What

Follow-up to #672 / PR #749. After PR #749, _searchByTag no longer pollutes search-input, so a tag pill click on a fresh session leaves search-input empty. doSearch() (packages/memtomem/src/memtomem/web/static/app.js:1675) currently early-returns on empty q:

async function doSearch() {
  const q = qs('search-input').value.trim();
  if (!q) return;  // ← blocks tag-only path
  const tf = qs('tag-filter').value.trim();
  ...
}

…and the backend (packages/memtomem/src/memtomem/web/routes/search.py) declares q as required:

async def search(
    q: str = Query(...),          # ← min length enforced; empty q → 422
    tag_filter: str | None = Query(None),
    ...
):

So the click currently acts as a "filter prep" — Search tab opens with tag-filter populated, but no results render until the user types a query.

Why

The intuitive UX (per #672 discussion) is that clicking a tag pill = "show me all memos with this tag". Today that requires an extra keystroke. Lifting the empty-q guard on both layers makes the click work as expected without any further _searchByTag change — PR #749 already calls doSearch() unconditionally, so this fix is purely additive.

Suggested change

1. Frontend (app.js:1673-1722): replace if (!q) return; with if (!q && !tf) return; so the search runs whenever either axis has a value.
2. Backend (packages/memtomem/src/memtomem/web/routes/search.py): make q optional (q: str | None = Query(None)) and validate that at least one of q / tag_filter is present (return 400 otherwise — keeps the "no-op search" guard at the API layer).
3. pipeline.search (packages/memtomem/src/memtomem/search/pipeline.py): verify it tolerates query=None / query="". The pipeline order is fixed (per CLAUDE.md); the BM25/dense stages should short-circuit cleanly when there's no query and the tag/source filter stage takes over as the primary selector. Add a regression test for tag_filter-only retrieval.
4. CHANGELOG: [Unreleased] / Changed — "Tag pill click now runs a tag-only search on a fresh session (was a no-op filter prep after _searchByTag pollutes search-input with tag text #672)."

Out of scope

• Multi-tag selection on tag-filter (separate UX question — issue _searchByTag pollutes search-input with tag text #672 noted the field is currently single-value text).
• Sort order for tag-only results (BM25 won't rank without a query — likely fall back to recency or importance boost; needs a small UX call).

Test plan

• Backend: pytest packages/memtomem/tests/test_user_workflows.py::TestSearch::test_tag_filter_search extended to cover q=None.
• Frontend (manual): Tags tab → click pill on fresh session → Search tab activates and results render immediately, scoped to the tag.

References

• _searchByTag pollutes search-input with tag text #672 — original bug (axis duplication), fixed in PR fix(web): drop search-input pollution from _searchByTag (closes #672) #749
• PR fix(web): drop search-input pollution from _searchByTag (closes #672) #749 — leaves the empty-q UX as filter prep so this issue can land independently