sections "..." and --hidden-tags "Leadership".

Sync from Google Drive / Google Sheets

1. Install sync dependencies: uv sync --extra sync
2. Sync a Drive folder to input/:

uv run python scripts/sync_drive_to_input.py "https://drive.google.com/drive/folders/FOLDER_ID"

3. Sync a Google Sheet to candidates.csv:

uv run python scripts/sync_sheet_to_csv.py "https://docs.google.com/spreadsheets/d/SHEET_ID"

• Drive folder must be shared with "Anyone with the link" (or use a different sync tool for private folders).
• Google Sheet: public sheets work without credentials; for private sheets, place service account JSON at ~/.config/gspread/service_account.json and share the sheet with the service account's client_email.
• These sync scripts only move files and tabular data. They are not approved for extracting candidate fields from resumes.

Manual Review Artifacts

The current repo state includes local-only review artifacts from a manual resume pass:

• output/manual_review/candidates_manual.csv: merged manual-review candidate data
• output/manual_review/chunks/: per-chunk manual review CSVs
• output/manual_review/sheet_vs_local_diff.txt: current diff between the reverted Google Sheet snapshot and the local candidates.csv

These are local review outputs; they are not applied back to Google Sheets automatically.

Options

• --title: Cover page title (default: "GreyHat Resume Book")
• --subtitle: Cover page subtitle (optional)

How It Works

1. Typst generates cover + TOC (no resume embedding). Names are plain text with optional tag badges and LinkedIn/GitHub icons; section headings link to section title pages.
2. Section pages: A separate Typst file generates one title page per graduation group (e.g., "Dec 2024", "May 2025").
3. Merge: cover_toc.pdf + section title pages + resume PDFs are merged (PyMuPDF, with section pages inserted between groups).
4. Link injection: PyMuPDF adds internal link annotations on TOC pages—each name becomes a clickable link to its resume page.
5. Resume text stays selectable because PDFs are merged, not embedded as images.

The current code also fixes a previous TOC-link bug:

• scripts/generate_book.py now refreshes the actual TOC page count after the second Typst pass.
• src/merge/pdf_merger.py uses the correct 1-based to 0-based page conversion for TOC and section links.

Project Structure

resume-book-generator/
├── assets/                    # Club logo
├── input-raw/                 # Raw resumes (PDF + DOCX)
├── input/                     # Prepared PDFs (from prepare_input)
├── output/                    # cover_toc.typ/pdf, section_pages.typ/pdf, resume_book.pdf
├── docs/
│   └── CHATGPT_RESUME_EXTRACTION_PROMPT.md
├── candidates.csv             # Candidate data
├── src/
│   ├── typst/
│   │   ├── generator.py       # Typst cover + TOC + section page generation
│   │   └── templates/
│   │       ├── cover.typ      # Terminal-style cover template
│   │       └── section_page.typ # Section title page template
│   └── merge/pdf_merger.py    # PDF merge + internal link injection
├── scripts/
│   ├── generate_book.py       # Main CLI
│   ├── prepare_input.py       # Copy PDFs + convert DOCX → input/
│   ├── convert_docx_to_pdf.py  # DOCX to PDF only
│   ├── generate_csv_from_resumes.py # Template CSV from filenames
│   ├── sync_drive_to_input.py # Sync Google Drive folder → input/
│   └── sync_sheet_to_csv.py   # Sync Google Sheet → candidates.csv
├── shell.nix                  # typst, ghostscript, libreoffice
└── pyproject.toml

Data Extraction

Resume-field extraction is intentionally manual AI-assisted review, not repo automation.

• Use docs/CHATGPT_RESUME_EXTRACTION_PROMPT.md as a review prompt.
• Read the actual resumes and confirm each row before treating it as canonical.
• Do not rely on repo scripts to infer major, grad_date, contact links, or tags from resume text.