fix: import list-marker stripping eats leading numbers in resume content

[Mubashirrrr]

Problem

When importing a resume via the AI / PDF path, list-style fields (experience details, project descriptions, education descriptions, skills) are corrupted whenever a line legitimately starts with a number. Examples of what the import produces today:

"3.5 GPA"                     -> "GPA"
"5 years of experience"       -> "years of experience"
"10x improvement in latency"  -> "x improvement in latency"
"1000+ users served"          -> "+ users served"
"2019 - present"              -> "present"

This is silent data loss: the user uploads a resume and the imported copy is missing leading numbers that are very common in resume bullets (GPAs, years of experience, metrics, dates).

Root cause

toStringArray() in src/app/app/dashboard/resumes/utils.ts splits a multi-line string into items and strips a leading list marker from each line with:

line.replace(/^[-*•\d.)\s]+/, "")

\d, . and ) are inside the same character class with a + quantifier, so the regex greedily removes any run of leading digits / dots / parens — not just an actual list marker. Any line beginning with a number loses that number.

This runs on the real import path: handlePdfFileChange -> /api/resume-import -> createResumeFromAIResult -> toListHtml -> toStringArray (the AI is also prompted to return description/details as text, which toStringArray handles via its string branch).

Fix

Replace the over-greedy class with a regex that strips only genuine list markers:

const LIST_MARKER_PREFIX_REGEX = /^\s*(?:[-*•]+\s*|\d+[.)]\s+)/;

• bullets (-, *, •, optionally repeated), or
• an ordered marker (1., 2)) that is digits followed by ./) and whitespace.

Bare leading numbers are preserved, while "- item", "1. item" and "2) item" are still stripped exactly as before. The only intentional behavior change is that a no-space token like "1.Built" is now kept (it is ambiguous and far more likely real content than a marker).

Test

The repo had no test runner, so this PR adds a minimal vitest setup:

• vitest devDependency + "test": "vitest run" script
• vitest.config.ts reusing the existing vite-tsconfig-paths for @/ alias resolution (no other plugins, so it stays fast and isolated)
• src/app/app/dashboard/resumes/utils.test.ts covering the regression plus guards that real markers are still stripped and the array/empty branches behave.

pnpm test
 ✓ src/app/app/dashboard/resumes/utils.test.ts (4 tests)
 Test Files  1 passed (1)
      Tests  4 passed (4)

Verified fail-before / pass-after: with the original regex the regression test fails with ["GPA", "years of experience with Python", "x improvement in latency", "+ users served", "present"]; with the fix all 4 tests pass. The 3 guard tests pass in both states, confirming marker-stripping is unchanged.

Repro

1. Import a resume (PDF/AI) whose bullets start with numbers, e.g. 5 years of experience, 3.5 GPA, 10x faster.
2. Open the imported resume — the leading numbers are gone.

Or directly: toStringArray("3.5 GPA\n5 years of experience") returns ["GPA", "years of experience"] before the fix and ["3.5 GPA", "5 years of experience"] after.

Note: pnpm-lock.yaml changes are purely additive (vitest and its transitive deps); no existing dependency versions were modified.