⚡ ResumeTailor — ChatGPT Edition

Tailor your resume to any job description using ChatGPT Plus you already pay for — no API keys, no extra cost.

Most resume tools rewrite your resume from scratch — destroying your template, changing your formatting, and bloating or shrinking the page count.

ResumeTailor does something different. It treats your resume as a fixed grid and only swaps out the weakest bullets for targeted replacements. Everything else stays exactly the same.

Your Resume (DOCX)  +  Job Description
          ↓
    AI scores every bullet 0–10
          ↓
    Weakest bullets identified
          ↓
    Targeted replacements generated
          ↓
Same template · Same page count · Better targeting

You use ChatGPT for the AI step — the tool generates the perfect prompt, you paste it into chat.openai.com, paste the response back, and the tool handles the rest.

No API billing needed. The tool acts as a bridge between your resume data and ChatGPT:

1. Upload your DOCX resume
2. Paste the job description
3. Tool generates a scoring prompt → you copy it
4. Paste into ChatGPT → copy the JSON response back
5. Tool generates an improvement prompt → repeat
6. Review every proposed swap — approve, reject, or edit inline
7. Download your tailored DOCX

Two prompts to ChatGPT. Total time: ~3–5 minutes.

• ✅ Your template and formatting
• ✅ Page count (validated before download)
• ✅ Bullet count per section
• ✅ Job titles, dates, company names, degrees
• ✅ Bold formatting on skill names

• The text of your weakest bullets — replaced with targeted content grounded in your actual experience

git clone https://github.com/PSR94/resume-tailor.git
cd resume-tailor

cd server
python3 -m venv .venv
source .venv/bin/activate        # Windows: .venv\Scripts\activate
pip install -r requirements.txt

cd ..
npm install

You need two terminals — one for the server, one for the web app.

Terminal 1 — Start the DOCX server:

cd server
source .venv/bin/activate
python main.py

Server starts on http://localhost:7842

Terminal 2 — Start the web app:

npm run dev

App opens at http://localhost:3001

Or use the one-command script from the root:

./start.sh

Use this flow when changing the project locally:

1. Start the app for manual testing:

   ./start.sh

2. Open http://localhost:3001 and test the resume tailoring workflow.
3. Before sharing changes, run the full verification suite:

   npm run verify

./start.sh is for running the local server and frontend together. npm run verify is for tests, build, and backend compile checks.

Upload your .docx resume template. Your file is processed locally — it never leaves your machine.

Copy the full job description from LinkedIn, Greenhouse, Lever, or any job board. The more detail you include, the better the targeting.

The tool generates a prompt. Copy it → open chat.openai.com → paste → send. ChatGPT returns a JSON object scoring every bullet in your resume 0–10 for relevance to this role.

Paste ChatGPT's response back. The tool identifies your weakest bullets and generates a second prompt. Repeat the copy-paste with ChatGPT.

For any skill the AI wants to add that isn't already in your resume, you get a confirmation dialog. Only approve what you can genuinely defend in an interview.

See exactly what changes before anything is applied. Toggle off swaps you don't want. Edit any generated bullet inline. When you're happy, click Apply & Download — your tailored DOCX downloads instantly.

resume-tailor/
├── server/                    # Python FastAPI — DOCX processing
│   ├── main.py                # API server (localhost:7842)
│   ├── docx_engine.py         # DOCX parsing, swap engine, bold preservation
│   └── requirements.txt
│
├── src/                       # React web app — ChatGPT bridge
│   ├── pages/
│   │   ├── Landing.tsx        # Landing page
│   │   └── App.tsx            # Full tailoring tool (all steps)
│   ├── lib/
│   │   ├── prompts.ts         # ChatGPT prompt templates
│   │   └── storage.ts         # Session persistence (localStorage)
│   └── styles/
│       ├── global.css         # CSS variables, resets
│       ├── landing.css        # Landing page styles
│       └── app.css            # App tool styles
│
├── index.html
├── package.json               # Frontend scripts (dev/build/preview)
├── vite.config.ts             # Vite config (localhost:3001)
│
└── start.sh                   # One-command startup script

The local Python server handles all DOCX operations. It runs entirely on your machine.

The AI (ChatGPT) never receives your DOCX binary — only the extracted plain text.

When a new bullet is inserted, the engine:

1. Scans every bold run in your existing resume to build a bold vocabulary (e.g. Python, AWS, Docker)
2. Splits the new bullet text into segments
3. Bolds any segment matching the vocabulary
4. Falls back to bolding CamelCase and ALL-CAPS tech terms not already in the vocabulary

Result: if your resume had **Python** and **AWS** bolded, new bullets mentioning those terms will automatically bold them.

• Your resume is stored in your browser's localStorage — never sent to any server except localhost
• The DOCX binary is processed locally by the Python server — never uploaded to any cloud
• ChatGPT receives only plain text (resume text + job description) — no file upload
• No analytics, no tracking, no accounts

Run all checks

npm run verify

Server not running (red indicator in the app)

cd server && source .venv/bin/activate && python main.py

Run backend tests

cd server
pytest

ChatGPT returns text instead of JSON Add this to the end of the prompt before sending:

"Return only the JSON object with no explanation, no markdown, no code fences."

Formatting changed after download The server validates bullet count before and after every swap. If the count changes, the swap is rejected. If you see formatting issues, check if your DOCX uses tables for the experience section — these need special handling.

Page goes back to start after closing popup
Session state is saved to localStorage. Reopening the tab restores exactly where you were.

This is a personal tool built for solo use. If you fork it and make improvements — especially to the DOCX engine's formatting preservation — PRs are welcome.

MIT — use it, fork it, improve it.