Takes frontend/npm run lint from 23 errors / 40 warnings to 0 errors /
40 warnings, then removes continue-on-error from the CI lint step so it
gates merges.
Errors fixed outright:
- @typescript-eslint/no-explicit-any (5, ChatView.tsx): dropped
redundant (msg as any) casts — Message already declares files,
workflow, and error with the exact shapes the props expect.
- react/no-unescaped-entities (1, support/page.tsx): "We'll" ->
"We'll".
Targeted disables (17, all react-hooks/set-state-in-effect): every site
is an intentional effect-driven state pattern — SSR/hydration mount
gates (Modal, useSelectedModel), reset-on-prop/identity-change
(CaseLawPanel x3, ChatView chat switch, CitationQuotesHeader,
ChatHistoryContext logout, useFetchDocxBytes), sync fast paths of async
fetch/check effects (CaseLawPanel, MfaLoginGate x2), DOM-measured state
(ChatView scroll button, message-visibility restore), and timed UI
latches (PreResponseWrapper, TRChatPanel, AskInputPopup auto-submit).
Rewriting any of them would change runtime behavior, so each carries a
// eslint-disable-next-line with a one-line reason instead of a fix or
a repo-wide rule change.
CI: lint step is now blocking; comments updated to say the backlog is
at zero.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Adds a Testing section documenting every suite the testing PR series
introduces (unit/integration via vitest, Playwright e2e, offline evals,
gated real-Supabase stack tests), the expectation that changes carry tests
at the lowest layer that catches the regression, and a PR template (ported
from the amal66 fork, Open-Legal-Products/mike#205) whose checklist asks how
the change was verified. Intended as the capstone of the series — the
commands it documents are introduced by the sibling test PRs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Ported from amal66/mike#46 onto current main, extended for this tree:
backend job runs npm ci, tests (--if-present, so it is safe to merge in any
order relative to the test-harness PR) and tsc build; frontend job runs
tests, eslint as an advisory step (main currently carries 23 lint errors —
flip to blocking once burned down), and a production next build with
placeholder NEXT_PUBLIC_* env (verified locally that the build succeeds and
contacts nothing); evals job runs node evals/run.mjs when present, else
skips.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
