Can I vibecode Resume Worded?

Add job tailoring to the resume checker in this folder. Requirements:

- A page where I paste a job posting; an LLM extracts required and preferred skills,
  tools and responsibilities as Zod-validated JSON.
- Show a coverage table: each item, whether the resume text mentions it, and the line
  where it does.
- Suggest bullet edits only for items the resume already supports. List unsupported items
  as gaps and never write them into the resume.
- Save each posting with its coverage score so I can compare postings.
- Update the README.

Add version history to the resume checker in this folder. Requirements:

- Uploading a resume with the same label creates a new version instead of a new resume.
- A history page with the score per version on a small line chart and a word-level diff
  between any two versions.
- The report shows which rule findings were fixed and which are new since the last version.
- Update the README.

Add an export check to the resume checker in this folder. Requirements:

- After generating the DOCX, convert it to PDF with LibreOffice (`soffice --headless`) and
  extract the text again with `pdfjs-dist`.
- Compare it with the source text and report missing sections, reordered columns, lost
  characters and table cells that merged.
- Show the extracted text next to the original, with differences highlighted.
- Fail clearly if LibreOffice isn't installed. Update the README.

Add a local-model mode to the resume checker in this folder. Requirements:

- LLM_BASE_URL and LLM_MODEL in .env switch between a hosted API and Ollama; no code path
  may call another host.
- An `npm run eval` command that runs the five resumes in ./samples through the current model
  and prints a table of valid JSON, bullets rewritten and invented facts found.
- Detect invented numbers by checking every digit in a rewrite against the source text.
- Update the README with a comparison table and the model I ran.