phase: 170_edit_bubble_sizing

All verification complete. Final report:

**Phase 170 — final verification pass (all 3 tasks were in `complete/`):**

**Verified implemented:**
- Editor renders inside the bubble: `bubble.replaceChildren(form)`, `display:none` rule deleted, `form.style.minWidth` seeded with pre-edit width, `restoreBubble()` (markdown + image re-prepend) before `closeEditor()` on Cancel/Escape, `closeEditor()` reordered before submit re-render
- Label: `editSaveBtn.textContent = "Send"` (app.js L1241)

**Gates (all 100% green):**
- `uv run pytest` → 4066 passed
- `uv run pytest --cov=app --cov-report=term-missing` → 99% on `app/` (gate >90%)
- `uv run ruff check . && uv run pyright` → clean (0 errors, 0 warnings)
- `uv run pytest tests/e2e/test_edit_bubble_sizing.py -v --no-cov` (in isolation) → 6/6 passed
- Regression set in isolation: `test_edit_and_resend.py` 7✓, `test_retry_answer.py` 4✓, `test_failed_turn_retry.py` 4✓, `test_chat_persistence.py` 4✓

**Completion criteria:** in-bubble editor + width pin ✅ · "Send" label + cancel restore + trim re-ask + "Question too long" driver case ✅ · suite/coverage/lint ✅ · E2E + regression isolation ✅ · phase dir moved to `complete/` ✅ (atomic commit is the harness's to make — changes left in the working tree per protocol)

**Documentation:** README.md reviewed in full — edit section already aligned to phase 170 (in-bubble editor + "Send"), pages table matches actual routes, all 29 `.env.example` vars documented, setup commands accurate; AGENTS.md reviewed — all referenced files/commands exist. Both verified consistent — no changes needed.

**Notable decisions:** none — verification pass only, no code changes required.

**Next pending phase:** `171_admin_settings_layout`

tests/e2e/test_edit_and_resend.py

@@ -26,10 +26,10 @@ The contract under test (phase 159, owner-locked 2026-10-05):
   earlier bubbles are never editable.
 * The Edit opens an INLINE editor in the bubble (the ``.tune-form``
   family — the textarea prefilled with the question,
-  ``maxlength`` = the 4000 cap, Save/Cancel, a ``role=alert`` error
+  ``maxlength`` = the 4000 cap, Send/Cancel, a ``role=alert`` error
   line); Cancel/``Escape`` restore the bubble and return focus to the
   toggle.
-* Save = REPLACE + TRUNCATE + RE-ASK (the phase-49 redo-in-place rule,
+* Send = REPLACE + TRUNCATE + RE-ASK (the phase-49 redo-in-place rule,
   one level up): the edited text replaces the record in place, every
   record after it is dropped (DOM + ``bor.chat.v1`` + the saved chat
   row), and a fresh answer streams from the edited text with
@@ -48,7 +48,7 @@ Test → contract mapping:
 
 1. ``test_edit_last_question_reasks_in_place`` — send, grounded answer
    (ONE old brain bubble), the user bubble carries Edit; the editor
-   prefills the EXACT question; append " in one line" + Save: exactly
+   prefills the EXACT question; append " in one line" + Send: exactly
    ONE user bubble (the edited text), the old answer bubble is gone,
    the new answer streams with the marker, ``bor.chat.v1`` holds
    ``[user(edited), brain]`` with NO duplicate question, and the
@@ -76,7 +76,7 @@ Test → contract mapping:
    this identical settle path, its exact detail string is
    real-endpoint-only, see the module docstring); Edit exists on the
    question BESIDE the failed bubble's Retry; trimming (the trigger
-   stripped) + Save streams the grounded answer (the failed record is
+   stripped) + Send streams the grounded answer (the failed record is
    part of the dropped tail — ``failed: true`` never survives).
 6. ``test_edit_is_inert_while_the_turn_runs`` — the ``write a long
    answer`` trigger (~8 s): NO ``.edit-btn`` while the turn runs
@@ -252,7 +252,7 @@ def _poll_saved_messages(
     phase-55 upserts are fire-and-forget — there is no status line to
     wait on; the ``#send-label`` settle only proves the LAST save
     point fired, not that its PUT landed). The row's title keeps the
-    FIRST question on re-Save (title-keep), so the edit's row is the
+    FIRST question on re-Send (title-keep), so the edit's row is the
     same row, polled by the original title."""
     deadline = time.monotonic() + 15
     last: list[dict] | None = None
@@ -338,7 +338,7 @@ def _save_editor(page: Page, new_text: str) -> None:
 
 
 # --------------------------------------------------------------------------
-# 1 — the flow: send → Edit (prefilled) → trim + Save → the old answer
+# 1 — the flow: send → Edit (prefilled) → trim + Send → the old answer
 #    is gone, the new one streams, the record + the row are truncated
 # --------------------------------------------------------------------------
 
@@ -380,7 +380,7 @@ def test_edit_last_question_reasks_in_place(
     assert MOCK_ANSWER_MARKER in stored["messages"][1]["text"]
 
     # The auto-saved row (polled — the title keeps the FIRST question
-    # on re-Save): the truncated shape [user(edited), brain], matching
+    # on re-Send): the truncated shape [user(edited), brain], matching
     # the localStorage shape key-for-key (the DB is the durable boundary).
     row = _poll_saved_messages(
         Q, lambda msgs: len(msgs) == 2 and msgs[0].get("text") == Q_TEST1_EDITED
@@ -507,7 +507,7 @@ def test_edit_drops_a_stale_tail_and_survives_refresh(
 
 # --------------------------------------------------------------------------
 # 5 — the driver case (TODO.md L3): a failed turn's question is
-#    editable; trimming + Save re-asks successfully (the phase-114
+#    editable; trimming + Send re-asks successfully (the phase-114
 #    "Question too long" frame takes this identical settle path — its
 #    exact detail string is real-endpoint-only, see the module
 #    docstring)