---

🚪 Why Flash Resume

Stop the manual 3-tab shuffle — Job Board ↔ ChatGPT ↔ Overleaf. Full-document AI rewrites take 15+ seconds, hallucinate experience, and push single-page resumes onto page 2.

Applying to jobs that way means juggling three tabs — job board, ChatGPT, and a LaTeX/Overleaf editor — then praying the result still fits on one page. When it doesn't, you're staring at an instant auto-reject.

Flash Resume kills that loop. It surgically injects the job's keywords into bullets you already have, compiles locally at sub-100ms, and verifies the single-page fit before you upload. What used to take a coffee break now takes one tap.

---

✨ What it does

🎯 Targeted ATS diffing	Injects high-density keywords from the job description directly into your existing skills and bullets — under a strict ±2-word budget so nothing overflows.
⚡ Near-instant or top-quality	Pick Groq for a ~3s tailor, or Gemini for the absolute best ATS keyword quality (~5-9s). Switch anytime.
📄 Verified 1-page PDF	Local Typst compilation in <100ms, then pypdf inspects the output to guarantee a true single-page fit — no page-2 surprises.
🧩 1-click browser extension	A blue ⚡ button floats on LinkedIn, Indeed, and Greenhouse. One tap → PDF saved to your folder. No download dialog, no copy-paste.
🔒 100% truthful	The engine only rephrases experience you actually have. It never fabricates companies, degrees, metrics, or domains.

---

🧭 How it works

┌─────────────┐   mirror keyword   ┌──────────────┐   compile    ┌──────────────────┐   verify    ┌───────────────┐
│ Job post    │ ───────────────▶  │  Your master │ ──────────▶   │   Typst engine   │ ─────────▶  │ 1-page PDF +  │
│  (JD text)  │   diff against    │    resume    │  (<0.1s)      │  (bundled, local) │  (pypdf)    │   review diff  │
└─────────────┘                   └──────────────┘               └──────────────────┘              └───────────────┘
        ▲                                                                                                  ▲
        │                                                          AI keyword plan (Gemini / Groq)          │
        │                                                                                                  │
   clipboard / one-tap FAB / pasted text                                                            saved to your folder

The flow:

1. You get the JD — copy it to your clipboard, or just tap the ⚡ button on a job page.
2. Flash Resume reads it — auto-infers the target company and role.
3. The AI maps keywords — matches JD terms against your master resume, the LLM (Gemini or Groq) selects high-impact ATS keywords that fit your real experience.
4. It rewrites with discipline — surgically edits only relevant bullets/skills, keeping every rewrite within ±2 words and 100% truthful.
5. It compiles & verifies — builds the PDF locally in <100ms and checks the output is exactly one page.
6. You get everything — Company_Role.pdf, a structured .json, and a .diff.md review report, plus a rich terminal diff showing exactly what changed.

---

🏗️ Architecture

High level — the extension is a thin client; all intelligence and configuration live in the local engine:

flowchart LR
    subgraph Browser["Your browser (Chrome / Brave / Edge)"]
        FAB["⚡ Floating button<br/>content.js"] -->|"one tap"| EXT["Extension<br/>popup.html"]
    end

    subgraph Local["Your machine — 127.0.0.1:13450"]
        API["FastAPI companion daemon<br/>(fs serve)"] --> TAILOR["TailorEngine<br/>pipeline"]
        TAILOR --> LLM{"AI provider"}
        LLM -->|quality| GEM["Gemini Flash<br/>~5–9s"]
        LLM -->|speed| GROQ["Groq gpt-oss-20b<br/>~3s"]
        TAILOR --> TYPST["Typst compiler<br/><0.1s"] --> PYPDF["pypdf<br/>1-page verification"]
        TYPST --> PDF["PDF + JSON + diff"]
    end

    EXT -->|fetch /api/tailor| API
    FAB -->|fetch /api/tailor| API
    PDF --> SAVE[("Your save folder")]

    subgraph Config["Configuration & assets"]
        CFG["config.json<br/>provider · key · save-folder"]
        RES["Master resume"]
        CFG -.->|read| API
        RES -.->|read| TAILOR
    end

Key design decision: your API key and master resume live only on the local server. The extension never sees them — it just POSTs the job description to 127.0.0.1:13450. Nothing leaves your machine except the JD you choose to send to your AI provider.

Project layout:

flash-resume/
├── pyproject.toml               # Dependencies & fs scripts
├── templates/
│   └── resume.typ               # Modern, ATS-optimized Typst template
├── extension/                   # Manifest V3 Chrome extension
│   ├── manifest.json
│   ├── popup.html / js          # Tailor UI + saved-path view
│   ├── content.js               # Job extractor + one-tap ⚡ floating button
│   └── icons/                   # "fr" toolbar icons (16/48/128)
├── src/flash_resume/
│   ├── cli.py                   # fs CLI (setup, init, tailor, serve, …)
│   ├── config.py                # ~/.config/flash-resume/config.json
│   ├── bootstrap.py             # Silent server entrypoint (pythonw -m …)
│   ├── autostart.py             # Windows Run-key at-login registration
│   ├── extension.py             # Bundled extension installer
│   ├── models/                  # resume.py · job.py · tailoring.py
│   ├── services/
│   │   ├── compiler.py          # Typst + pypdf page validation
│   │   ├── llm.py               # Gemini Flash provider
│   │   ├── groq_llm.py          # Groq gpt-oss-20b fast provider
│   │   ├── tailor.py            # Pipeline orchestrator & diff generator
│   │   └── server.py            # FastAPI companion daemon
│   └── utils/diff.py            # Rich terminal visualization
├── examples/                    # Sample master resume + JD
└── tests/                       # test_models · test_compiler · test_tailor

---

🚀 Quick Start (Windows)

1 · Install

pip install flash-resume

That's it — the package bundles everything (Typst engine, AI clients, extension files). No extra dependencies.

2 · One-time setup

fs setup

A short interactive wizard that walks you through:

1. AI provider — Gemini (quality) or Groq (speed), plus your free API key
2. Save folder — where tailored PDFs land
3. Master resume — loads your real resume (runs the fs init wizard if you haven't yet)
4. Autostart — registers the engine to start silently at login, no terminal ever needed again
5. Extension — copies the extension to a stable folder and prints where to load it

3 · Load the extension (once)

1. Open chrome://extensions/ (works in Chrome, Brave, Edge)
2. Enable Developer mode (top-right toggle)
3. Click Load unpacked and select the folder fs setup printed (default: %LOCALAPPDATA%\flash-resume\extension)

That's it. The engine autostarts at login and the extension reconnects automatically — no reloads, no terminal, no "keep this window open."

---

🧑‍💻 Daily usage

The one-tap way (fastest)

Open any job on LinkedIn, Indeed, or Greenhouse and tap the ⚡ floating button in the bottom-right corner. Flash Resume auto-detects the job, tailors, and shows a toast with your ATS score and saved path.

The detailed way (review / retry)

Open the popup and hit "⚡ Tailor 1-Page Resume" — company, role, and JD are auto-filled (including anything you've text-selected on the page).

The terminal way (everywhere else)

Copy the job description (Ctrl+C), then:

fs tailor

Flash Resume reads the clipboard, infers the company/role, tailors, and drops three files in your save folder:

File	What it is
Company_Role.pdf	✅ Ready to upload
Company_Role.json	Structured tailored resume
Company_Role.diff.md	Reviewable before/after report

…plus a rich terminal diff:

✓ Flash Resume Tailoring Complete
Datadog — Backend Software Engineer
ATS Match Score: 94%  |  Pages: 1 page (Layout Verified)  |  Compile: 72.5ms  |  Total: 1.84s

Integrated ATS Keywords:  FastAPI    PostgreSQL    Docker    Redis

Handy overrides

# Tailor from a file or string instead of the clipboard
fs tailor --jd ./jobs/datadog_swe.txt

# Paste a long JD safely (quotes & apostrophes included)
Get-Content ./jobs/joveo.txt | fs tailor --jd -
cat ./jobs/joveo.txt | fs tailor --jd -          # Git Bash

# Override company / role
fs tailor --company "Google" --role "Site Reliability Engineer"

# Custom output folder
fs tailor --out ./applications/2026/

---

🛠️ CLI reference

Command	What it does
fs setup	Interactive one-time configuration (provider, key, save folder, resume, autostart, extension)
fs init	Create/edit your master resume
fs import	Bring in an existing resume
fs tailor	Read JD → tailor → verified 1-page PDF (clipboard by default)
fs serve	Run the local engine on 127.0.0.1:13450 (the extension needs it)
fs doctor	Check that everything is healthy
fs autostart status	Is the at-login autostart registered?
fs autostart enable / disable	Turn the at-login engine on / off
fs extension-path	Print the folder to load in chrome://extensions
fs --version	Show version

---

🛡️ Product principles

• 100% factual — the engine only maps keywords and rephrases bullets that match your real experience. It never invents companies, degrees, metrics, or domains.
• Layout-preserving budgets — every replacement bullet is constrained to ±2 words of the original, so the layout can't blow up.
• Sub-100ms local compilation — pure local Typst, no external render queues.
• Local-first & private — your key and resume never leave your machine except for the JD you choose to send your AI provider.

---

🧪 Development

git clone https://github.com/PavanBollepalli/flash-resume.git
cd flash-resume
uv sync
uv run fs init
uv run fs doctor
uv run pytest

---

🙌 Contributing

Found a bug, want a new job-board parser, or a macOS/Linux heartbeat? Open an issue or a PR. Contributions of the "it still fits on one page" energy are especially welcome. ⚡

---