jakeforge

A Jake's-style resume editor, live at jakeforge.xinyiklin.com. Edit a resume directly on the page — the same single-column, ATS-friendly layout from Jake Gutierrez's LaTeX template — and export it as a clean browser PDF, a Tectonic-compiled LaTeX PDF, or .tex source.

The app is deployed on EC2 behind Caddy (HTTPS), but stays private by design: your resume lives in your browser's localStorage and is sent only to the app's own LaTeX endpoints for rendering — never stored server-side.

It's a focused extraction of the resume editor from the role-fit-ai sibling project: the structured editor, live document, and LaTeX pipeline, with the AI tailoring, job tracker, and applications stripped out.

Features

• On-page editing — name, contact chips, and Education / Experience / Projects / Skills / Summary sections. Add, remove, and drag-reorder sections, entries, and bullets inline (@dnd-kit, pointer + keyboard).
• Faithful Jake's styling — serif document, ruled section headings, bold titles with right-aligned dates, italic subtitles, automatic multi-page pagination that keeps headings with their first entry.
• Layout & spacing controls — page zoom (also ⌘/Ctrl +/- / 0) and Compact/Normal/Relaxed presets. A Fine-tune button opens a spacing flyout with sliders for line height and header/section/entry/list gaps; save the current spacing as a reusable Custom preset. Settings persist to localStorage.
• Typography controls — collapsible Typography / Entries / Skills cards: section-heading case (small caps / uppercase / normal) plus bold and underline, and a configurable contact divider (quick-pick glyphs or custom 1-2 char input) under Typography; bold titles and italic subtitles under Entries; bold skill labels under Skills. One-click reset to Jake's defaults.
• Exports (each download opens a rename dialog pre-filled with a resume-derived file name)
  • PDF - LaTeX — the resume rendered through the Jake's template and compiled by Tectonic. Requires Tectonic.
  • LaTeX source (.tex) — download the rendered template source.
  • Clean PDF — press Cmd+P / Ctrl+P; the print CSS isolates the resume into a selectable, ATS-readable page (choose Save as PDF). No dependencies.
• Import — drag-and-drop or browse for a file (.txt, .md, .tex, .docx) directly from the sidebar. LaTeX is auto-detected. PDF import is intentionally unsupported — it extracts too poorly to be useful.
• In-app PDF preview — compile and view the LaTeX PDF without downloading.
• Autosave — the structured resume is persisted to localStorage, so a reload keeps your work. Rendering only calls the app's own LaTeX endpoints.

Getting started

npm install
npm run dev        # http://localhost:5186

npm run dev runs server.mjs, which serves the Vite app and the LaTeX API from one process.

Optional: LaTeX PDF output

Clean PDF via Cmd+P / Ctrl+P works out of the box. For the high-fidelity PDF - LaTeX export and the in-app preview, install Tectonic:

brew install tectonic

Then restart the dev server. Without it, those buttons are disabled and the sidebar shows the clean-print hint.

Scripts

Command	Description
npm run dev	Dev server (frontend + LaTeX API) on port 5186
npm run build	Typecheck (tsc) + production build to dist/
npm run preview	Serve the production build (NODE_ENV=production)

Deploying (Docker)

The Dockerfile builds a self-contained image: the production bundle, server.mjs, Tectonic (arch-matched, so LaTeX PDF works in the container), and unzip for DOCX import.

docker build -t jakeforge .
docker run -p 5186:5186 -e ALLOWED_HOSTS=resume.example.com,203.0.113.7 jakeforge

ALLOWED_HOSTS is required: a comma-separated list of every hostname or IP the app is reached by. It backs the same-origin/Host guard on the API, and the server refuses to start without it when bound beyond loopback. Loopback names (localhost, 127.0.0.1) are always allowed, so on-box smoke tests and container health checks work regardless.

A single small instance (e.g. an EC2 t3.micro) is plenty — the server is one Node process with no database.

For a small EC2 deployment, bind the container to loopback and put an HTTPS reverse proxy (e.g. Caddy) in front:

docker run -d \
  --name jakeforge \
  --restart unless-stopped \
  -p 127.0.0.1:5186:5186 \
  -e ALLOWED_HOSTS=jakeforge.example.com,203.0.113.7,localhost,127.0.0.1 \
  jakeforge

Caddy handles HTTPS, automatic Let's Encrypt certificates, and HTTP→HTTPS redirects. A minimal Caddyfile:

jakeforge.example.com {
    reverse_proxy 127.0.0.1:5186
}

Run Caddy as a Docker container on the same host:

docker run -d \
  --name caddy \
  --restart unless-stopped \
  --network host \
  -v caddy_data:/data \
  -v /path/to/Caddyfile:/etc/caddy/Caddyfile \
  caddy:2

To smoke-test on the host, hit the loopback binding directly (curl -I http://127.0.0.1:5186). Publishing the container on port 80 (-p 80:5186) only works before Caddy is installed — once Caddy owns 80/443, that mapping fails with "address already in use".

GitHub Actions deploy to EC2

The workflow in .github/workflows/deploy.yml keeps pull requests as CI-only builds. Pushes to main SSH into the EC2 host, copy the source archive, build the Docker image on the instance, and restart the jakeforge container bound to 127.0.0.1:5186. A separately managed Caddy container owns ports 80/443 and reverse-proxies to the app.

Configure these repository secrets before enabling the deploy job:

Secret	Value
EC2_HOST	Public IPv4 address or DNS name of the EC2 instance
EC2_USER	SSH user, typically ec2-user on Amazon Linux
EC2_SSH_KEY	Private key contents for the EC2 key pair
ALLOWED_HOSTS	Comma-separated public hostnames/IPs, e.g. jakeforge.xinyiklin.com,100.60.78.4,ec2-100-60-78-4.compute-1.amazonaws.com

The EC2 instance must already have Docker and Caddy running (see the Caddy setup above). The workflow uses plain docker when available, or passwordless sudo docker on default Amazon Linux setups.

Architecture

src/
  lib/            resume data model + parse/serialize/LaTeX-extract helpers
  hooks/          useResumeEditor · useDocStyle · useTemplates · useResumeExport
  components/     reusable Modal shell and ImportModal
  sections/
    editor/       the editable on-page resume (sections, entries, bullets, skills)
    Resume*.tsx   read-only document + off-screen print layer
  styles/         design tokens, resume document/editor CSS, app shell
public/
  favicon.svg     forge brand mark (anvil app-icon, also the sidebar lockup)
  fonts/          embedded LM Roman faces for PDF-faithful on-page rendering
server/
  latex/          Jake's template renderer, resume parser, Tectonic wrapper
  docx.mjs        DOCX text extractor (zero-dep; shells to unzip)
server.mjs        serves the app + /api/{templates,render-resume-latex,import-resume-tex,import-resume-docx}

The editor holds a structured resume model (ResumeData). LaTeX exports render straight from that model through the template — the same path the preview uses — so downloads match what you see. Plain-text serialization backs the clean-print mirror.

Privacy

Local-first and personal. The resume lives in localStorage and is sent solely to the app's own LaTeX endpoints for rendering — on your machine when running locally, or on your own server when self-hosting the Docker image. There is no account, no database, and no third-party service.