feat(frontend): follow-up parser hardening, unit tests, and chip affordance (FE-001)

[avsolatorio]

Summary

Closes #63

Implements FE-001 (follow-up parser hardening, unit tests, and clearer suggestion chips). Suggested follow-ups are easier to parse when the model adds blank lines or markdown heading prefixes, and the UI makes chips read more clearly as clickable actions.

Changes

• parseFollowUps (frontend/lib/parse-follow-ups.ts)

  • Documents the expected heading and list formats (for alignment with backend/prompt work, e.g. BE-001).
  • Tolerates optional ### -style headings and the existing bold/colon variants.
  • Skips blank lines after the heading and between list items instead of stopping at the first empty line, so multi-line spacing from the model does not drop later bullets.

• Unit tests (frontend/lib/__tests__/parse-follow-ups.test.ts)

  • Covers missing heading, bullets, heading variants, * / • / numbered lists, blank lines, the four-item cap, and intro text before the first bullet.

• Suggestion chips (frontend/components/message.tsx)

  • Keeps outline Button-style chips; adds a small message icon (aria-hidden) so the affordance matches “suggested reply” without changing the visible label text.

• Scripts (frontend/package.json)

  • Adds pnpm test:unit using Node’s test runner via tsx.

• Tasks

  • Marks FE-001 as done and refreshes TODO/README.md.

How to test

cd frontend && pnpm test:unit

Manually: complete an assistant reply that includes a Suggested follow-ups: block with blank lines between items; confirm all bullets show as chips and behave as before (send / populate input per existing behavior).

Related

• Task: TODO/FE-001-follow-ups-parser-and-chips.md
• Coordinate heading wording with BE-001 if the model output format is tightened on the backend.