phase: 150_drop_sync_result

**Phase 150 final verification pass — complete, 100% green.** All 6 task files were already in `complete/`, so this was verification + doc pass.

- Verified: `#sync-progress` is the sole sync text surface + sole `aria-live` region (`role="status"`, polite, no `aria-hidden`); `span#sync-result` gone from HTML/JS/CSS/tests (incl. runtime-built `RETIRED_ID` absence pins in the unit suites); `setSyncProgress` single writer with text-driven visibility; running-only box write in `setSyncLabel`; all 9 settle/cancel sites routed; `requestCancel` pre-fetch clear is a NO-OP.
- Tests: `uv run pytest` → **3463 passed**; `uv run pytest --cov=app` → **99% on app/** (>90%); `uv run ruff check .` + `uv run pyright` → clean (0 errors); `tests/e2e/test_sync_progress_box.py --no-cov` → **8 passed in isolation**; all 18 migrated sync/image-PDF E2E files green in isolation (77 tests).
- Completion criteria: 1) grep clean ✅ 2) sole live region, attrs correct ✅ 3) in-flight/settle/cancel/idle box states proven by E2E + unit pins ✅ 4) icon-only running button + restore ✅ 5) full gates green ✅ 6) no completed-phase behavior change (18 family suites + full suite green) ✅ 7) commit — left to harness per pass instructions (working tree untouched, no `git add`/`commit`).
- Documentation: README.md fixed 2 pre-existing inconsistencies (stale `~/Homelab`+`~/Deployments` source model in the intro — retired in phase 61 — now describes the git/local/archive Sources model; "Requires AIPI_KEY" in retrieval-quality now states `BOR_LLM_API_KEY` in `.env` or `$AIPI_KEY`, matching the scripts' real convention). Everything else (URL table, env defaults vs `app/config.py`, sync/phase-150 copy, prod claims) verified accurate. AGENTS.md: verified consistent — no changes needed.
- Notable: no code changes were required — the phase implementation was correct as left; no live infrastructure touched.
- Next pending phase: `151_pdf_image_extraction` in `.agents/phases/todo/`.

tests/e2e/test_sync_changed_only.py

@@ -19,7 +19,8 @@ bump ``indexed_at``); a source change syncs ONLY the changed part (new
 file → ``1 added``; modified file → ``1 updated`` with only its own
 row moving). NO PDF here — the unit matrix (task 02) carries the
 vision-skip proof; this suite pins the WIRE (the owner's visible
-``#sync-result`` line) + the UI settle + the row-level proof, fast and
+``#sync-progress`` result line — phase 150: the box is the single sync
+text surface) + the UI settle + the row-level proof, fast and
 deterministic.
 
 Per-module app env (the E2E conftest pattern, module-scoped): this
@@ -259,11 +260,13 @@ def _click_sync_and_wait_settle(page: Page) -> None:
 
 
 def _expect_settle_line(page: Page, expected: str) -> None:
-    """The owner's surface: the VISIBLE ``#sync-result`` text in its
-    house count shape (sources.js's fmtSyncResult — the zero
+    """The owner's surface: the VISIBLE ``#sync-progress`` box text in
+    its house count shape (sources.js's fmtSyncResult — the zero
     "updated" part is omitted: ``0 added · N unchanged``,
-    ``1 added · N unchanged``, ``0 added · 1 updated · N unchanged``)."""
-    expect(page.locator("#sync-result")).to_have_text(
+    ``1 added · N unchanged``, ``0 added · 1 updated · N unchanged``).
+    Phase 150: the box IS the result line's surface (the retired
+    announcer span is gone)."""
+    expect(page.locator("#sync-progress-text")).to_have_text(
         expected, timeout=SYNC_TIMEOUT_MS
     )