From ecc921098a8c00dceb5403c0e0b67374f4b5bf32 Mon Sep 17 00:00:00 2001
From: ducoterra <git@ducoterra.net>
Date: Sat, 12 Sep 2026 19:45:00 -0400
Subject: [PATCH] phase: 104_chip_sizing_question_cap
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

All completion criteria verified. Final report:

**Phase 104 — final verification pass: all green**
- Verified (no code changes needed): single-line ellipsized `.suggestion-chip` CSS + deleted `.maybe-try` override, `renderChips` full-text `title` + `aria-label`-when-clipped, `maxlength="4000"` + `#char-count` counter + `handleSend` over-cap guard, unit pins file, schemas boundary pins, dedicated E2E suite.
- E2E (isolation): `uv run pytest tests/e2e/test_chip_sizing_question_cap.py -v --no-cov` → **6 passed**; regressions: `test_suggestion_chips.py` 8 passed, `test_pinned_composer.py` 4 passed, `test_responsive_polish.py` 7 passed, `test_chat_history.py` 5 passed.
- `uv run pytest` → **2102 passed**; `--cov=app` → **99%** (>90%); `uv run ruff check . && uv run pyright` → clean, 0 errors.
- Criteria: chip E2E (single-line, clipped, title+aria-label full text) ✅; paste caps at exactly 4,000, send streams, counter hides ✅; programmatic 5,000-char fill → banner, no turn, text kept ✅; 4,000/4,001 boundary pinned + HTML maxlength == JS constant cross-file pin ✅.
- Diff scope: `frontend/`, new unit file, `tests/unit/test_schemas.py`, new E2E file, phase files — **no `app/` diff, no migration, no `shared.js` diff**.
- Deviations: 4 regression test files touched — 2 genuine DOM-pin conflicts from the new `#char-count` child (explicitly anticipated by the overview) + 3 documented **pre-existing E2E flake fixes** (smooth-scroll race, tab-walk heuristic, 10 ms timeout), each verified pre-existing on the pre-phase-104 tree.
- No commit made (harness commits per the execution protocol override).
- Next pending phase: `98_sync_summary_visibility`.
---
 .../104_chip_sizing_question_cap/00_phase.md  |   0
 .../01_chip_ellipsis_css.md                   |   0
 .../02_chip_tooltip_aria.md                   |   0
 .../03_composer_question_cap.md               |   0
 .../04_e2e_suite_commit.md                    |   0
 ..._chip_sizing_question_cap__00_phase.a1.err |   0
 ...4_chip_sizing_question_cap__00_phase.a1.md |  11 +
 ..._sizing_question_cap__00_phase.a1.validate |  97 +++
 ..._question_cap__01_chip_ellipsis_css.a1.err |   0
 ...g_question_cap__01_chip_ellipsis_css.a1.md |  18 +
 ...tion_cap__01_chip_ellipsis_css.a1.validate |  97 +++
 ..._question_cap__02_chip_tooltip_aria.a1.err |   0
 ...g_question_cap__02_chip_tooltip_aria.a1.md |  15 +
 ...tion_cap__02_chip_tooltip_aria.a1.validate |  97 +++
 ...stion_cap__03_composer_question_cap.a1.err |   0
 ...estion_cap__03_composer_question_cap.a1.md |  14 +
 ..._cap__03_composer_question_cap.a1.validate |  97 +++
 ...g_question_cap__04_e2e_suite_commit.a1.err |   0
 ...ng_question_cap__04_e2e_suite_commit.a1.md |  11 +
 ...stion_cap__04_e2e_suite_commit.a1.validate |  97 +++
 frontend/assets/app.js                        |  55 +-
 frontend/assets/styles.css                    |  39 +-
 frontend/index.html                           |  13 +
 tests/e2e/test_chip_sizing_question_cap.py    | 521 +++++++++++++++++
 tests/e2e/test_pinned_composer.py             |  51 +-
 tests/e2e/test_responsive_polish.py           |  34 +-
 tests/e2e/test_suggestion_chips.py            |   7 +-
 tests/unit/test_chip_sizing_question_cap.py   | 550 ++++++++++++++++++
 tests/unit/test_pinned_composer.py            |  33 +-
 tests/unit/test_save_chat_ui.py               |  27 +-
 tests/unit/test_schemas.py                    |  26 +
 31 files changed, 1874 insertions(+), 36 deletions(-)
 rename .agents/phases/{todo => complete}/104_chip_sizing_question_cap/00_phase.md (100%)
 rename .agents/phases/{todo => complete}/104_chip_sizing_question_cap/01_chip_ellipsis_css.md (100%)
 rename .agents/phases/{todo => complete}/104_chip_sizing_question_cap/02_chip_tooltip_aria.md (100%)
 rename .agents/phases/{todo => complete}/104_chip_sizing_question_cap/03_composer_question_cap.md (100%)
 rename .agents/phases/{todo => complete}/104_chip_sizing_question_cap/04_e2e_suite_commit.md (100%)
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.err
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.md
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.validate
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.err
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.md
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.validate
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.err
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.md
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.validate
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.err
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.md
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.validate
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.err
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.md
 create mode 100644 .agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.validate
 create mode 100644 tests/e2e/test_chip_sizing_question_cap.py
 create mode 100644 tests/unit/test_chip_sizing_question_cap.py

diff --git a/.agents/phases/todo/104_chip_sizing_question_cap/00_phase.md b/.agents/phases/complete/104_chip_sizing_question_cap/00_phase.md
similarity index 100%
rename from .agents/phases/todo/104_chip_sizing_question_cap/00_phase.md
rename to .agents/phases/complete/104_chip_sizing_question_cap/00_phase.md
diff --git a/.agents/phases/todo/104_chip_sizing_question_cap/01_chip_ellipsis_css.md b/.agents/phases/complete/104_chip_sizing_question_cap/01_chip_ellipsis_css.md
similarity index 100%
rename from .agents/phases/todo/104_chip_sizing_question_cap/01_chip_ellipsis_css.md
rename to .agents/phases/complete/104_chip_sizing_question_cap/01_chip_ellipsis_css.md
diff --git a/.agents/phases/todo/104_chip_sizing_question_cap/02_chip_tooltip_aria.md b/.agents/phases/complete/104_chip_sizing_question_cap/02_chip_tooltip_aria.md
similarity index 100%
rename from .agents/phases/todo/104_chip_sizing_question_cap/02_chip_tooltip_aria.md
rename to .agents/phases/complete/104_chip_sizing_question_cap/02_chip_tooltip_aria.md
diff --git a/.agents/phases/todo/104_chip_sizing_question_cap/03_composer_question_cap.md b/.agents/phases/complete/104_chip_sizing_question_cap/03_composer_question_cap.md
similarity index 100%
rename from .agents/phases/todo/104_chip_sizing_question_cap/03_composer_question_cap.md
rename to .agents/phases/complete/104_chip_sizing_question_cap/03_composer_question_cap.md
diff --git a/.agents/phases/todo/104_chip_sizing_question_cap/04_e2e_suite_commit.md b/.agents/phases/complete/104_chip_sizing_question_cap/04_e2e_suite_commit.md
similarity index 100%
rename from .agents/phases/todo/104_chip_sizing_question_cap/04_e2e_suite_commit.md
rename to .agents/phases/complete/104_chip_sizing_question_cap/04_e2e_suite_commit.md
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.err b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.err
new file mode 100644
index 0000000..e69de29
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.md b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.md
new file mode 100644
index 0000000..fd193f6
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.md
@@ -0,0 +1,11 @@
+All completion criteria verified. Final report:
+
+**Phase 104 — final verification pass: all green**
+- Verified (no code changes needed): single-line ellipsized `.suggestion-chip` CSS + deleted `.maybe-try` override, `renderChips` full-text `title` + `aria-label`-when-clipped, `maxlength="4000"` + `#char-count` counter + `handleSend` over-cap guard, unit pins file, schemas boundary pins, dedicated E2E suite.
+- E2E (isolation): `uv run pytest tests/e2e/test_chip_sizing_question_cap.py -v --no-cov` → **6 passed**; regressions: `test_suggestion_chips.py` 8 passed, `test_pinned_composer.py` 4 passed, `test_responsive_polish.py` 7 passed, `test_chat_history.py` 5 passed.
+- `uv run pytest` → **2102 passed**; `--cov=app` → **99%** (>90%); `uv run ruff check . && uv run pyright` → clean, 0 errors.
+- Criteria: chip E2E (single-line, clipped, title+aria-label full text) ✅; paste caps at exactly 4,000, send streams, counter hides ✅; programmatic 5,000-char fill → banner, no turn, text kept ✅; 4,000/4,001 boundary pinned + HTML maxlength == JS constant cross-file pin ✅.
+- Diff scope: `frontend/`, new unit file, `tests/unit/test_schemas.py`, new E2E file, phase files — **no `app/` diff, no migration, no `shared.js` diff**.
+- Deviations: 4 regression test files touched — 2 genuine DOM-pin conflicts from the new `#char-count` child (explicitly anticipated by the overview) + 3 documented **pre-existing E2E flake fixes** (smooth-scroll race, tab-walk heuristic, 10 ms timeout), each verified pre-existing on the pre-phase-104 tree.
+- No commit made (harness commits per the execution protocol override).
+- Next pending phase: `98_sync_summary_visibility`.
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.validate b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.validate
new file mode 100644
index 0000000..15c88de
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__00_phase.a1.validate
@@ -0,0 +1,97 @@
+........................................................................ [  3%]
+........................................................................ [  6%]
+........................................................................ [ 10%]
+........................................................................ [ 13%]
+........................................................................ [ 17%]
+........................................................................ [ 20%]
+........................................................................ [ 23%]
+........................................................................ [ 27%]
+........................................................................ [ 30%]
+........................................................................ [ 34%]
+........................................................................ [ 37%]
+........................................................................ [ 41%]
+........................................................................ [ 44%]
+........................................................................ [ 47%]
+........................................................................ [ 51%]
+........................................................................ [ 54%]
+........................................................................ [ 58%]
+........................................................................ [ 61%]
+........................................................................ [ 65%]
+........................................................................ [ 68%]
+........................................................................ [ 71%]
+........................................................................ [ 75%]
+........................................................................ [ 78%]
+........................................................................ [ 82%]
+........................................................................ [ 85%]
+........................................................................ [ 89%]
+........................................................................ [ 92%]
+........................................................................ [ 95%]
+........................................................................ [ 99%]
+..............                                                           [100%]
+=============================== warnings summary ===============================
+.venv/lib/python3.13/site-packages/fastapi/testclient.py:1
+  /var/home/ducoterra/Projects/Personal/brain_of_reese/.venv/lib/python3.13/site-packages/fastapi/testclient.py:1: StarletteDeprecationWarning: Using `httpx` with `starlette.testclient` is deprecated; install `httpx2` instead.
+    from starlette.testclient import TestClient as TestClient  # noqa
+
+-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
+================================ tests coverage ================================
+_______________ coverage: platform linux, python 3.13.13-final-0 _______________
+
+Name                           Stmts   Miss  Cover
+--------------------------------------------------
+app/__init__.py                    1      0   100%
+app/api/__init__.py                0      0   100%
+app/api/auth.py                   52      0   100%
+app/api/chat.py                  181      0   100%
+app/api/chats.py                 110      0   100%
+app/api/config.py                 13      0   100%
+app/api/doc_drafts.py             94      0   100%
+app/api/docs.py                  134      1    99%
+app/api/git_sources.py           229      0   100%
+app/api/health.py                 10      0   100%
+app/api/steering.py               42      0   100%
+app/api/suggestions.py            33      0   100%
+app/api/sync.py                  115      0   100%
+app/api/tokens.py                 40      0   100%
+app/api/ui_settings.py            55      0   100%
+app/config.py                    140      0   100%
+app/core/__init__.py               0      0   100%
+app/core/auth.py                  45      0   100%
+app/core/caching.py              124      0   100%
+app/core/debugging.py             29      2    93%
+app/core/docs_push.py             39      0   100%
+app/core/errors.py                 5      0   100%
+app/core/logging.py               13      0   100%
+app/core/rate_limit.py            44      0   100%
+app/core/security_headers.py      20      0   100%
+app/core/theming.py               38      0   100%
+app/core/tokens.py                44      0   100%
+app/db.py                         21      0   100%
+app/main.py                       66      0   100%
+app/models.py                    125      0   100%
+app/rag/__init__.py                0      0   100%
+app/rag/agent.py                 316      1    99%
+app/rag/archive_upload.py        128      0   100%
+app/rag/chunker.py               206      4    98%
+app/rag/folder_summaries.py      121      0   100%
+app/rag/git_sources.py            14      0   100%
+app/rag/importer.py              197      3    98%
+app/rag/llm.py                   243      1    99%
+app/rag/overview.py               71      0   100%
+app/rag/prompts.py                88      0   100%
+app/rag/retriever.py             150      3    98%
+app/rag/scaffolding.py            55      0   100%
+app/rag/source_removal.py         41      0   100%
+app/rag/sources_meta.py           16      0   100%
+app/rag/suggestions.py            27      0   100%
+app/rag/summarizer.py             24      0   100%
+app/schemas.py                   307      0   100%
+--------------------------------------------------
+TOTAL                           3866     15    99%
+coverage gate: app/ 99% (>90%) OK
+All checks passed!
+0 errors, 0 warnings, 0 informations
+WARNING: there is a new pyright version available (v1.1.411 -> v1.1.414).
+Please install the new version or set PYRIGHT_PYTHON_FORCE_VERSION to `latest`
+
+validation OK
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.err b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.err
new file mode 100644
index 0000000..e69de29
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.md b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.md
new file mode 100644
index 0000000..bdba8c7
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.md
@@ -0,0 +1,18 @@
+Task 01 complete — all gates green.
+
+**Implemented**
+- `frontend/assets/styles.css`: base `.suggestion-chip` rule gains the single-line contract (`white-space: nowrap`, `overflow: hidden`, `text-overflow: ellipsis`, `max-width: 100%`, `min-width: 0`) with the phase-104 provenance comment (incl. the folded-in phase-07 origin).
+- Deleted the fully subsumed `.maybe-try .suggestion-chip` override (its phase-07 provenance lives in the base rule's comment, phrased without the literal selector so the `count == 0` pin holds).
+- Created `tests/unit/test_chip_sizing_question_cap.py` — the phase's unit-pins file with the three task-01 CSS pins (ellipsis quartet + pill contract kept, override gone / `.maybe-try` group stays, ≤640px `flex: 0 0 auto` track untouched).
+- Mobile rule and `.shared-shell` static-chip rule untouched (A6); no JS/HTML/E2E changes (later tasks).
+
+**Test / lint / coverage**
+- `uv run pytest tests/unit/ -q` → 1564 passed
+- `uv run pytest` (unit+integration, DB up) → 2089 passed
+- `uv run pytest --cov=app --cov-report=term-missing` → TOTAL 99% (>90% gate)
+- `uv run ruff check .` → All checks passed; `uv run pyright` → 0 errors, 0 warnings
+- `git diff --stat` → `frontend/assets/styles.css` only (+ new unit file; no `app/` diff)
+
+**Decisions:** none deviating — followed the phase design verbatim.
+
+**Next pending task:** `.agents/phases/todo/104_chip_sizing_question_cap/02_chip_tooltip_aria.md`
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.validate b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.validate
new file mode 100644
index 0000000..9ca2a44
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__01_chip_ellipsis_css.a1.validate
@@ -0,0 +1,97 @@
+........................................................................ [  3%]
+........................................................................ [  6%]
+........................................................................ [ 10%]
+........................................................................ [ 13%]
+........................................................................ [ 17%]
+........................................................................ [ 20%]
+........................................................................ [ 24%]
+........................................................................ [ 27%]
+........................................................................ [ 31%]
+........................................................................ [ 34%]
+........................................................................ [ 37%]
+........................................................................ [ 41%]
+........................................................................ [ 44%]
+........................................................................ [ 48%]
+........................................................................ [ 51%]
+........................................................................ [ 55%]
+........................................................................ [ 58%]
+........................................................................ [ 62%]
+........................................................................ [ 65%]
+........................................................................ [ 68%]
+........................................................................ [ 72%]
+........................................................................ [ 75%]
+........................................................................ [ 79%]
+........................................................................ [ 82%]
+........................................................................ [ 86%]
+........................................................................ [ 89%]
+........................................................................ [ 93%]
+........................................................................ [ 96%]
+........................................................................ [ 99%]
+.                                                                        [100%]
+=============================== warnings summary ===============================
+.venv/lib/python3.13/site-packages/fastapi/testclient.py:1
+  /var/home/ducoterra/Projects/Personal/brain_of_reese/.venv/lib/python3.13/site-packages/fastapi/testclient.py:1: StarletteDeprecationWarning: Using `httpx` with `starlette.testclient` is deprecated; install `httpx2` instead.
+    from starlette.testclient import TestClient as TestClient  # noqa
+
+-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
+================================ tests coverage ================================
+_______________ coverage: platform linux, python 3.13.13-final-0 _______________
+
+Name                           Stmts   Miss  Cover
+--------------------------------------------------
+app/__init__.py                    1      0   100%
+app/api/__init__.py                0      0   100%
+app/api/auth.py                   52      0   100%
+app/api/chat.py                  181      0   100%
+app/api/chats.py                 110      0   100%
+app/api/config.py                 13      0   100%
+app/api/doc_drafts.py             94      0   100%
+app/api/docs.py                  134      1    99%
+app/api/git_sources.py           229      0   100%
+app/api/health.py                 10      0   100%
+app/api/steering.py               42      0   100%
+app/api/suggestions.py            33      0   100%
+app/api/sync.py                  115      0   100%
+app/api/tokens.py                 40      0   100%
+app/api/ui_settings.py            55      0   100%
+app/config.py                    140      0   100%
+app/core/__init__.py               0      0   100%
+app/core/auth.py                  45      0   100%
+app/core/caching.py              124      0   100%
+app/core/debugging.py             29      2    93%
+app/core/docs_push.py             39      0   100%
+app/core/errors.py                 5      0   100%
+app/core/logging.py               13      0   100%
+app/core/rate_limit.py            44      0   100%
+app/core/security_headers.py      20      0   100%
+app/core/theming.py               38      0   100%
+app/core/tokens.py                44      0   100%
+app/db.py                         21      0   100%
+app/main.py                       66      0   100%
+app/models.py                    125      0   100%
+app/rag/__init__.py                0      0   100%
+app/rag/agent.py                 316      1    99%
+app/rag/archive_upload.py        128      0   100%
+app/rag/chunker.py               206      4    98%
+app/rag/folder_summaries.py      121      0   100%
+app/rag/git_sources.py            14      0   100%
+app/rag/importer.py              197      3    98%
+app/rag/llm.py                   243      1    99%
+app/rag/overview.py               71      0   100%
+app/rag/prompts.py                88      0   100%
+app/rag/retriever.py             150      3    98%
+app/rag/scaffolding.py            55      0   100%
+app/rag/source_removal.py         41      0   100%
+app/rag/sources_meta.py           16      0   100%
+app/rag/suggestions.py            27      0   100%
+app/rag/summarizer.py             24      0   100%
+app/schemas.py                   307      0   100%
+--------------------------------------------------
+TOTAL                           3866     15    99%
+coverage gate: app/ 99% (>90%) OK
+All checks passed!
+0 errors, 0 warnings, 0 informations
+WARNING: there is a new pyright version available (v1.1.411 -> v1.1.414).
+Please install the new version or set PYRIGHT_PYTHON_FORCE_VERSION to `latest`
+
+validation OK
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.err b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.err
new file mode 100644
index 0000000..e69de29
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.md b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.md
new file mode 100644
index 0000000..a538074
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.md
@@ -0,0 +1,15 @@
+Task 02 complete. All checks green.
+
+**Implemented**
+- `frontend/assets/app.js` `renderChips`: `btn.title = text` on every chip (full-text hover tooltip, owner A2, source-chip precedent) + `aria-label` = full text only when `btn.scrollWidth > btn.clientWidth` (clipped case), with the phase-104 provenance comment; nothing else in `renderChips` touched
+- `tests/unit/test_chip_sizing_question_cap.py`: added `APP_JS`/`_app_js()`/`_render_chips_body()` (brace-matched extraction, handles the `{ onSelect }` destructuring) + two task-02 pins (`btn.title = text`; the guard-anchored `setAttribute("aria-label", text)`) + module docstring updated
+
+**Results**
+- `uv run pytest tests/unit/` → 1566 passed
+- `uv run pytest` (unit + integration) → 2091 passed
+- `uv run pytest --cov=app --cov-report=term` → 2091 passed, `app/` coverage 99% (>90% gate)
+- `uv run ruff check .` → All checks passed; `uv run pyright` → 0 errors
+
+**Decisions:** `shared.js` untouched (its "no renderChips" pin stays green — it guards a different file).
+
+**Next pending task:** `.agents/phases/todo/104_chip_sizing_question_cap/03_composer_question_cap.md`
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.validate b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.validate
new file mode 100644
index 0000000..98f85ea
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__02_chip_tooltip_aria.a1.validate
@@ -0,0 +1,97 @@
+........................................................................ [  3%]
+........................................................................ [  6%]
+........................................................................ [ 10%]
+........................................................................ [ 13%]
+........................................................................ [ 17%]
+........................................................................ [ 20%]
+........................................................................ [ 24%]
+........................................................................ [ 27%]
+........................................................................ [ 30%]
+........................................................................ [ 34%]
+........................................................................ [ 37%]
+........................................................................ [ 41%]
+........................................................................ [ 44%]
+........................................................................ [ 48%]
+........................................................................ [ 51%]
+........................................................................ [ 55%]
+........................................................................ [ 58%]
+........................................................................ [ 61%]
+........................................................................ [ 65%]
+........................................................................ [ 68%]
+........................................................................ [ 72%]
+........................................................................ [ 75%]
+........................................................................ [ 79%]
+........................................................................ [ 82%]
+........................................................................ [ 86%]
+........................................................................ [ 89%]
+........................................................................ [ 92%]
+........................................................................ [ 96%]
+........................................................................ [ 99%]
+...                                                                      [100%]
+=============================== warnings summary ===============================
+.venv/lib/python3.13/site-packages/fastapi/testclient.py:1
+  /var/home/ducoterra/Projects/Personal/brain_of_reese/.venv/lib/python3.13/site-packages/fastapi/testclient.py:1: StarletteDeprecationWarning: Using `httpx` with `starlette.testclient` is deprecated; install `httpx2` instead.
+    from starlette.testclient import TestClient as TestClient  # noqa
+
+-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
+================================ tests coverage ================================
+_______________ coverage: platform linux, python 3.13.13-final-0 _______________
+
+Name                           Stmts   Miss  Cover
+--------------------------------------------------
+app/__init__.py                    1      0   100%
+app/api/__init__.py                0      0   100%
+app/api/auth.py                   52      0   100%
+app/api/chat.py                  181      0   100%
+app/api/chats.py                 110      0   100%
+app/api/config.py                 13      0   100%
+app/api/doc_drafts.py             94      0   100%
+app/api/docs.py                  134      1    99%
+app/api/git_sources.py           229      0   100%
+app/api/health.py                 10      0   100%
+app/api/steering.py               42      0   100%
+app/api/suggestions.py            33      0   100%
+app/api/sync.py                  115      0   100%
+app/api/tokens.py                 40      0   100%
+app/api/ui_settings.py            55      0   100%
+app/config.py                    140      0   100%
+app/core/__init__.py               0      0   100%
+app/core/auth.py                  45      0   100%
+app/core/caching.py              124      0   100%
+app/core/debugging.py             29      2    93%
+app/core/docs_push.py             39      0   100%
+app/core/errors.py                 5      0   100%
+app/core/logging.py               13      0   100%
+app/core/rate_limit.py            44      0   100%
+app/core/security_headers.py      20      0   100%
+app/core/theming.py               38      0   100%
+app/core/tokens.py                44      0   100%
+app/db.py                         21      0   100%
+app/main.py                       66      0   100%
+app/models.py                    125      0   100%
+app/rag/__init__.py                0      0   100%
+app/rag/agent.py                 316      1    99%
+app/rag/archive_upload.py        128      0   100%
+app/rag/chunker.py               206      4    98%
+app/rag/folder_summaries.py      121      0   100%
+app/rag/git_sources.py            14      0   100%
+app/rag/importer.py              197      3    98%
+app/rag/llm.py                   243      1    99%
+app/rag/overview.py               71      0   100%
+app/rag/prompts.py                88      0   100%
+app/rag/retriever.py             150      3    98%
+app/rag/scaffolding.py            55      0   100%
+app/rag/source_removal.py         41      0   100%
+app/rag/sources_meta.py           16      0   100%
+app/rag/suggestions.py            27      0   100%
+app/rag/summarizer.py             24      0   100%
+app/schemas.py                   307      0   100%
+--------------------------------------------------
+TOTAL                           3866     15    99%
+coverage gate: app/ 99% (>90%) OK
+All checks passed!
+0 errors, 0 warnings, 0 informations
+WARNING: there is a new pyright version available (v1.1.411 -> v1.1.414).
+Please install the new version or set PYRIGHT_PYTHON_FORCE_VERSION to `latest`
+
+validation OK
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.err b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.err
new file mode 100644
index 0000000..e69de29
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.md b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.md
new file mode 100644
index 0000000..a2a2ccd
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.md
@@ -0,0 +1,14 @@
+All gates green. Final report:
+
+**Task 03 complete — visible 4,000-char question cap (maxlength + counter + guard + server-boundary pin)**
+
+- `#message-input` gains `maxlength="4000"` with provenance comment; `#char-count` counter added inside `.chat-bottom` between the actions row and composer, hidden by default, with the phase-104 decision comment (not a live region; B3 text+color)
+- `app.js`: `MAX_QUESTION_CHARS=4000` / `CHAR_COUNT_SHOW_AT=3200` (80%), `updateCharCount()` (RAW length, `.is-max` + "— character limit" copy at/over cap, honest over-cap reading), called at all four `input.value` mutation sites; `handleSend` over-cap guard (banner, no turn, text kept) before the clear
+- `styles.css`: `.char-count` / `.char-count.is-max` rules — AA-verified on `--bg` (8.6:1, 10.4:1) with ratios recorded in the comment
+- Unit pins: `test_chip_sizing_question_cap.py` extended (HTML/JS/CSS pins + single-source cross-file pin HTML==JS==schema); `test_schemas.py` 4,000/4,001 `ChatRequest.message` boundary; two conflicting unit pins (`test_pinned_composer.py`, `test_save_chat_ui.py`) updated to carry the new counter child (genuine conflicts anticipated by the phase)
+- **Defect fixes in prior work:** `e2e/test_pinned_composer.py::test_stop_is_reachable_from_scrolled_up` was racy (proven on the pristine tree — the test scrolled up mid smooth-`scrollReveal`; reads caught y=23–119) → deterministic arrival-wait on the captured scroll target; `e2e/test_suggestion_chips.py` `_api_suggestions` `timeout=10` (10 ms) sat below the ~12 ms steady-state localhost round trip → 10 s house pattern
+
+- Results: `uv run pytest tests/unit/ -q` → 1577 passed; `uv run pytest --cov=app` → 2102 passed, **99%** coverage (>90%); `uv run ruff check .` → clean; `uv run pyright` → 0 errors
+- Regression E2Es (isolated): `test_pinned_composer.py` 4 passed (×6 consecutive), `test_suggestion_chips.py` 8 passed (×2), `test_responsive_polish.py` 7 passed, `test_chat_history.py` 5 passed; no `app/` diff
+
+- Next pending task: `04_e2e_suite_commit.md` (dedicated E2E suite + full gate + atomic commit)
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.validate b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.validate
new file mode 100644
index 0000000..15c88de
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__03_composer_question_cap.a1.validate
@@ -0,0 +1,97 @@
+........................................................................ [  3%]
+........................................................................ [  6%]
+........................................................................ [ 10%]
+........................................................................ [ 13%]
+........................................................................ [ 17%]
+........................................................................ [ 20%]
+........................................................................ [ 23%]
+........................................................................ [ 27%]
+........................................................................ [ 30%]
+........................................................................ [ 34%]
+........................................................................ [ 37%]
+........................................................................ [ 41%]
+........................................................................ [ 44%]
+........................................................................ [ 47%]
+........................................................................ [ 51%]
+........................................................................ [ 54%]
+........................................................................ [ 58%]
+........................................................................ [ 61%]
+........................................................................ [ 65%]
+........................................................................ [ 68%]
+........................................................................ [ 71%]
+........................................................................ [ 75%]
+........................................................................ [ 78%]
+........................................................................ [ 82%]
+........................................................................ [ 85%]
+........................................................................ [ 89%]
+........................................................................ [ 92%]
+........................................................................ [ 95%]
+........................................................................ [ 99%]
+..............                                                           [100%]
+=============================== warnings summary ===============================
+.venv/lib/python3.13/site-packages/fastapi/testclient.py:1
+  /var/home/ducoterra/Projects/Personal/brain_of_reese/.venv/lib/python3.13/site-packages/fastapi/testclient.py:1: StarletteDeprecationWarning: Using `httpx` with `starlette.testclient` is deprecated; install `httpx2` instead.
+    from starlette.testclient import TestClient as TestClient  # noqa
+
+-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
+================================ tests coverage ================================
+_______________ coverage: platform linux, python 3.13.13-final-0 _______________
+
+Name                           Stmts   Miss  Cover
+--------------------------------------------------
+app/__init__.py                    1      0   100%
+app/api/__init__.py                0      0   100%
+app/api/auth.py                   52      0   100%
+app/api/chat.py                  181      0   100%
+app/api/chats.py                 110      0   100%
+app/api/config.py                 13      0   100%
+app/api/doc_drafts.py             94      0   100%
+app/api/docs.py                  134      1    99%
+app/api/git_sources.py           229      0   100%
+app/api/health.py                 10      0   100%
+app/api/steering.py               42      0   100%
+app/api/suggestions.py            33      0   100%
+app/api/sync.py                  115      0   100%
+app/api/tokens.py                 40      0   100%
+app/api/ui_settings.py            55      0   100%
+app/config.py                    140      0   100%
+app/core/__init__.py               0      0   100%
+app/core/auth.py                  45      0   100%
+app/core/caching.py              124      0   100%
+app/core/debugging.py             29      2    93%
+app/core/docs_push.py             39      0   100%
+app/core/errors.py                 5      0   100%
+app/core/logging.py               13      0   100%
+app/core/rate_limit.py            44      0   100%
+app/core/security_headers.py      20      0   100%
+app/core/theming.py               38      0   100%
+app/core/tokens.py                44      0   100%
+app/db.py                         21      0   100%
+app/main.py                       66      0   100%
+app/models.py                    125      0   100%
+app/rag/__init__.py                0      0   100%
+app/rag/agent.py                 316      1    99%
+app/rag/archive_upload.py        128      0   100%
+app/rag/chunker.py               206      4    98%
+app/rag/folder_summaries.py      121      0   100%
+app/rag/git_sources.py            14      0   100%
+app/rag/importer.py              197      3    98%
+app/rag/llm.py                   243      1    99%
+app/rag/overview.py               71      0   100%
+app/rag/prompts.py                88      0   100%
+app/rag/retriever.py             150      3    98%
+app/rag/scaffolding.py            55      0   100%
+app/rag/source_removal.py         41      0   100%
+app/rag/sources_meta.py           16      0   100%
+app/rag/suggestions.py            27      0   100%
+app/rag/summarizer.py             24      0   100%
+app/schemas.py                   307      0   100%
+--------------------------------------------------
+TOTAL                           3866     15    99%
+coverage gate: app/ 99% (>90%) OK
+All checks passed!
+0 errors, 0 warnings, 0 informations
+WARNING: there is a new pyright version available (v1.1.411 -> v1.1.414).
+Please install the new version or set PYRIGHT_PYTHON_FORCE_VERSION to `latest`
+
+validation OK
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.err b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.err
new file mode 100644
index 0000000..e69de29
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.md b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.md
new file mode 100644
index 0000000..c54e8d2
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.md
@@ -0,0 +1,11 @@
+All gates green. Final report:
+
+**Phase 104 · Task 04 — E2E suite + regression gates**
+
+- New dedicated suite `tests/e2e/test_chip_sizing_question_cap.py` (6 tests: single-line ellipsized long chip with `title`/`aria-label` full text, seed-chip tooltip-without-aria contrast, counter threshold 100/3500, paste-path hard cap at 4,000 + clean submit at cap, over-cap programmatic-fill guard (banner, no turn, text kept), short-flow never-visible counter with MutationObserver tripwire) — drafted in the interrupted prior attempt, verified green
+- Fixed a **pre-existing flake** in `test_responsive_polish.py::test_a11y_landmarks_and_labels` (verified identical on the pristine pre-phase-104 tree): `_tab_outline_walk` keyed elements by 24-char text prefix, so two identical opener chips left in shared-DB `saved_chats` by `test_pinned_composer.py` cut the walk short (2 < 3); now uses true element identity via a stamped `data-tabwalk-id`. No behavior change to the asserted contract
+- E2E in isolation: `uv run pytest tests/e2e/test_chip_sizing_question_cap.py -v --no-cov` → 6 passed; regressions each green in isolation: `test_suggestion_chips` 8, `test_pinned_composer` 4, `test_responsive_polish` 7 (incl. adversarial post-pinned_composer state, 3 runs), `test_chat_history` 5
+- Full gate: `uv run pytest` → 2102 passed; `uv run pytest --cov=app --cov-report=term-missing` → 2102 passed, `app/` **99%** (>90%); `uv run ruff check . && uv run pyright` → clean, 0 errors
+- `git diff --stat` scoped: `frontend/` (app.js/styles.css/index.html), new unit+E2E files, `test_schemas.py`, pin updates (`test_pinned_composer` unit, `test_save_chat_ui`), prior-task e2e flake fixes (`test_pinned_composer`, `test_suggestion_chips`) + this fix; NO `app/` code diff, no migration, no `shared.js`, no `pyproject.toml`/`uv.lock`
+- Phase dir move + atomic `--no-gpg-sign` commit: left to the harness per executor rules (overrides task step 5)
+- Next pending task: none in this phase (04 is the last); next phase = whatever follows `104` in `.agents/phases/todo/`
diff --git a/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.validate b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.validate
new file mode 100644
index 0000000..15c88de
--- /dev/null
+++ b/.agents/reports/104_chip_sizing_question_cap/104_chip_sizing_question_cap__04_e2e_suite_commit.a1.validate
@@ -0,0 +1,97 @@
+........................................................................ [  3%]
+........................................................................ [  6%]
+........................................................................ [ 10%]
+........................................................................ [ 13%]
+........................................................................ [ 17%]
+........................................................................ [ 20%]
+........................................................................ [ 23%]
+........................................................................ [ 27%]
+........................................................................ [ 30%]
+........................................................................ [ 34%]
+........................................................................ [ 37%]
+........................................................................ [ 41%]
+........................................................................ [ 44%]
+........................................................................ [ 47%]
+........................................................................ [ 51%]
+........................................................................ [ 54%]
+........................................................................ [ 58%]
+........................................................................ [ 61%]
+........................................................................ [ 65%]
+........................................................................ [ 68%]
+........................................................................ [ 71%]
+........................................................................ [ 75%]
+........................................................................ [ 78%]
+........................................................................ [ 82%]
+........................................................................ [ 85%]
+........................................................................ [ 89%]
+........................................................................ [ 92%]
+........................................................................ [ 95%]
+........................................................................ [ 99%]
+..............                                                           [100%]
+=============================== warnings summary ===============================
+.venv/lib/python3.13/site-packages/fastapi/testclient.py:1
+  /var/home/ducoterra/Projects/Personal/brain_of_reese/.venv/lib/python3.13/site-packages/fastapi/testclient.py:1: StarletteDeprecationWarning: Using `httpx` with `starlette.testclient` is deprecated; install `httpx2` instead.
+    from starlette.testclient import TestClient as TestClient  # noqa
+
+-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
+================================ tests coverage ================================
+_______________ coverage: platform linux, python 3.13.13-final-0 _______________
+
+Name                           Stmts   Miss  Cover
+--------------------------------------------------
+app/__init__.py                    1      0   100%
+app/api/__init__.py                0      0   100%
+app/api/auth.py                   52      0   100%
+app/api/chat.py                  181      0   100%
+app/api/chats.py                 110      0   100%
+app/api/config.py                 13      0   100%
+app/api/doc_drafts.py             94      0   100%
+app/api/docs.py                  134      1    99%
+app/api/git_sources.py           229      0   100%
+app/api/health.py                 10      0   100%
+app/api/steering.py               42      0   100%
+app/api/suggestions.py            33      0   100%
+app/api/sync.py                  115      0   100%
+app/api/tokens.py                 40      0   100%
+app/api/ui_settings.py            55      0   100%
+app/config.py                    140      0   100%
+app/core/__init__.py               0      0   100%
+app/core/auth.py                  45      0   100%
+app/core/caching.py              124      0   100%
+app/core/debugging.py             29      2    93%
+app/core/docs_push.py             39      0   100%
+app/core/errors.py                 5      0   100%
+app/core/logging.py               13      0   100%
+app/core/rate_limit.py            44      0   100%
+app/core/security_headers.py      20      0   100%
+app/core/theming.py               38      0   100%
+app/core/tokens.py                44      0   100%
+app/db.py                         21      0   100%
+app/main.py                       66      0   100%
+app/models.py                    125      0   100%
+app/rag/__init__.py                0      0   100%
+app/rag/agent.py                 316      1    99%
+app/rag/archive_upload.py        128      0   100%
+app/rag/chunker.py               206      4    98%
+app/rag/folder_summaries.py      121      0   100%
+app/rag/git_sources.py            14      0   100%
+app/rag/importer.py              197      3    98%
+app/rag/llm.py                   243      1    99%
+app/rag/overview.py               71      0   100%
+app/rag/prompts.py                88      0   100%
+app/rag/retriever.py             150      3    98%
+app/rag/scaffolding.py            55      0   100%
+app/rag/source_removal.py         41      0   100%
+app/rag/sources_meta.py           16      0   100%
+app/rag/suggestions.py            27      0   100%
+app/rag/summarizer.py             24      0   100%
+app/schemas.py                   307      0   100%
+--------------------------------------------------
+TOTAL                           3866     15    99%
+coverage gate: app/ 99% (>90%) OK
+All checks passed!
+0 errors, 0 warnings, 0 informations
+WARNING: there is a new pyright version available (v1.1.411 -> v1.1.414).
+Please install the new version or set PYRIGHT_PYTHON_FORCE_VERSION to `latest`
+
+validation OK
diff --git a/frontend/assets/app.js b/frontend/assets/app.js
index 89c1636..cbb8dd4 100644
--- a/frontend/assets/app.js
+++ b/frontend/assets/app.js
@@ -299,6 +299,7 @@ bindSharedHeaderControls();
 const messagesEl = document.querySelector("#messages");
 const emptyState = document.querySelector("#empty-state");
 const suggestionsEl = document.querySelector("#suggestions");
+const charCountEl = document.querySelector("#char-count"); // phase 104: the question-length counter (ships hidden)
 const composer = document.querySelector("#composer");
 const input = document.querySelector("#message-input");
 const sendBtn = document.querySelector("#send-btn");
@@ -1012,6 +1013,7 @@ function appendTruncatedNote(wrap, argument, charsShown, charsTotal) {
 function submitSuggestion(text) {
   input.value = text;
   autoGrow();
+  updateCharCount(); // phase 104: the chip fill bypasses maxlength — count what landed
   input.focus();
   composer.requestSubmit();
 }
@@ -1028,11 +1030,16 @@ function renderChips(container, items, { onSelect } = {}) {
     btn.className = "suggestion-chip";
     btn.setAttribute("role", "listitem");
     btn.textContent = text;
+    // phase 104 (owner 2026-09-12): the single-line chip clips long
+    // questions — `title` is the hover reveal, `aria-label` the
+    // clipped-case accessible name (the source-chip pattern).
+    btn.title = text;
     btn.addEventListener("click", () => {
       submitSuggestion(text);
       if (onSelect) onSelect(text, btn);
     });
     container.appendChild(btn);
+    if (btn.scrollWidth > btn.clientWidth) btn.setAttribute("aria-label", text);
   }
   return container;
 }
@@ -1255,11 +1262,42 @@ export function setUiState(state, errorDetail = "") {
   if (state === UI_STATE.error) showErrorBanner(errorDetail);
 }
 
+/* Phase 104 (owner 2026-09-12, A3/A4): the visible question-length cap.
+   MAX_QUESTION_CHARS mirrors ChatRequest.message max_length=4000
+   (app/schemas.py — the source of truth; the server 422s beyond it) and
+   must stay equal to the #message-input maxlength (cross-file pin in
+   tests/unit/test_chip_sizing_question_cap.py). The counter shows only
+   from 80% of the cap — no noise on normal use (owner A4). */
+const MAX_QUESTION_CHARS = 4000;
+const CHAR_COUNT_SHOW_AT = 3200; // 80% of the cap — visible only when it matters
+
 function autoGrow() {
   input.style.height = "auto";
   input.style.height = `${Math.min(input.scrollHeight, 192)}px`;
 }
 
+/* Phase 104 (owner 2026-09-12, A4): the counter state — hidden below the
+   80% threshold, plain `len/4000` above it, and `len/4000 — character
+   limit` + the .is-max (–err-*) treatment at/over the cap. The over-cap
+   reading is the HONEST length (the programmatic chip-fill path can
+   exceed maxlength — e.g. `5123/4000 — character limit`). */
+function updateCharCount() {
+  // RAW length (no trim): raw ≤ cap ⟹ trimmed ≤ cap, so the raw
+  // count is a safe superset of what the server validates.
+  const len = input.value.length;
+  if (len < CHAR_COUNT_SHOW_AT) {
+    charCountEl.hidden = true;
+    charCountEl.classList.remove("is-max");
+    return;
+  }
+  charCountEl.hidden = false;
+  const atMax = len >= MAX_QUESTION_CHARS;
+  charCountEl.classList.toggle("is-max", atMax);
+  charCountEl.textContent = atMax
+    ? `${len}/${MAX_QUESTION_CHARS} — character limit`
+    : `${len}/${MAX_QUESTION_CHARS}`;
+}
+
 /* ---------- chat turn (SSE streaming, PLAN §4) ---------- */
 
 /* Cancel a response body without leaking an unhandled rejection:
@@ -2014,6 +2052,7 @@ function startNewChat() {
   setUiState(UI_STATE.idle);
   input.value = "";
   autoGrow();
+  updateCharCount(); // phase 104: the cleared composer hides the counter again
   input.focus();
   sendStatus.textContent = "New chat started — previous conversation cleared.";
 }
@@ -2111,12 +2150,21 @@ async function handleSend(e) {
   }
   const text = input.value.trim();
   if (!text || sendBtn.disabled) return;
+  // Phase 104 (owner 2026-09-12, A5): maxlength caps typing + pastes, but
+  // a programmatic fill (the chip one-tap path) bypasses it — this guard
+  // is the never-stale backstop (PLAN §7.4): no turn, no clear, the user
+  // trims the kept text (out-of-turn banner, the saveAsDoc precedent).
+  if (text.length > MAX_QUESTION_CHARS) {
+    showErrorBanner("Questions are limited to 4,000 characters — trim the question and try again.");
+    return;
+  }
   // Phase 49: the user append + persistence save point 1 moved into
   // runTurn with the rest of the turn — the `reask` flag skips them on
   // the redo-in-place retry path (the question is already in the DOM +
   // conversation); handleSend keeps only the form-level pre-work.
   input.value = "";
   autoGrow();
+  updateCharCount(); // phase 104: the sent question clears the counter with the input
   clearErrorBanner();
   await runTurn(text, { reask: false });
 }
@@ -2481,7 +2529,12 @@ async function runTurn(text, { reask = false } = {}) {
   }
 }
 
-input.addEventListener("input", autoGrow);
+// Phase 104: every input-path change (keystroke, paste — maxlength
+// caps both at 4,000) re-runs the counter alongside the auto-grow.
+input.addEventListener("input", () => {
+  autoGrow();
+  updateCharCount();
+});
 input.addEventListener("keydown", (e) => {
   if (e.key === "Enter" && !e.shiftKey) {
     e.preventDefault();
diff --git a/frontend/assets/styles.css b/frontend/assets/styles.css
index 323e694..8cd153e 100644
--- a/frontend/assets/styles.css
+++ b/frontend/assets/styles.css
@@ -703,14 +703,6 @@ details.thinking .thinking-text ul { margin: 0 0 0.5rem; }
   gap: 0.45rem;
   padding-inline: 0.25rem;
 }
-/* As flex items these chips must be allowed to shrink (min-width:auto
-   would let a long title-derived chip exceed the column on phones —
-   phase 07 overflow fix); the label text then wraps inside the pill. */
-.maybe-try .suggestion-chip {
-  min-width: 0;
-  max-width: 100%;
-}
-
 /* ---------- Steering notes (phase 15) ---------- */
 /* "Tune" button in the meta row of every completed brain bubble: ghost
    pill, ≥44px, right-aligned after the source chips. ink-soft on surface
@@ -1230,6 +1222,21 @@ details.thinking .thinking-text ul { margin: 0 0 0.5rem; }
   margin: 0;
   padding: 0;
 }
+/* The shared chip pill — both chip rows (the onboarding row in the
+   empty state and the "Maybe try" row under a deflected bubble) reuse
+   it. Phase 104 (owner 2026-09-12, A1): a chip is ONE line at every
+   viewport width — `white-space: nowrap` stops a 400-char question
+   from wrapping the pill into a multi-line "chonk"; `overflow: hidden`
+   zeroes the flex item's automatic minimum size (min-width:auto would
+   pin the item to the full text width), so `max-width: 100%` binds and
+   `text-overflow: ellipsis` clips the text at the row edge — in the
+   desktop wrap row 100% is the chat column, in the ≤640px row
+   (nowrap + overflow-x: auto) 100% is the VISIBLE width and the row
+   scrolls. That clipping is the phase-07 overflow fix the old
+   "Maybe try" chip override used to carry (min-width:0 + max-width:
+   100%) — now fully subsumed here and deleted, provenance folded into
+   this comment so the history lives with the contract. The full text
+   stays one hover away (the title tooltip, phase 104 task 02). */
 .suggestion-chip {
   font: inherit;
   font-size: 0.92rem;
@@ -1242,6 +1249,11 @@ details.thinking .thinking-text ul { margin: 0 0 0.5rem; }
   min-height: 44px;
   cursor: pointer;
   transition: background 0.15s ease, transform 0.05s ease;
+  white-space: nowrap;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  max-width: 100%;
+  min-width: 0;
 }
 .suggestion-chip:hover { background: var(--brand-soft); }
 .suggestion-chip:active { transform: scale(0.98); }
@@ -1321,6 +1333,17 @@ details.thinking .thinking-text ul { margin: 0 0 0.5rem; }
 #view-chat:not(.chat-booted) .composer {
   position: static;
 }
+/* Phase 104 (owner 2026-09-12, A4): the question-length counter —
+   right-aligned above the composer, a child of the .chat-bottom sticky
+   unit (between the actions row and the form), hidden until 80% of the
+   4,000-char cap (app.js updateCharCount). It sits on the APP
+   background behind the transparent .chat-bottom unit, so both pairings
+   are verified against --bg: --ink-soft on --bg = 8.6:1 (≥4.5:1, WCAG
+   AA) and the .is-max state's --err-ink on --bg = 10.4:1 (AA). The
+   .is-max state ALSO changes the copy ("— character limit") — text +
+   color, never color alone (B3). */
+.char-count { margin: 0; text-align: right; font-size: 0.75rem; line-height: 1.2; color: var(--ink-soft); }
+.char-count.is-max { color: var(--err-ink); }
 .composer {
   display: flex;
   align-items: flex-end;
diff --git a/frontend/index.html b/frontend/index.html
index 4925500..c8d9f69 100644
--- a/frontend/index.html
+++ b/frontend/index.html
@@ -252,6 +252,15 @@
           </button>
         </div>
 
+        <!-- Phase 104 (owner 2026-09-12): the question-length counter —
+             hidden until 80% of the 4,000-char cap (app.js
+             updateCharCount); the .is-max state uses the --err-* family
+             PLUS a copy change (B3: text + color, never color alone). NOT
+             a live region: per-keystroke feedback is decorative — the
+             over-cap failure path announces through the error banner
+             (role=alert). -->
+        <p class="char-count" id="char-count" hidden></p>
+
           <!-- Composer (phase 48, 2026-08-29, TODO.md L3): one button, two
                roles — #send-btn reads "Send" when idle and morphs into the
                enabled "Stop" control (.is-stop, rose treatment) while a turn
@@ -270,12 +279,16 @@
                precedent as the tuning form's noValidate). -->
         <form class="composer" id="composer" novalidate>
           <label class="visually-hidden" for="message-input">Ask Brain of Reese a question</label>
+          <!-- maxlength=4000 mirrors ChatRequest.message max_length=4000
+               (app/schemas.py) — the server 422s beyond; the #char-count
+               line makes the cap visible (app.js updateCharCount). -->
           <textarea
             id="message-input"
             name="message"
             rows="1"
             placeholder="Ask me anything…"
             autocomplete="off"
+            maxlength="4000"
           ></textarea>
           <button type="submit" class="send-btn" id="send-btn">
             <span class="spinner" aria-hidden="true" hidden></span>
diff --git a/tests/e2e/test_chip_sizing_question_cap.py b/tests/e2e/test_chip_sizing_question_cap.py
new file mode 100644
index 0000000..cbc4367
--- /dev/null
+++ b/tests/e2e/test_chip_sizing_question_cap.py
@@ -0,0 +1,521 @@
+"""Phase 104 E2E (Playwright): single-line suggestion chips (hover reveals
+the full text) + the visible 4,000-char question cap.
+
+Source: owner request 2026-09-12 — "…tweak the size of the suggestion
+chips on the chat page. Some users submit truly massive queries and the
+'chips' become more like 'chonks'. Hovering over the chips should still
+show the full message … put a character cap on the chat submission box,
+to prevent ultra-long context overflowing queries. Users should still be
+able to paste code example of a few dozen lines, but nothing much longer
+than that."
+
+The contract (owner decisions A1–A6, recorded in 00_phase.md):
+
+* **A1/A2 — single-line chips, full text one hover away:** every
+  suggestion chip is ONE line at every viewport width (never a wrapped
+  multi-line "chonk") — long text ellipsizes at the row edge, and the
+  FULL text is recoverable via the native ``title`` tooltip on every
+  chip plus the ``aria-label`` accessible name when the visible text is
+  actually clipped;
+* **A3 — the hard cap through the input path:** the composer textarea
+  carries ``maxlength="4000"`` mirroring the pre-existing server cap
+  (``ChatRequest.message max_length=4000`` — the server 422s beyond);
+  a paste of more than 4,000 chars lands as EXACTLY 4,000, and
+  submitting at the cap is a clean turn (no 422 error state);
+* **A4 — the counter:** ``#char-count`` is hidden while the RAW length
+  is below 80% of the cap (3,200 — no noise on normal use), shows
+  ``len/4000`` from there, and reads ``len/4000 — character limit`` +
+  the ``.is-max`` (err family + copy change, B3) at/over the cap;
+* **A5 — the over-cap guard:** a programmatic fill (the chip one-tap
+  path) bypasses ``maxlength`` — ``handleSend`` refuses trimmed text
+  over the cap with the out-of-turn error banner (the ``saveAsDoc``
+  precedent), NO turn, and the input KEEPS the text (the user trims it
+  — never stale, PLAN §7.4).
+
+The states pinned here (six tests):
+
+1. **truncated-chip** (A1/A2 core) — a saved chat whose FIRST user
+   question is LONG (720 chars — a readable repeated phrase, 300+ per
+   the phase) with a short follow-up → a fresh page load shows EXACTLY
+   ONE onboarding chip (the phase-103 opener semantics — the follow-up
+   never surfaces): computed ``white-space: nowrap`` /
+   ``overflow: hidden`` / ``text-overflow: ellipsis``, visually clipped
+   (``scrollWidth > clientWidth``), single line (``44 <=
+   offsetHeight <= 60`` — a one-line pill sits at the 44px
+   ``min-height`` floor, border-box; a wrapped two-liner is ~65px —
+   the chonk), and
+   ``title`` + ``aria-label`` == the FULL long text;
+2. **seed-contrast** — a fresh deployment's (seed) short chips have
+   ``title`` set AND NO ``aria-label`` (not truncated — the attribute
+   is absent by design; the ``textContent`` already carries the full
+   text);
+3. **counter-threshold** (A4) — ``#char-count`` hidden at boot; 100
+   chars typed → still hidden; exactly 3,500 chars (a dispatched
+   ``input`` event — ``locator.fill`` does this) → visible, text
+   ``3500/4000``, NO ``.is-max``;
+4. **paste-cap** (A3) — ``keyboard.insert_text`` of 6,500 chars (CDP
+   ``Input.insertText`` = the paste path — ``maxlength`` applies) →
+   the textarea holds EXACTLY 4,000 chars; the counter reads
+   ``4000/4000 — character limit`` + ``.is-max``; Send → NO 422 error
+   state → the mock answer streams to ``done`` (the brain bubble with
+   the ``MOCK_ANSWER_MARKER``) → the input is cleared and the counter
+   hidden again;
+5. **over-cap-guard** (A5) — ``page.evaluate`` sets
+   ``#message-input.value`` to 5,000 chars + dispatches an ``input``
+   event (the programmatic path ``maxlength`` cannot stop) → counter
+   ``5000/4000 — character limit`` + ``.is-max`` → Send → the error
+   banner shows the "4,000 characters" cap copy, NO bubble is
+   appended, the input STILL holds the 5,000 chars (kept for
+   trimming — never stale);
+6. **short-flow-regression** — a short question submits cleanly and
+   the counter NEVER becomes visible during the turn (a
+   MutationObserver flags any hidden→visible transition of
+   ``#char-count`` for the whole turn).
+
+The endpoint and the chat are authed (phase 79, ``require_user``), so
+every test signs in as admin first (``auth_helpers.login``). The
+onboarding chips are the phase-103 session openers read from
+``saved_chats`` — AND the send tests auto-save a row per turn — so the
+autouse fixture truncates that table before and after EVERY test,
+which keeps each test on an empty deployment (the seed state) and
+stops cross-test leakage.
+
+Run in isolation (DB must be up: ``podman compose up -d db``):
+
+    uv run pytest tests/e2e/test_chip_sizing_question_cap.py -v --no-cov
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import re
+from collections.abc import Iterator
+from pathlib import Path
+from threading import Thread
+from typing import Any
+
+import pytest
+from playwright.sync_api import Page, expect
+from sqlalchemy import text
+
+from app.config import Settings
+from app.db import SessionLocal
+from app.rag.importer import ImportSummary, import_sources
+from app.rag.llm import LLMClient
+from e2e.auth_helpers import login
+
+REPO = Path(__file__).resolve().parents[2]
+FIXTURES = REPO / "tests" / "fixtures" / "docs"
+MOCK_ANSWER_MARKER = "Deterministic mock answer for E2E"
+
+#: A 720-char opener (300+ per the phase) — a readable repeated phrase,
+#: far wider than the 46rem chat column, so the single-line pill MUST
+#: clip. The phase-103 opener semantics surface exactly this text as
+#: the chat's one onboarding chip.
+LONG_OPENER = "What are the correct arguments for " * 20 + "qwen on llama.cpp?"
+SHORT_FOLLOW_UP = "And which of the two needs the most VRAM?"
+
+#: A 6,500-char paste (4,000+ = over the cap) built from a word the
+#: fixture KB indexes ("kubernetes" — ``tests/fixtures/docs/homelab/
+#: kubernetes.md``): after ``maxlength`` trims it to EXACTLY 4,000 the
+#: question still carries the token, so the FTS leg of the hybrid
+#: retrieval hits and the mock answers the grounded (non-deflected)
+#: path with the marker — a clean turn at the cap, no 422.
+PASTE_QUESTION = "kubernetes " * 500
+PASTE_AT_CAP = PASTE_QUESTION[:4000]  # what maxlength=4000 keeps
+
+#: The over-cap programmatic fill (A5 — the chip one-tap path
+#: ``maxlength`` cannot stop): 5,000 chars, trimmed to 5,000.
+OVER_CAP_FILL = "x" * 5000
+
+#: The short-flow question (well under the 3,200 counter threshold and
+#: grounded in the fixture KB — the mock answers it with the marker).
+SHORT_QUESTION = "How do I deploy a new service on the homelab node?"
+
+
+@pytest.fixture(autouse=True)
+def clean_chats(db_ready: None) -> Iterator[None]:
+    """``saved_chats`` is the state the onboarding chips read (the
+    phase-103 openers) AND the send tests auto-save a row per turn —
+    truncate it before and after EVERY test so each starts from (and
+    leaves) an empty deployment (the seed-chip state), exactly the
+    phase-80/103 autouse pattern in test_suggestion_chips.py."""
+    with SessionLocal() as db:
+        db.execute(text("TRUNCATE saved_chats"))
+        db.commit()
+    yield
+    with SessionLocal() as db:
+        db.execute(text("TRUNCATE saved_chats"))
+        db.commit()
+
+
+async def _import_fixtures(mock_port: int) -> ImportSummary:
+    kwargs: dict[str, Any] = {"_env_file": None, "llm_base_url": f"http://127.0.0.1:{mock_port}/v1"}
+    settings = Settings(**kwargs)  # pyright: ignore[reportCallIssue]
+    return await import_sources([FIXTURES], LLMClient(settings))
+
+
+def _run_in_thread(coro: Any) -> Any:
+    """Run a coroutine on a worker thread (Playwright owns the test loop)."""
+    box: dict[str, Any] = {}
+
+    def runner() -> None:
+        try:
+            box["value"] = asyncio.run(coro)
+        except BaseException as e:  # noqa: BLE001 — re-raised on the test thread
+            box["error"] = e
+
+    t = Thread(target=runner)
+    t.start()
+    t.join()
+    if "error" in box:
+        raise box["error"]
+    return box["value"]
+
+
+def _seed_kb(mock_port: int) -> ImportSummary:
+    """Deterministic KB: truncate the KB tables, import the fixture
+    docs (needed by the send tests' grounded answers). ``saved_chats``
+    is the autouse fixture's job."""
+    with SessionLocal() as db:
+        db.execute(text("TRUNCATE chunks, documents, query_log"))
+        db.commit()
+    summary = _run_in_thread(_import_fixtures(mock_port))
+    assert summary is not None and summary.added == 13  # A9 formats (phase 47 added quadlet+j2)
+    return summary
+
+
+def _user(q: str) -> dict[str, Any]:
+    return {"who": "user", "text": q}
+
+
+def _brain(text: str = "Grounded mock brain reply.") -> dict[str, Any]:
+    return {"who": "brain", "text": text}
+
+
+def _save_chat(page: Page, app_url: str, messages: list[dict[str, Any]]) -> dict[str, Any]:
+    """Save one conversation as the signed-in admin (``POST /api/chats``)
+    and return the 201 body."""
+    r = page.request.post(
+        f"{app_url}/api/chats",
+        data=json.dumps({"messages": messages}),
+        headers={"Content-Type": "application/json"},
+        timeout=10_000,
+    )
+    assert r.status == 201, r.text
+    return r.json()
+
+
+def _chip(page: Page, index: int = 0) -> Any:
+    return page.locator("#suggestions .suggestion-chip").nth(index)
+
+
+def _wait_chat_booted(page: Page) -> None:
+    """Wait until app.js has FINISHED booting the chat page. The login
+    helper returns on the URL change (navigation commit) — the page's
+    module script may still be executing, and an ``input`` event
+    dispatched before its top-level listener registrations land on a
+    page whose listeners do not exist yet (the event is simply lost).
+    ``#view-chat.chat-booted`` is added two frames after the boot
+    settles (AFTER every top-level listener), so it is the "the app's
+    JS is live" sentinel (house pattern: test_mobile_chat_hamburger_
+    boot.py's post-login wait)."""
+    page.wait_for_function(
+        "() => document.getElementById('view-chat')?."
+        "classList.contains('chat-booted')",
+        timeout=15_000,
+    )
+    expect(page.locator("#view-chat")).to_have_class(re.compile(r"\bchat-booted\b"))
+
+
+def test_long_chip_is_single_line_ellipsized_with_full_text_tooltip(
+    page: Page, app_url: str, db_ready: None
+) -> None:
+    """A1/A2 core: a saved chat whose FIRST user question is 720 chars
+    (a readable repeated phrase) with a short follow-up → a fresh page
+    load shows EXACTLY ONE onboarding chip (the phase-103 opener
+    semantics — the follow-up never surfaces), and it is:
+
+    * ONE line — computed ``white-space: nowrap``, ``overflow: hidden``,
+      ``text-overflow: ellipsis``;
+    * visually clipped — ``scrollWidth > clientWidth`` (720 chars of
+      ~7px/char far exceeds the 46rem column);
+    * the pill, not the chonk — ``44 <= offsetHeight <= 60`` (a
+      one-line pill sits at the 44px ``min-height`` floor, border-box
+      — the global ``* { box-sizing: border-box }`` makes the 44px
+      the OUTER height; a wrapped two-liner is ~65px);
+    * the FULL text one hover away — ``title`` == the full long text;
+    * the clipped-case accessible name — ``aria-label`` == the full
+      long text.
+    """
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+
+    # ONE multi-turn chat: the LONG opener + one short follow-up
+    # (the follow-up must never surface — phase-103 semantics).
+    _save_chat(
+        page,
+        app_url,
+        [
+            _user(LONG_OPENER), _brain(),
+            _user(SHORT_FOLLOW_UP), _brain(),
+        ],
+    )
+
+    # A FRESH page load (a new boot fetch): EXACTLY ONE chip — the
+    # opener, and nothing else.
+    page.goto(app_url + "/")
+    chips = page.locator("#suggestions .suggestion-chip")
+    expect(chips.first).to_be_visible(timeout=30_000)
+    expect(chips).to_have_count(1, timeout=15_000)
+    # The chip carries the EXACT full opener in the DOM (the clipping
+    # is visual — textContent is never truncated).
+    assert chips.first.inner_text().strip() == LONG_OPENER
+
+    chip = chips.first
+    # A1 — the single-line computed contract (the phase-104 CSS).
+    style = chip.evaluate("el => getComputedStyle(el)")
+    assert style["whiteSpace"] == "nowrap", "the chip must never wrap (owner A1)"
+    assert style["overflow"] in {"hidden", "hidden hidden"}, (
+        "`overflow: hidden` is what zeroes the flex auto-min so the "
+        "ellipsis clip at the row edge binds"
+    )
+    assert style["textOverflow"] == "ellipsis", "the clipped text shows the ellipsis"
+
+    # Clipped + single line: the text overflows the pill (ellipsis) but
+    # the pill itself stays ONE 44px-min line tall — never the chonk.
+    # offsetHeight (border-box): the global `* { box-sizing: border-box }`
+    # makes the 44px `min-height` the OUTER floor — a one-line pill is
+    # exactly 44px, a wrapped two-liner is ~65px (63px content + 2px
+    # border), so the 44..60 band is single-line only.
+    dims = chip.evaluate(
+        "el => ({ sw: el.scrollWidth, cw: el.clientWidth, oh: el.offsetHeight })"
+    )
+    assert dims["sw"] > dims["cw"], (
+        "a 720-char question must be visually clipped in the 46rem column"
+    )
+    assert 44 <= dims["oh"] <= 60, (
+        f"one-line pill: 44px min-height floor (border-box), a wrapped "
+        f"two-liner is ~65px — got {dims['oh']}px"
+    )
+
+    # A2 — the full text is one hover away (the native tooltip), and
+    # the accessible name of the CLIPPED chip is the full text.
+    assert chip.get_attribute("title") == LONG_OPENER, (
+        "the title tooltip must carry the FULL question (owner A2)"
+    )
+    assert chip.get_attribute("aria-label") == LONG_OPENER, (
+        "a clipped chip's aria-label must be the full text (the "
+        "source-chip pattern)"
+    )
+
+
+def test_short_seed_chip_has_tooltip_but_no_aria_label(
+    page: Page, app_url: str, db_ready: None
+) -> None:
+    """The contrast pin: a fresh deployment (seed chips — the autouse
+    fixture guarantees the empty state) renders four SHORT seed chips,
+    none of them clipped in the 1280px desktop column. Every chip still
+    has ``title`` = its full text (A2 is universal), but NO ``aria-label``
+    — the attribute is absent by design when the visible text is not
+    clipped (the ``textContent`` already carries the full text)."""
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+
+    chips = page.locator("#suggestions .suggestion-chip")
+    expect(chips.first).to_be_visible(timeout=30_000)
+    expect(chips).to_have_count(4, timeout=15_000)  # the built-in seed (4 entries)
+
+    for i in range(4):
+        chip = chips.nth(i)
+        full_text = chip.inner_text().strip()
+        assert full_text, "every seed chip needs non-empty label text"
+        # Not clipped: the pill fits its text in the desktop column.
+        dims = chip.evaluate(
+            "el => ({ sw: el.scrollWidth, cw: el.clientWidth, h: el.clientHeight })"
+        )
+        assert dims["sw"] <= dims["cw"], f"seed chip {i!r} must not be clipped"
+        # A2 universal: the tooltip is set on EVERY chip…
+        assert chip.get_attribute("title") == full_text
+        # …but the accessible name is added ONLY when clipped.
+        assert chip.get_attribute("aria-label") is None, (
+            f"an unclipped chip must NOT carry an aria-label (chip {i!r}) — "
+            "the textContent already reads the full text"
+        )
+
+
+def test_counter_hidden_below_threshold_and_visible_above(
+    page: Page, app_url: str, db_ready: None
+) -> None:
+    """A4: the counter is hidden at boot, stays hidden at 100 chars
+    (below the 3,200 = 80% threshold — no noise on normal use), and at
+    exactly 3,500 chars (a dispatched ``input`` event — ``locator.fill``
+    fires it) it is visible reading ``3500/4000`` with NO ``.is-max``
+    (that state is for at/over the cap)."""
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+    _wait_chat_booted(page)
+
+    counter = page.locator("#char-count")
+    expect(counter).to_be_hidden()  # boot state (hidden by default in the HTML)
+
+    page.fill("#message-input", "a" * 100)
+    expect(counter).to_be_hidden()  # 100 < 3,200 — still no noise
+
+    page.fill("#message-input", "b" * 3500)
+    expect(counter).to_be_visible(timeout=5_000)
+    expect(counter).to_have_text("3500/4000")
+    expect(counter).not_to_have_class(re.compile("is-max"))
+
+
+def test_paste_path_hard_caps_at_the_cap_and_sends(
+    page: Page, app_url: str, mock_llm: int, db_ready: None
+) -> None:
+    """A3: the paste path (CDP ``Input.insertText`` — the browser
+    applies ``maxlength`` to it, like a paste) hard-caps at EXACTLY
+    4,000: a 6,500-char paste lands as 4,000, the counter reads
+    ``4000/4000 — character limit`` + ``.is-max``, and submitting a
+    4,000-char question PASSES the server cap (no 422 error state) —
+    the mock answer streams to ``done`` and the turn clears the input
+    + the counter."""
+    _seed_kb(mock_llm)
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+    _wait_chat_booted(page)
+
+    counter = page.locator("#char-count")
+    input_el = page.locator("#message-input")
+    expect(counter).to_be_hidden()
+
+    # The paste path: one CDP insert of 6,500 chars — maxlength trims
+    # the OVERFLOW, the textarea keeps EXACTLY the first 4,000.
+    input_el.click()
+    page.keyboard.insert_text(PASTE_QUESTION)
+    expect(input_el).to_have_value(PASTE_AT_CAP)  # EXACTLY 4,000 chars
+
+    # At the cap: the honest reading + the B3 state (copy + err color).
+    expect(counter).to_be_visible(timeout=5_000)
+    expect(counter).to_have_text("4000/4000 — character limit")
+    expect(counter).to_have_class(re.compile("is-max"))
+
+    # Submit at the cap: the 4,000-char question passes the server cap
+    # (no 422 error state — a 422 would raise the error banner).
+    banner = page.locator("#kb-banner")
+    expect(banner).to_be_hidden()
+    page.click("#send-btn")
+
+    # A clean turn: the user bubble is the exact 4,000-char question and
+    # the grounded mock answer streams to done (the KB carries
+    # "kubernetes" — the FTS leg hits, non-deflected).
+    expect(page.locator(".msg.user .bubble")).to_have_count(1, timeout=30_000)
+    expect(page.locator(".msg.user .bubble")).to_have_text(PASTE_AT_CAP)
+    brain = page.locator(".msg.brain .bubble").first
+    expect(brain).to_contain_text(MOCK_ANSWER_MARKER, timeout=30_000)
+    expect(page.locator(".msg.brain.is-deflected")).to_have_count(0)
+    expect(banner).to_be_hidden()  # NO 422 error state at the cap
+
+    # Never stale: the turn cleared the input AND the counter, and the
+    # send button recovered.
+    expect(input_el).to_have_value("")
+    expect(counter).to_be_hidden()
+    expect(page.locator("#send-btn")).to_be_enabled()
+
+
+def test_over_cap_programmatic_fill_hits_the_guard(
+    page: Page, app_url: str, db_ready: None
+) -> None:
+    """A5: the one path ``maxlength`` cannot stop — a programmatic
+    ``value`` assignment (the chip one-tap fill). A 5,000-char fill +
+    a dispatched ``input`` event → the counter reads the honest
+    ``5000/4000 — character limit`` + ``.is-max``; Send → the over-cap
+    guard fires: the error banner shows the "4,000 characters" cap
+    copy, NO turn (no user or brain bubble), and the input STILL holds
+    the 5,000 chars (kept for trimming — never stale, PLAN §7.4)."""
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+    _wait_chat_booted(page)
+
+    counter = page.locator("#char-count")
+    expect(counter).to_be_hidden()
+
+    # The programmatic path: assign + dispatch (maxlength never sees it).
+    page.evaluate(
+        """(value) => {
+            const el = document.querySelector("#message-input");
+            el.value = value;
+            el.dispatchEvent(new Event("input", { bubbles: true }));
+        }""",
+        OVER_CAP_FILL,
+    )
+    expect(counter).to_be_visible(timeout=5_000)
+    expect(counter).to_have_text("5000/4000 — character limit")
+    expect(counter).to_have_class(re.compile("is-max"))
+
+    # Send over the cap: the guard refuses BEFORE the request.
+    page.click("#send-btn")
+    banner = page.locator("#kb-banner")
+    expect(banner).to_be_visible(timeout=5_000)
+    expect(banner).to_have_attribute("role", "alert")
+    expect(page.locator("#kb-banner-text")).to_contain_text("4,000 characters")
+
+    # NO turn: nothing was appended, and the input keeps the text the
+    # user trims (a cleared input would be stale — PLAN §7.4).
+    expect(page.locator(".msg.user .bubble")).to_have_count(0)
+    expect(page.locator(".msg.brain .bubble")).to_have_count(0)
+    expect(page.locator("#message-input")).to_have_value(OVER_CAP_FILL)
+    # The counter (which mirrors the untouched input) and the send
+    # button (never mid-turn) are consistent with the kept text.
+    expect(counter).to_have_text("5000/4000 — character limit")
+    expect(page.locator("#send-btn")).to_be_enabled()
+
+
+def test_short_flow_never_shows_the_counter(
+    page: Page, app_url: str, mock_llm: int, db_ready: None
+) -> None:
+    """Short-flow regression: a 49-char question submits cleanly (the
+    grounded mock answer lands) and ``#char-count`` NEVER becomes
+    visible during the whole turn — a MutationObserver flags ANY
+    hidden→visible transition of the counter (with ``oldValue``, so
+    even a same-task set/unset is caught)."""
+    _seed_kb(mock_llm)
+    page.set_default_timeout(30_000)
+    login(page, app_url, next="/")
+    _wait_chat_booted(page)
+
+    # Arm the never-visible tripwire for the rest of this page's life.
+    page.evaluate(
+        """() => {
+            window.__char_count_ever_visible = false;
+            const el = document.querySelector("#char-count");
+            new MutationObserver((records) => {
+                for (const r of records) {
+                    if (r.attributeName === "hidden" && r.oldValue === "true") {
+                        window.__char_count_ever_visible = true;
+                    }
+                }
+            }).observe(el, {
+                attributes: true,
+                attributeFilter: ["hidden"],
+                attributeOldValue: true,
+            });
+        }"""
+    )
+
+    counter = page.locator("#char-count")
+    expect(counter).to_be_hidden()
+    page.fill("#message-input", SHORT_QUESTION)  # 49 chars << 3,200
+    expect(counter).to_be_hidden()
+
+    page.click("#send-btn")
+    expect(page.locator(".msg.user .bubble")).to_have_count(1, timeout=30_000)
+    expect(page.locator(".msg.user .bubble")).to_have_text(SHORT_QUESTION)
+    brain = page.locator(".msg.brain .bubble").first
+    expect(brain).to_contain_text(MOCK_ANSWER_MARKER, timeout=30_000)
+    expect(page.locator("#message-input")).to_have_value("")
+    expect(page.locator("#send-btn")).to_be_enabled()
+
+    # The whole turn: the counter never once became visible.
+    assert page.evaluate("() => window.__char_count_ever_visible") is False, (
+        "a short question must never surface the counter — not at the "
+        "keystrokes, not at the send, not during the turn"
+    )
diff --git a/tests/e2e/test_pinned_composer.py b/tests/e2e/test_pinned_composer.py
index ea5697a..3b828d5 100644
--- a/tests/e2e/test_pinned_composer.py
+++ b/tests/e2e/test_pinned_composer.py
@@ -186,6 +186,38 @@ def scroll_state(page: Page) -> dict[str, float]:
     )
 
 
+def wait_reveal_scroll_done(page: Page, target: int, timeout: int = 10_000) -> None:
+    """Block until the submit's ``scrollReveal`` smooth scroll has ARRIVED
+    at its target (the document bottom captured right after the click,
+    ``target`` — the first streamed delta lands tens of ms later and
+    would inflate a later read). Needed before a test scroll: Chromium's
+    smooth-scroll duration is distance- and environment-dependent (roughly
+    half a second to a second for a full page), and a ``window.scrollTo``
+    issued mid-animation does not settle the measured position (the
+    animation keeps moving the page after the test's scroll, so a
+    position read lands mid-flight, e.g. ``y=71`` instead of ``0``).
+
+    The wait is an ARRIVAL wait (``scrollY`` reached the target bottom),
+    not a "looks still" wait: mid-animation frame stalls under load make
+    a short stillness window pass while the animation is still pending
+    (observed: a 120 ms window passed mid-scroll; 800 ms is the fallback
+    for the vanishingly-rare late-target read, where the arrival
+    threshold is inflated by an already-landed delta). Phase 104 (task
+    03, 2026-09-12): pinned deterministically when the phase-104 E2E
+    regression exposed the race (it reproduces on the pre-phase-104 tree
+    — a pre-existing flake, not a phase-104 regression)."""
+    page.wait_for_function(
+        """(target) => {
+          if (window.scrollY >= target - window.innerHeight - 2) return true;
+          const y0 = window.scrollY;
+          return new Promise((resolve) =>
+            setTimeout(() => resolve(window.scrollY === y0), 800));
+        }""",
+        arg=target,
+        timeout=timeout,
+    )
+
+
 def wait_settled(page: Page, timeout: int = 30_000) -> None:
     """The turn is over: the label is back to "Send" (phase 48 — the
     in-flight state is the enabled Stop control, so the label carries the
@@ -194,11 +226,18 @@ def wait_settled(page: Page, timeout: int = 30_000) -> None:
     expect(page.locator("#send-label")).to_have_text("Send", timeout=timeout)
 
 
-def submit(page: Page, question: str) -> None:
-    """Submit through the composer (the real-user flow)."""
+def submit(page: Page, question: str) -> int:
+    """Submit through the composer (the real-user flow). Returns the
+    document ``scrollHeight`` read right after the click — the submit's
+    ``scrollReveal`` target (``window.scrollTo({top: scrollHeight})``
+    runs in the click's own event dispatch, and the first streamed delta
+    lands tens of ms later and would inflate a later read)."""
     page.fill("#message-input", question)
     page.click("#send-btn")
     expect(page.locator(".msg.user .bubble").last).to_contain_text(question)
+    # The bubble exists by now and the first streamed delta is still tens
+    # of ms away — the read is the scrollReveal's own target.
+    return page.evaluate("() => document.documentElement.scrollHeight")
 
 
 def build_conversation(page: Page, n: int = 6) -> None:
@@ -371,7 +410,7 @@ def test_stop_is_reachable_from_scrolled_up(
     # SECOND answer bubble exists and carries deltas (`.last` alone would
     # still resolve to turn 1's finished bubble: the typing indicator is
     # `.bubble.typing` and excluded from ANSWER).
-    submit(page, LONG_QUESTION)
+    reveal_target = submit(page, LONG_QUESTION)
     page.wait_for_function(
         "() => { const els = document.querySelectorAll('.msg.brain .bubble:not(.typing)');"
         " return els.length >= 2 && els[els.length - 1].innerText.length > 80; }",
@@ -380,6 +419,12 @@ def test_stop_is_reachable_from_scrolled_up(
     answer = page.locator(ANSWER).nth(1)
     expect(page.locator("#send-label")).to_have_text("Stop", timeout=10_000)
 
+    # The submit's own smooth scrollReveal (to the document bottom) must
+    # ARRIVE before the test scrolls up — mid-animation, the browser
+    # keeps moving the page after the test's scrollTo and the position
+    # read lands mid-flight (the race wait_reveal_scroll_done documents).
+    wait_reveal_scroll_done(page, reveal_target)
+
     # The user scrolls UP to read earlier content. Phase 42 leaves them
     # there — the app never follows the stream.
     page.evaluate("() => window.scrollTo(0, 0)")
diff --git a/tests/e2e/test_responsive_polish.py b/tests/e2e/test_responsive_polish.py
index 5defd76..632648c 100644
--- a/tests/e2e/test_responsive_polish.py
+++ b/tests/e2e/test_responsive_polish.py
@@ -130,8 +130,24 @@ def _assert_no_doc_overflow(page: Page, label: str) -> None:
 
 
 def _tab_outline_walk(page: Page, max_tabs: int = 60) -> list[dict[str, str]]:
-    """Real keyboard Tab walk; returns each focused element's outline."""
-    first_key: str | None = None
+    """Real keyboard Tab walk; returns each focused element's outline.
+
+    The walk covers a FULL focus cycle (it starts wherever Chromium
+    resumes after the skip-link check — mid-document, right after
+    #main — and ends when focus wraps back to the FIRST element it
+    visited). The wrap is detected by TRUE element identity: each
+    newly focused element is stamped with a unique ``data-tabwalk-id``
+    and the cycle ends when a STAMPED element is focused again.
+    Phase 104 (task 04, 2026-09-12): the pre-104 heuristic keyed on
+    the first 24 chars of the element's text — two chips sharing that
+    prefix (e.g. the two identical "How is my Kubernetes cluster set
+    up? write a long answer" opener chips left in ``saved_chats`` by
+    test_pinned_composer.py, which runs suites in sequence on the
+    shared e2e DB) collided and cut the walk short at two entries,
+    flaking ``len(seen) >= 3``. Verified pre-existing on the
+    pre-phase-104 tree (the mid-document Tab start is Chromium
+    behavior, not a phase-104 change)."""
+    first_id: str | None = None
     seen: list[dict[str, str]] = []
     for _ in range(max_tabs):
         page.keyboard.press("Tab")
@@ -142,7 +158,13 @@ def _tab_outline_walk(page: Page, max_tabs: int = 60) -> list[dict[str, str]]:
                 const cls = String(el.className).split(" ")[0];
                 const label = (el.getAttribute("aria-label")
                     || el.textContent || "").trim().slice(0, 24);
+                let id = el.getAttribute("data-tabwalk-id");
+                if (!id) {
+                    id = "tw" + ((window.__twSeq = (window.__twSeq || 0) + 1));
+                    el.setAttribute("data-tabwalk-id", id);
+                }
                 return {
+                    id: id,
                     key: el.tagName + "#" + (el.id || "") + "." + cls + ":" + label,
                     outline_style: cs.outlineStyle,
                     outline_width: cs.outlineWidth,
@@ -151,11 +173,11 @@ def _tab_outline_walk(page: Page, max_tabs: int = 60) -> list[dict[str, str]]:
         )
         if info["key"].startswith("BODY"):
             continue  # focus has not entered the document yet
-        if first_key is None:
-            first_key = info["key"]
+        if first_id is None:
+            first_id = info["id"]
         seen.append(info)
-        if len(seen) > 1 and info["key"] == first_key:
-            break  # wrapped back to the first focusable
+        if len(seen) > 1 and info["id"] == first_id:
+            break  # wrapped back to the first focusable (same ELEMENT)
     return seen
 
 
diff --git a/tests/e2e/test_suggestion_chips.py b/tests/e2e/test_suggestion_chips.py
index 6d9af7a..827b9cc 100644
--- a/tests/e2e/test_suggestion_chips.py
+++ b/tests/e2e/test_suggestion_chips.py
@@ -181,7 +181,12 @@ def _save_chat(page: Page, app_url: str, messages: list[dict[str, Any]]) -> dict
 def _api_suggestions(page: Page, app_url: str) -> list[str]:
     # Phase 79: the endpoint is require_user-gated — the request rides
     # the page's signed-in context (each test signs in above).
-    r = page.request.get(f"{app_url}/api/suggestions", timeout=10)
+    # Phase 104 (task 03, 2026-09-12): the original 10 ms timeout sat
+    # BELOW a normal localhost round trip (~12 ms steady-state) and
+    # flaked the whole suite on a loaded box — the sibling _save_chat
+    # helper's 10 s is the house pattern; the server is up by the time
+    # this probe runs (the chips it mirrors already rendered).
+    r = page.request.get(f"{app_url}/api/suggestions", timeout=10_000)
     assert r.status == 200, r.text
     return r.json()["suggestions"]
 
diff --git a/tests/unit/test_chip_sizing_question_cap.py b/tests/unit/test_chip_sizing_question_cap.py
new file mode 100644
index 0000000..dcec31f
--- /dev/null
+++ b/tests/unit/test_chip_sizing_question_cap.py
@@ -0,0 +1,550 @@
+"""Unit: the phase-104 chip-sizing contract in the static frontend
+(owner request 2026-09-12 — "tweak the size of the suggestion chips on
+the chat page. Some users submit truly massive queries and the 'chips'
+become more like 'chonks'. Hovering over the chips should still show
+the full message … put a character cap on the chat submission box").
+
+The browser geometry (the measured clientHeight / scrollWidth / the
+title + aria-label / the counter / the guard) is E2E-gated by the
+phase's dedicated suite (tests/e2e/test_chip_sizing_question_cap.py,
+task 04); this module pins the static sources the contract stands on,
+in the house source-pin pattern (the test_pinned_composer.py
+``_rule`` style).
+
+Task 01 — single-line chips, never chonks (owner A1): a suggestion
+chip is ONE line at every viewport width — long text ellipsizes at the
+row edge instead of wrapping the pill into a multi-line "chonk". The
+pin is the CSS contract:
+
+* the base ``.suggestion-chip`` rule carries the single-line/ellipsis
+  set — ``white-space: nowrap``, ``overflow: hidden`` (zeroes the flex
+  item's automatic minimum size so ``max-width: 100%`` actually binds),
+  ``text-overflow: ellipsis``, ``max-width: 100%`` — plus
+  ``min-width: 0``, and KEEPS the pill it always was (the 44px
+  ``min-height`` floor, the 999px radius);
+* the phase-07 override ``.maybe-try .suggestion-chip``
+  (``min-width: 0; max-width: 100%``) is fully subsumed by the base
+  rule and DELETED — zero occurrences of the selector remain in the
+  file (its phase-07 provenance was folded into the base rule's
+  comment, so the history lives with the contract); the ``.maybe-try``
+  GROUP rule itself stays (the deflection row still wraps at every
+  width — only the chip override is gone);
+* the ≤640px block keeps its phase-07 contract untouched: the
+  ``.suggestions`` row is the single horizontal-scroll track
+  (``flex-wrap: nowrap; overflow-x: auto``) and
+  ``.suggestion-chip { flex: 0 0 auto; }`` — with the base rule's
+  ``max-width: 100%`` a long chip now clips at the VISIBLE width while
+  the row scrolls, instead of letting the chip outgrow the viewport.
+
+Task 02 — the full text is always one hover away (owner A2): the
+single-line pill from task 01 CLIPS long questions, so every chip the
+shared ``renderChips`` component builds (onboarding row AND "Maybe try"
+row) carries:
+
+* ``btn.title = text`` — the native tooltip set to the FULL text,
+  always (the house source-chip precedent, ``chip.title = label``);
+* the truncation-aware accessible name — ONLY when the visible text is
+  clipped (``btn.scrollWidth > btn.clientWidth``) does the chip get
+  ``aria-label`` = the full text (the source-chip pattern); when the
+  pill is not clipped the attribute stays absent, because
+  ``textContent`` already reads the full text to screen readers.
+
+Task 03 extends this module with the composer cap pins (``maxlength`` /
+the counter / the guard / the single-source cross-file pin).
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+from app.schemas import ChatRequest
+
+FRONTEND = Path(__file__).resolve().parents[2] / "frontend"
+STYLES_CSS = FRONTEND / "assets" / "styles.css"
+APP_JS = FRONTEND / "assets" / "app.js"
+INDEX_HTML = FRONTEND / "index.html"
+
+
+def _css() -> str:
+    assert STYLES_CSS.is_file(), f"missing {STYLES_CSS}"
+    return STYLES_CSS.read_text(encoding="utf-8")
+
+
+def _app_js() -> str:
+    assert APP_JS.is_file(), f"missing {APP_JS}"
+    return APP_JS.read_text(encoding="utf-8")
+
+
+def _index_html() -> str:
+    assert INDEX_HTML.is_file(), f"missing {INDEX_HTML}"
+    return INDEX_HTML.read_text(encoding="utf-8")
+
+
+def _function_body(js: str, header: str) -> str:
+    """The full text of the function whose header is ``header`` — from
+    the header to its brace-matched closing ``}``. The naive brace count
+    is safe for the pinned functions: their template literals carry
+    balanced ``${…}`` pairs and no string literal holds a stray brace."""
+    start = js.index(header)
+    body_open = js.index("{", start)
+    depth = 0
+    for j in range(body_open, len(js)):
+        if js[j] == "{":
+            depth += 1
+        elif js[j] == "}":
+            depth -= 1
+            if depth == 0:
+                return js[start : j + 1]
+    raise AssertionError(f"unbalanced braces in {header!r}")
+
+
+def _single_line_rule(css: str, selector: str) -> str:
+    """The body of a ONE-LINE rule (``selector { … }`` on a single
+    line) — the .char-count rules are written house one-liners."""
+    block = re.search(rf"^{re.escape(selector)} \{{([^}}]*)\}}", css, re.MULTILINE)
+    assert block, f"styles.css must carry a `{selector} {{ … }}` rule"
+    return block.group(1)
+
+
+def _message_input_tag(html: str) -> str:
+    """The FULL opening tag of the ``#message-input`` textarea (the
+    composer's — index.html carries other textareas too)."""
+    tag = re.search(r'<textarea\b[^>]*id="message-input"[^>]*>', html)
+    assert tag, "index.html must carry the #message-input textarea"
+    return tag.group(0)
+
+
+def _schema_question_cap() -> int:
+    """The server cap the UI mirrors — ``ChatRequest.message``'s
+    ``max_length`` (the single conceptual source of the 4,000)."""
+    for meta in ChatRequest.model_fields["message"].metadata:
+        if getattr(meta, "max_length", None):
+            return int(meta.max_length)
+    raise AssertionError("ChatRequest.message must keep its max_length")
+
+
+def _render_chips_body(js: str) -> str:
+    """The full text of the ``renderChips`` function — from its
+    signature to its brace-matched closing ``}`` (brace counting, so
+    the nested click-handler braces are exact). The body's opening
+    brace is the LAST ``{`` on the signature line — the ``{ onSelect }
+    = {}`` destructuring pair sits before it."""
+    start = js.index("function renderChips(")
+    line_end = js.index("\n", start)
+    body_open = js.rindex("{", start, line_end)
+    depth = 0
+    for j in range(body_open, len(js)):
+        if js[j] == "{":
+            depth += 1
+        elif js[j] == "}":
+            depth -= 1
+            if depth == 0:
+                return js[start : j + 1]
+    raise AssertionError("unbalanced braces in renderChips")
+
+
+def _rule(css: str, selector: str) -> str:
+    """The body of the rule whose selector line is exactly `selector`
+    (multi-line block) — same slicing style as test_pinned_composer.py."""
+    block = re.search(rf"^{re.escape(selector)} \{{\n([\s\S]*?)\n\}}", css, re.MULTILINE)
+    assert block, f"styles.css must carry a `{selector} {{ … }}` rule"
+    return block.group(1)
+
+
+def _mobile_block(css: str) -> str:
+    """The ≤640px media query body (the phase-07 responsive block)."""
+    mobile = re.search(r"@media \(max-width: 640px\) \{([\s\S]*?)\n\}\n", css)
+    assert mobile, "the mobile media query must exist"
+    return mobile.group(1)
+
+
+# ---------- task 01: the single-line chip contract ----------
+
+
+def test_base_chip_rule_is_single_line_and_ellipsized() -> None:
+    """The base ``.suggestion-chip`` rule carries the phase-104
+    single-line/ellipsis contract (owner A1) AND keeps the pill it
+    always was — a chip is a one-line 44px pill at every width, never
+    a wrapped multi-line block:
+
+    * ``white-space: nowrap`` — the text never wraps inside the pill;
+    * ``overflow: hidden`` (≠ visible) — zeroes the flex item's
+      automatic minimum size, which is what makes ``max-width: 100%``
+      bind at all;
+    * ``text-overflow: ellipsis`` + ``max-width: 100%`` — the text
+      clips at the row edge (desktop wrap row: the chat column;
+      ≤640px row: the visible width) instead of the pill outgrowing it;
+    * ``min-width: 0`` — the same auto-min zeroing, pinned for clarity.
+    """
+    body = _rule(_css(), ".suggestion-chip")
+    assert "white-space: nowrap;" in body, (
+        "the chip must never wrap — a 400-char question may not turn the "
+        "pill into a multi-line 'chonk' (owner A1)"
+    )
+    assert "overflow: hidden;" in body, (
+        "`overflow: hidden` zeroes the flex item's automatic minimum size "
+        "so `max-width: 100%` actually binds (visible overflow would let "
+        "the pill push past the row edge)"
+    )
+    assert "text-overflow: ellipsis;" in body, (
+        "the clipped text must show the ellipsis — the full text stays "
+        "one hover away (the title tooltip, task 02)"
+    )
+    assert "max-width: 100%;" in body, (
+        "the chip clips at the row edge — 100% of the row it sits in"
+    )
+    assert "min-width: 0;" in body, (
+        "the flex item's automatic minimum size must be zero — the "
+        "phase-07 shrink allowance, now on every chip row"
+    )
+    # The pill contract survives the sizing change (the 44px touch
+    # floor + the round pill — the E2E's 44 <= clientHeight <= 60 pin
+    # stands on this min-height).
+    assert "min-height: 44px;" in body, (
+        "the chip keeps its ≥44px touch target (WCAG 2.1 AA, phase 07)"
+    )
+    assert "border-radius: 999px;" in body, "the chip stays the round pill"
+
+
+def test_subsumed_maybe_try_chip_override_is_gone() -> None:
+    """The phase-07 ``.maybe-try .suggestion-chip { min-width: 0;
+    max-width: 100%; }`` override is fully subsumed by the new base
+    rule and DELETED — zero occurrences of the selector remain anywhere
+    in the file (comments included: the provenance was folded into the
+    base rule's comment WITHOUT the selector string, so this pin
+    doubles as the "no second cap / no drift" guard). The ``.maybe-try``
+    GROUP rule itself stays exactly once — the deflection row still
+    wraps at every width (phase 04), only the chip override is gone."""
+    css = _css()
+    assert css.count(".maybe-try .suggestion-chip") == 0, (
+        "the subsumed phase-07 override must be deleted — the base rule "
+        "carries min-width:0 + max-width:100% now, and a second cap "
+        "would be dead weight / a drift hazard"
+    )
+    assert css.count(".maybe-try {") == 1, (
+        "the .maybe-try group rule stays (the deflection row's wrap "
+        "contract) — only the chip override was removed"
+    )
+
+
+def test_mobile_chip_row_keeps_the_phase_07_track() -> None:
+    """The ≤640px block is UNTOUCHED by task 01: the ``.suggestions``
+    row keeps the phase-07 single horizontal-scroll track
+    (``flex-wrap: nowrap; overflow-x: auto``) and
+    ``.suggestion-chip { flex: 0 0 auto; }``. With the base rule's
+    ``max-width: 100%`` + ``overflow: hidden``, a long chip on the
+    phone clips at the VISIBLE width and the row scrolls — the
+    phase-07 overflow contract now lives in the base rule (task 01
+    folded the provenance there), so nothing in the mobile block may
+    have changed."""
+    mobile = _mobile_block(_css())
+    mobile_chip = re.search(r"\.suggestion-chip \{([^}]*)\}", mobile)
+    assert mobile_chip, "the ≤640px block must keep the .suggestion-chip rule"
+    assert ".suggestions { flex-wrap: nowrap; overflow-x: auto;" in mobile, (
+        "the mobile row keeps the phase-07 single horizontal-scroll "
+        "track (nowrap + overflow-x: auto)"
+    )
+    assert mobile_chip.group(1).strip() == "flex: 0 0 auto;", (
+        "the mobile chip must stay EXACTLY the phase-07 fixed flex item "
+        "(flex: 0 0 auto) on the scroll track — the chip does not shrink "
+        "below its content on the phone, the ROW scrolls; the long-chip "
+        "clip is the base rule's max-width: 100%, not a mobile-local one"
+    )
+
+
+# ---------- task 02: the full-text tooltip + accessible name ----------
+
+
+def test_render_chips_sets_full_text_title_on_every_chip() -> None:
+    """Every chip ``renderChips`` builds (onboarding row AND "Maybe try"
+    row — one shared component) carries ``title`` = the FULL text
+    (owner A2): the task-01 pill clips a long question to one
+    ellipsized line, and the native hover tooltip is the way the user
+    reads the rest. The pin is the house source-chip precedent
+    (``chip.title = label``) — and it is set to the SAME ``text``
+    variable ``textContent`` gets, so the tooltip can never drift
+    short of the full question."""
+    body = _render_chips_body(_app_js())
+    assert "btn.textContent = text;" in body, (
+        "the chip's visible text is the (trimmed) full question"
+    )
+    assert "btn.title = text;" in body, (
+        "every chip must carry the full text as its native title "
+        "tooltip — hovering is the contract for the clipped text "
+        "(owner A2: 'Hovering over the chips should still show the "
+        "full message')"
+    )
+
+
+def test_render_chips_sets_aria_label_only_when_clipped() -> None:
+    """The accessible name is the full text ONLY when the visible text
+    is actually clipped (the source-chip truncation pattern, the
+    ``chip.scrollWidth > chip.clientWidth`` → ``aria-label`` loop): a
+    chip whose text fits carries NO ``aria-label`` — its ``textContent``
+    already reads the full text to screen readers — while the clipped
+    chip (the ellipsized one) announces the full question. The pin is
+    the guard plus the ``setAttribute`` call inside it."""
+    body = _render_chips_body(_app_js())
+    guard = re.search(
+        r'if \(btn\.scrollWidth > btn\.clientWidth\) '
+        r'btn\.setAttribute\("aria-label", text\);',
+        body,
+    )
+    assert guard, (
+        "the clipped chip must set aria-label to the FULL text under "
+        "the scrollWidth > clientWidth guard (the source-chip pattern) "
+        "— a clipped chip's accessible name must be the full question, "
+        "and an unclipped chip must stay attribute-free"
+    )
+
+
+# ---------- task 03: the visible 4,000-char question cap ----------
+
+
+def test_message_input_carries_the_4000_maxlength() -> None:
+    """The composer's textarea hard-caps the input path (typing AND
+    paste — the browser enforces maxlength on both, E2E task 04) at
+    EXACTLY the server's cap, so a user can never meet the 422 blind
+    through the input path (owner A3). The provenance comment lives
+    with the attribute — the house pattern (the theme inputs' "
+    maxlength=300 mirrors the server's 300-char")."""
+    html = _index_html()
+    tag = _message_input_tag(html)
+    assert 'maxlength="4000"' in tag, (
+        "#message-input must carry maxlength=4000 — the server already "
+        "rejects >4,000 (ChatRequest.message) and the counter makes it "
+        "visible; without it the input path 422s with zero feedback"
+    )
+    assert "maxlength=4000 mirrors ChatRequest.message max_length=4000" in html, (
+        "the provenance comment must live with the attribute (house "
+        "pattern) — the schema is the source of truth the HTML mirrors"
+    )
+
+
+def test_char_count_sits_in_chat_bottom_above_the_composer() -> None:
+    """The counter element is a child of the .chat-bottom sticky unit —
+    between the actions row and the composer form (source order) and
+    ``hidden`` by default (it only appears from 80% of the cap). A
+    hidden ``<p>`` adds zero height, so the pinned-cluster geometry
+    (tests/unit/test_pinned_composer.py, the .chat-bottom-last-child
+    pin) is untouched by the new child."""
+    html = _index_html()
+    unit = re.search(r'<div class="chat-bottom">([\s\S]*?)</form>\s*</div>', html)
+    assert unit, "the .chat-bottom sticky unit must exist (phase 65)"
+    inner = unit.group(1)
+    el = re.search(r'<p class="char-count" id="char-count"[^>]*></p>', inner)
+    assert el, "the #char-count counter element must exist in .chat-bottom"
+    assert "hidden" in el.group(0), (
+        "the counter ships HIDDEN — it is feedback near the cap, not a "
+        "permanent chrome line (owner A4: no noise on normal use)"
+    )
+    # Source order: AFTER the actions row closes, BEFORE the composer —
+    # the counter sits between the row and the form inside the unit.
+    actions_close = inner.index("</div>")  # the .chat-actions wrapper closes first
+    assert actions_close < inner.index('id="char-count"') < inner.index('id="composer"'), (
+        "the counter must sit inside .chat-bottom between the chat-actions "
+        "row and the composer form (its flow position above the box)"
+    )
+
+
+def test_counter_comment_records_the_not_live_region_decision() -> None:
+    """The phase-104 decision record on the element: the .is-max state
+    is --err-* PLUS a copy change (B3 — text + color, never color
+    alone) and the counter is NOT a live region (per-keystroke
+    feedback is decorative; the over-cap failure announces through the
+    role=alert error banner)."""
+    comment = re.search(
+        r"<!--\s*Phase 104\s*\(owner 2026-09-12\): the question-length counter[\s\S]*?-->",
+        _index_html(),
+    )
+    assert comment, "the phase-104 provenance comment must sit on the counter"
+    text = comment.group(0)
+    assert "NOT" in text and "live region" in text.lower(), (
+        "the counter must stay out of the aria-live contract — the "
+        "over-cap failure path announces through the error banner"
+    )
+    assert "B3" in text, (
+        "the .is-max state must be recorded as text + color (B3), "
+        "never color alone"
+    )
+
+
+def test_cap_constants_mirror_the_server_cap() -> None:
+    """The JS cap constants: ``MAX_QUESTION_CHARS = 4000`` (mirrors the
+    schema — the executor must NOT change app/schemas.py) and
+    ``CHAR_COUNT_SHOW_AT = 3200`` (80% of the cap, owner A4). The
+    single-source pin below proves the three copies (HTML / JS / schema)
+    cannot drift."""
+    js = _app_js()
+    m = re.search(r"const MAX_QUESTION_CHARS = (\d+);", js)
+    assert m, "app.js must define the MAX_QUESTION_CHARS constant"
+    assert int(m.group(1)) == _schema_question_cap(), (
+        "MAX_QUESTION_CHARS must mirror ChatRequest.message's "
+        "max_length — the cap lives in one place conceptually"
+    )
+    s = re.search(r"const CHAR_COUNT_SHOW_AT = (\d+);", js)
+    assert s, "app.js must define the CHAR_COUNT_SHOW_AT constant"
+    assert int(s.group(1)) == 0.8 * int(m.group(1)), (
+        "the counter appears at exactly 80% of the cap (owner A4) — "
+        "no noise on normal use, visible when it matters"
+    )
+
+
+def test_html_maxlength_equals_js_constant_cross_file() -> None:
+    """THE single-source cross-file pin: the HTML ``maxlength`` on
+    #message-input, the JS ``MAX_QUESTION_CHARS`` constant, and the
+    server's ``ChatRequest.message`` cap are the SAME number (regex-
+    parsed from both frontend files + the schema). Any one drifting is
+    a blind 422 (or a counter that lies about the cap)."""
+    html_m = re.search(r'maxlength="(\d+)"', _message_input_tag(_index_html()))
+    js_m = re.search(r"const MAX_QUESTION_CHARS = (\d+);", _app_js())
+    assert html_m and js_m, "both the HTML maxlength and the JS constant must exist"
+    html_len = int(html_m.group(1))
+    js_len = int(js_m.group(1))
+    assert html_len == js_len == _schema_question_cap() == 4000, (
+        f"HTML maxlength={html_len}, JS MAX_QUESTION_CHARS={js_len} and the "
+        "schema cap must all be the one number (4,000)"
+    )
+
+
+def test_update_char_count_contract() -> None:
+    """The counter's state machine (owner A4): RAW length (no trim —
+    raw ≤ cap ⟹ trimmed ≤ cap, a safe superset of what the server
+    validates), hidden below the 80% threshold, plain ``len/4000`` at
+    and above it, and ``len/4000 — character limit`` + the .is-max
+    (err family) treatment at/over the cap — the over-cap reading keeps
+    the HONEST length (the chip-fill path exceeds maxlength, e.g.
+    ``5123/4000 — character limit``)."""
+    body = _function_body(_app_js(), "function updateCharCount() {")
+    assert "const len = input.value.length;" in body, (
+        "the count is the RAW value (no trim) — a raw count is a safe "
+        "superset of what the server validates"
+    )
+    assert "len < CHAR_COUNT_SHOW_AT" in body, (
+        "below the 80% threshold the counter stays hidden"
+    )
+    assert 'charCountEl.hidden = true;' in body
+    assert 'charCountEl.classList.remove("is-max");' in body, (
+        "hiding the counter must also drop the .is-max state — a short "
+        "question after a maxed one must not keep the error color"
+    )
+    assert "len >= MAX_QUESTION_CHARS" in body, (
+        "the at/over-cap branch is >= (4,000 itself is already at the "
+        "limit — the server accepts exactly 4,000, so the counter says "
+        "so at 4,000)"
+    )
+    assert 'charCountEl.classList.toggle("is-max", atMax);' in body
+    assert "— character limit" in body, (
+        "the .is-max state must CHANGE THE COPY (B3) — color alone is "
+        "not the state; the words carry it"
+    )
+    assert "`${len}/${MAX_QUESTION_CHARS}`" in body, (
+        "the counter text is len/cap from the SAME constants — the "
+        "displayed cap can never drift from MAX_QUESTION_CHARS"
+    )
+    grabber = re.search(
+        r'const charCountEl = document\.querySelector\("#char-count"\);', _app_js()
+    )
+    assert grabber, "app.js must grab #char-count alongside the other controls"
+
+
+def test_handle_send_guard_blocks_over_cap_questions() -> None:
+    """The over-cap guard (owner A5): the one reachable path that
+    bypasses maxlength is the programmatic chip fill, and it is
+    blocked in handleSend — trimmed text over the cap gets the
+    out-of-turn error banner (the saveAsDoc precedent), NO turn, and
+    the input KEEPS the text (the user trims it — never stale, PLAN
+    §7.4). The guard sits AFTER the !text guard and BEFORE the clear.
+    """
+    body = _function_body(_app_js(), "async function handleSend(e) {")
+    guard = re.search(
+        r"if \(text\.length > MAX_QUESTION_CHARS\) \{\s*"
+        r"showErrorBanner\(\"Questions are limited to 4,000 characters"
+        r" — trim the question and try again\.\"\);\s*"
+        r"return;\s*\}",
+        body,
+    )
+    assert guard, (
+        "handleSend must refuse a trimmed question over the cap with the "
+        "4,000-characters banner copy (the UI makes the server's 422 "
+        "visible BEFORE the request)"
+    )
+    assert "if (!text || sendBtn.disabled) return;" in body
+    assert body.index("text.length > MAX_QUESTION_CHARS") < body.index(
+        'input.value = "";'
+    ), (
+        "the guard must run BEFORE the clear — the input keeps the kept "
+        "text so the user can trim it (a cleared input would be stale)"
+    )
+
+
+def test_all_four_mutation_sites_run_update_char_count() -> None:
+    """``updateCharCount()`` runs at the EXACT four ``input.value``
+    mutation sites (each already ran ``autoGrow()`` there) — the
+    counter can never show a stale length:
+
+    * the ``input`` listener — keystrokes + pastes (maxlength caps
+      both); the listener body now runs autoGrow AND updateCharCount;
+    * ``submitSuggestion`` — the chip one-tap fill (the programmatic
+      path maxlength cannot stop);
+    * the ``handleSend`` post-send clear — the sent question hides the
+      counter again;
+    * the ``startNewChat`` clear — same.
+    """
+    js = _app_js()
+    # (a) the input listener — both calls inside the one listener body
+    listener = re.search(
+        r'input\.addEventListener\("input", \(\) => \{[\s\S]*?autoGrow\(\);'
+        r"[\s\S]*?updateCharCount\(\);[\s\S]*?\}\);",
+        js,
+    )
+    assert listener, (
+        "the input listener must run autoGrow AND updateCharCount — the "
+        "counter follows every input-path change"
+    )
+    # (b) the chip one-tap fill
+    sub = _function_body(js, "function submitSuggestion(text) {")
+    assert sub.index("input.value = text;") < sub.index("autoGrow();") < sub.index(
+        "updateCharCount();"
+    ), "submitSuggestion must count the (possibly over-cap) filled text"
+    # (c) the post-send clear
+    send = _function_body(js, "async function handleSend(e) {")
+    assert send.index('input.value = "";') < send.index("autoGrow();") < send.index(
+        "updateCharCount();"
+    ), "the post-send clear must hide the counter with the input"
+    # (d) the new-chat clear
+    newchat = _function_body(js, "function startNewChat() {")
+    assert newchat.index('input.value = "";') < newchat.index("autoGrow();") < newchat.index(
+        "updateCharCount();"
+    ), "the new-chat clear must hide the counter with the input"
+
+
+def test_char_count_css_is_aa_on_the_app_background() -> None:
+    """The counter's two color pairings are WCAG-AA-verified against
+    the background it actually sits on — the APP background (--bg)
+    behind the transparent .chat-bottom unit — with each ratio recorded
+    in the rule's comment (house style): --ink-soft on --bg = 8.6:1,
+    --err-ink on --bg = 10.4:1 (both ≥4.5:1). The .is-max state pairs
+    the color with the "— character limit" copy change (B3)."""
+    css = _css()
+    body = _single_line_rule(css, ".char-count")
+    assert "text-align: right;" in body, (
+        "the counter right-aligns above the composer (house layout)"
+    )
+    assert "color: var(--ink-soft);" in body
+    is_max = _single_line_rule(css, ".char-count.is-max")
+    assert "color: var(--err-ink);" in is_max, (
+        "the at/over-cap state must use the --err-* semantic family (B3)"
+    )
+    comment = re.search(r"(/\*[^*]*?\*/)\s*\.char-count \{", css)
+    assert comment, "the .char-count rule must carry its provenance comment"
+    note = comment.group(1)
+    assert "8.6:1" in note, (
+        "the --ink-soft on --bg ratio must be recorded (verified "
+        "8.6:1 ≥ 4.5:1, WCAG AA)"
+    )
+    assert "10.4:1" in note, (
+        "the --err-ink on --bg ratio must be recorded (verified "
+        "10.4:1 ≥ 4.5:1, WCAG AA)"
+    )
diff --git a/tests/unit/test_pinned_composer.py b/tests/unit/test_pinned_composer.py
index 3ace264..044db49 100644
--- a/tests/unit/test_pinned_composer.py
+++ b/tests/unit/test_pinned_composer.py
@@ -312,12 +312,13 @@ def test_chat_bottom_unit_is_last_child_of_the_chat_shell() -> None:
     """Phase 65 (task 02, owner-locked A1): the LAST element child of
     `.chat-shell` is the `.chat-bottom` wrapper — NO id (nothing in JS
     binds it; the bindings live on the inner elements, the move is pure
-    HTML/CSS) — holding exactly the `.chat-actions` row and then the
-    `#composer` form, in that order: the row + composer are ONE sticky
-    unit, and the wrapper owns the shell's bottom slot, so the sticky
-    shift range is still that column's box (a sibling after it would
-    carve the range away and re-break the pin). The composer form keeps
-    `novalidate` and its contract ids."""
+    HTML/CSS) — holding the `.chat-actions` row, the phase-104
+    `#char-count` counter, and the `#composer` form, in that order: the
+    row + counter + composer are ONE sticky unit, and the wrapper owns
+    the shell's bottom slot, so the sticky shift range is still that
+    column's box (a sibling after it would carve the range away and
+    re-break the pin). The composer form keeps `novalidate` and its
+    contract ids."""
     shell = _tree().find("chat-shell")
     last = shell["children"][-1]
     assert last["tag"] == "div" and (
@@ -330,17 +331,29 @@ def test_chat_bottom_unit_is_last_child_of_the_chat_shell() -> None:
         "elements"
     )
     kids = last["children"]
-    assert len(kids) == 2, (
-        "the unit holds exactly two element children: .chat-actions, then "
-        "#composer"
+    assert len(kids) == 3, (
+        "the unit holds exactly three element children: .chat-actions, "
+        "then #char-count (phase 104), then #composer"
     )
-    row, form = kids
+    row, counter, form = kids
     assert row["tag"] == "div" and (
         row["attrs"].get("class") or ""
     ).split() == ["chat-actions"], (
         "the first child is the .chat-actions row (the New chat → Share "
         "DOM order is pinned by test_save_chat_ui.py)"
     )
+    # Phase 104 (owner 2026-09-12): the question-length counter — a
+    # hidden-by-default <p> between the row and the composer (zero
+    # height while hidden; a new child INSIDE the unit breaks no pin —
+    # the sticky geometry below is untouched).
+    assert counter["tag"] == "p" and counter["attrs"].get("id") == "char-count", (
+        "the second child is the phase-104 #char-count counter"
+    )
+    assert (counter["attrs"].get("class") or "") == "char-count"
+    assert "hidden" in counter["attrs"], (
+        "the counter ships hidden — it appears only from 80% of the "
+        "4,000-char cap (app.js updateCharCount)"
+    )
     assert form["tag"] == "form" and form["attrs"].get("id") == "composer"
     assert "novalidate" in form["attrs"], (
         "phase 48: the composer form stays `novalidate` (a `required` "
diff --git a/tests/unit/test_save_chat_ui.py b/tests/unit/test_save_chat_ui.py
index b189fc9..b7877e4 100644
--- a/tests/unit/test_save_chat_ui.py
+++ b/tests/unit/test_save_chat_ui.py
@@ -695,8 +695,10 @@ def test_chat_actions_wrapper_holds_both_pills_in_order() -> None:
     phase 65 (2026-09-01, ``TODO.md`` L3, owner confirmation) moved it
     from the top of the column to the bottom: below the ``#messages``
     section, directly above the composer; nothing but the row's own
-    comment lands between ``#messages`` and the row, and nothing but
-    the composer comment lands between the row and the composer. No
+    comment lands between ``#messages`` and the row, and nothing but the
+    phase-104 ``#char-count`` counter + the composer comment lands
+    between the row and the composer (the counter is hidden by default —
+    zero height, the pinned-cluster geometry untouched). No
     other page carries ``.chat-actions`` (chat-page only, like the
     pills)."""
     html = _index()
@@ -730,12 +732,23 @@ def test_chat_actions_wrapper_holds_both_pills_in_order() -> None:
         and "<form" not in between
     ), ("nothing but the row's comment lands between #messages and the row")
     after = html[end:composer_idx]
+    # Phase 104 (owner 2026-09-12): the ONE permitted child between the
+    # row and the composer is the hidden-by-default question-length
+    # counter — everything else (ids, buttons, sections, forms) is
+    # still excluded from the gap.
+    after_minus_counter = after.replace(
+        '<p class="char-count" id="char-count" hidden></p>', ""
+    )
     assert (
-        "id=" not in after
-        and "<button" not in after
-        and "<section" not in after
-        and "<form" not in after
-    ), ("nothing but the composer comment lands between the row and the composer")
+        after.count("<p") == 1
+        and "id=" not in after_minus_counter
+        and "<button" not in after_minus_counter
+        and "<section" not in after_minus_counter
+        and "<form" not in after_minus_counter
+    ), (
+        "nothing but the phase-104 counter + the composer comment lands "
+        "between the row and the composer"
+    )
     # Phase 76 (task 02): the folded view files are gone (the shell's
     # chat view is the one and only carrier of the row — pinned above);
     # the standalone pages carry none (task 03 dropped the last folded
diff --git a/tests/unit/test_schemas.py b/tests/unit/test_schemas.py
index 02bc04d..1704de9 100644
--- a/tests/unit/test_schemas.py
+++ b/tests/unit/test_schemas.py
@@ -15,6 +15,7 @@ from pydantic import ValidationError
 
 from app.schemas import (
     ChatMessage,
+    ChatRequest,
     SavedChatCreate,
     SavedChatUpdate,
     SourceRef,
@@ -34,6 +35,7 @@ TITLE_CAP = 500  # documents.title String(500)
 NAME_CAP = 100  # ToolCall.name
 ARGUMENT_CAP = 2000  # ToolCall.argument
 MESSAGES_CAP = 200  # SavedChatCreate/Update.messages
+QUESTION_CAP = 4_000  # ChatRequest.message — the cap the composer mirrors (phase 104)
 
 
 def _source_ref() -> dict:
@@ -174,6 +176,30 @@ def test_chat_message_tools_one_over_cap_rejects() -> None:
     _failed_loc(exc.value, "tools")
 
 
+# ---------------------------------------------------------------------------
+# ChatRequest.message (the 4,000-char question cap — phase 104, A3)
+# ---------------------------------------------------------------------------
+
+
+def test_chat_request_message_at_cap_validates() -> None:
+    """EXACTLY 4,000 chars passes — the cap admits "a code example of a
+    few dozen lines" (owner A3) and the composer's maxlength + counter
+    mirror this exact boundary (the 4,000/4,000 E2E submit must not
+    422)."""
+    req = ChatRequest.model_validate({"message": "a" * QUESTION_CAP})
+    assert len(req.message) == QUESTION_CAP
+
+
+def test_chat_request_message_one_over_cap_rejects() -> None:
+    """4,001 chars is the 422 the UI now prevents blind: the server cap
+    stays the backstop (pre-existing, untouched — this phase only pins
+    it at the boundary, phase-83 pattern), naming ``message`` in the
+    error loc."""
+    with pytest.raises(ValidationError) as exc:
+        ChatRequest.model_validate({"message": "a" * (QUESTION_CAP + 1)})
+    _failed_loc(exc.value, "message")
+
+
 # ---------------------------------------------------------------------------
 # SourceRef (documents column-length mirrors)
 # ---------------------------------------------------------------------------