fix(bot): close dialog engine gaps (C-SSRS repeat, scale intros, button payloads, step timeout)

- No same-day C-SSRS repeat: a risk screen completed within cssrs_cooldown_hours is not queued again
  after PHQ-9 item 9 >= 1 (scale crisis path, rule schedule_scale, scale_due); crisis script and alerts stay.
- Scale intro registry: register_scale_intro(scale_id, text_or_fn); start_scale(intro=None) uses it.
  veteran_daily migrated from per-scale on_enter hooks.
- Prefix triggers "button:<prefix>:*" with trigger_payload(event); medication buttons carry the dose
  (med:taken:<dose_id>, ...), med_pending dropped; payload-less legacy buttons answer as stale.
- Abandoned-step timeout: settings.abandon_after_hours (default 6 h) returns abandonable states
  (daily.*, meds.*, scale.* not in onboarding) to idle before the next event; never in crisis or with intake on.

Refs #56

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/test_dialog_engine_gaps.py

@@ -0,0 +1,385 @@
+"""Engine gaps found by the 13-day run (#56): no same-day C-SSRS repeat, scale intro registry, prefix triggers
+with a button payload (medication buttons carry the dose), abandoned-step timeout."""
+
+from collections.abc import AsyncIterator
+from datetime import datetime, timedelta
+from typing import Any
+
+import pytest
+from sqlalchemy import select
+
+from app.config import Settings
+from app.core import clock
+from app.core.rules import Action
+from app.core.scales import load_scale
+from app.db import repo, session
+from app.db.bootstrap import init_db
+from app.db.clinical import Dose, ItemResponse, Prescription
+from app.dialog import engine as engine_module
+from app.dialog import fsm, scale_session
+from app.dialog.contracts import Event, Reply
+from app.dialog.engine import STEP_AT, Engine
+from app.dialog.flows.meds import load_meds_texts
+from app.dialog.flows.veteran_daily import load_daily_texts
+from app.dialog.fsm import Ctx, Step, get_trigger, is_registered, trigger, trigger_payload
+from app.dialog.texts import load_dialog_texts
+from app.llm.gateway import LLMGateway
+from app.llm.mock import MockTasks
+
+TOKEN = "g" * 32
+OTHER = "h" * 32
+MEM = "sqlite+aiosqlite:///:memory:"
+NOW = datetime(2026, 10, 7, 9, 0)
+TEXTS = load_dialog_texts()
+MEDS = load_meds_texts()
+DAILY = load_daily_texts()
+SCREEN = TEXTS.settings.risk_screen_scale
+
+
+class ScriptedMonitoring:
+    def __init__(self) -> None:
+        self.actions: list[Action] = []
+
+    async def after_data_change(self, token: str) -> list[Action]:
+        out, self.actions = self.actions, []
+        return out
+
+    async def is_in_crisis(self, token: str) -> bool:
+        return True
+
+
+@pytest.fixture
+async def mon(monkeypatch: pytest.MonkeyPatch) -> AsyncIterator[ScriptedMonitoring]:
+    monkeypatch.setattr(clock, "get_settings", lambda: Settings(_env_file=None, demo_clock=True))
+    clock.set_now(NOW)
+    await session.configure(identity_url=MEM, clinical_url=MEM)
+    await init_db()
+    async with session.get_clinical_session() as s:
+        await repo.upsert_profile(
+            s, token=TOKEN, doctor_id="doc", modules=["veteran"], sex="m", birth_year=1990, therapy_start=NOW.date()
+        )
+        s.add(Prescription(token=TOKEN, drug="сертралин", dose="50 мг", schedule={"times": ["08:00"]}))
+        await repo.set_dialog_state(s, token=TOKEN, state="idle", data={"consent": True}, now=clock.now())
+    yield ScriptedMonitoring()
+    clock.reset()
+    await session.configure()
+
+
+@pytest.fixture
+def engine(mon: ScriptedMonitoring) -> Engine:
+    return Engine(gateway=LLMGateway(MockTasks()), monitoring=mon)
+
+
+async def press(e: Engine, data: str) -> list[Reply]:
+    return await e.handle(TOKEN, Event(type="button", button=data))
+
+
+async def text(e: Engine, t: str) -> list[Reply]:
+    return await e.handle(TOKEN, Event(type="text", text=t))
+
+
+async def system(e: Engine, kind: str, **params: Any) -> list[Reply]:
+    return await e.handle(TOKEN, Event(type="system", system=kind, params=params))  # type: ignore[arg-type]
+
+
+async def answer(e: Engine, scale_id: str, values: list[int]) -> list[Reply]:
+    out: list[Reply] = []
+    for item, value in zip(load_scale(scale_id).items, values, strict=False):
+        out += await press(e, f"scale:{scale_id}:{item.id}:{value}")
+    return out
+
+
+def buttons(replies: list[Reply]) -> list[str]:
+    return [b.data for r in replies for row in r.buttons for b in row]
+
+
+async def get_state() -> tuple[str, dict[str, Any]]:
+    async with session.get_clinical_session() as s:
+        row = await repo.get_dialog_state(s, TOKEN)
+        assert row is not None
+        return row.state, dict(row.data)
+
+
+async def results(scale: str) -> list[Any]:
+    async with session.get_clinical_session() as s:
+        return list(await repo.list_scale_results(s, TOKEN, scale))
+
+
+async def checkin() -> Any:
+    async with session.get_clinical_session() as s:
+        rows = await repo.list_checkins(s, TOKEN)
+        return rows[-1] if rows else None
+
+
+async def doses() -> list[Dose]:
+    async with session.get_clinical_session() as s:
+        return list((await s.scalars(select(Dose).where(Dose.token == TOKEN).order_by(Dose.id))).all())
+
+
+# ----------------------------------------------------------------------------------------------- 1. C-SSRS repeat
+
+
+async def test_item9_right_after_cssrs_does_not_repeat_screen(engine: Engine, mon: ScriptedMonitoring) -> None:
+    await system(engine, "scale_due", scale=SCREEN, scheduled=False)
+    assert (await get_state())[0] == f"scale.{SCREEN}"
+    await answer(engine, SCREEN, [0] * 6)
+    clock.advance(hours=1)
+    await system(engine, "scale_due", scale="phq9", scheduled=False)
+    # The rule phq9_item9 also asks for the screen — neither path may repeat it within the cooldown.
+    mon.actions = [Action(type="schedule_scale", code="phq9_item9", params={"scale": SCREEN, "scheduled": False})]
+    r = await answer(engine, "phq9", [1] * 8 + [1])
+    assert any(x.kind == "crisis" for x in r)  # the fixed crisis script still goes out
+    state, data = await get_state()
+    assert state == "crisis" and data["crisis"]["open"] is True and not data.get("pending_scales")
+    assert len(await results(SCREEN)) == 1
+
+
+async def test_item9_after_old_cssrs_asks_screen_again(engine: Engine) -> None:
+    await system(engine, "scale_due", scale=SCREEN, scheduled=False)
+    await answer(engine, SCREEN, [0] * 6)
+    clock.advance(hours=TEXTS.settings.cssrs_cooldown_hours + 1)
+    await system(engine, "scale_due", scale="phq9", scheduled=False)
+    r = await answer(engine, "phq9", [1] * 8 + [1])
+    assert any(x.kind == "crisis" for x in r)
+    assert (await get_state())[0] == f"scale.{SCREEN}" and TEXTS.scale.risk_screen_intro in r[-1].text
+
+
+async def test_unfinished_cssrs_does_not_count(engine: Engine) -> None:
+    await system(engine, "scale_due", scale=SCREEN, scheduled=False)
+    await answer(engine, SCREEN, [0, 0])  # left halfway: no complete result
+    await system(engine, "scale_due", scale="phq9", scheduled=False)  # queued behind the screen
+    async with session.get_clinical_session() as s:
+        assert not await scale_session.recent_risk_screen(_ctx(s), SCREEN)
+
+
+def _ctx(s: Any, state: str = "idle", data: dict[str, Any] | None = None) -> Ctx:
+    return Ctx(
+        token=TOKEN,
+        state=state,
+        data=data or {},
+        now=clock.now(),
+        session=s,
+        gateway=LLMGateway(MockTasks()),
+        monitoring=ScriptedMonitoring(),
+        texts=TEXTS,
+    )
+
+
+# ----------------------------------------------------------------------------------------------- 2. intros
+
+
+async def test_registered_intro_text_and_callable(engine: Engine, monkeypatch: pytest.MonkeyPatch) -> None:
+    seen: list[tuple[str, bool, str]] = []
+
+    def intro(ctx: Ctx, scale_id: str, scheduled: bool, return_state: str) -> str | None:
+        seen.append((scale_id, scheduled, return_state))
+        return "Вступление к GAD-7"
+
+    monkeypatch.setattr(scale_session, "_intros", dict(scale_session._intros))
+    scale_session.register_scale_intro("gad7", intro)
+    with pytest.raises(ValueError):
+        scale_session.register_scale_intro("gad7", "второе")
+    r = await system(engine, "scale_due", scale="gad7", scheduled=True)
+    assert r[0].text.startswith("Вступление к GAD-7") and seen == [("gad7", True, "idle")]
+
+    scale_session.register_scale_intro("gad7", "Текстовое вступление", replace=True)
+    async with session.get_clinical_session() as s:
+        step = await scale_session.start_scale(_ctx(s), "gad7", return_state="idle")
+        assert step.data["scales"]["gad7"]["intro"] == "Текстовое вступление"
+        explicit = await scale_session.start_scale(_ctx(s), "gad7", return_state="idle", intro="Своё")
+        assert explicit.data["scales"]["gad7"]["intro"] == "Своё"
+
+
+def test_daily_intros_use_registry_not_enter_hooks() -> None:
+    for scale_id in DAILY.settings.intro_scales:
+        assert not is_registered(f"scale.{scale_id}")  # no per-scale enter hooks any more
+        assert scale_id in scale_session._intros
+
+
+async def test_unscheduled_intro_from_registry(engine: Engine) -> None:
+    r = await system(engine, "scale_due", scale="phq9", scheduled=False)
+    assert DAILY.scale_intros.unscheduled["phq9"] in r[0].text
+
+
+# ----------------------------------------------------------------------------------------------- 3. payload buttons
+
+
+async def test_prefix_trigger_payload_and_precedence(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setattr(fsm, "_triggers", dict(fsm._triggers))
+    monkeypatch.setattr(fsm, "_prefix_triggers", dict(fsm._prefix_triggers))
+
+    def handler(tag: str) -> Any:
+        async def fn(ctx: Ctx, event: Event) -> Step:
+            return Step(data={"tag": tag, "payload": trigger_payload(event)})
+
+        return fn
+
+    trigger("button:zz:a:*")(handler("short"))
+    trigger("button:zz:a:b:*")(handler("long"))
+    trigger("button:zz:a:exact")(handler("exact"))
+    with pytest.raises(ValueError):
+        trigger("button:zz:a:*")(handler("dup"))
+
+    async def run(key: str) -> dict[str, Any] | None:
+        fn = get_trigger(key)
+        if fn is None:
+            return None
+        return (await fn(None, Event(type="button", button=key.split(":", 1)[1]))).data  # type: ignore[arg-type]
+
+    assert await run("button:zz:a:42") == {"tag": "short", "payload": "42"}
+    assert await run("button:zz:a:b:7:x") == {"tag": "long", "payload": "7:x"}
+    assert await run("button:zz:a:exact") == {"tag": "exact", "payload": None}
+    assert await run("button:zz:a:") is None and await run("button:zz:a") is None
+
+
+async def test_reminder_buttons_carry_dose_and_no_state_key(engine: Engine) -> None:
+    async with session.get_clinical_session() as s:
+        await repo.set_dialog_state(
+            s, token=TOKEN, state="idle", data={"consent": True, "med_pending": {"dose_id": 1}}, now=clock.now()
+        )
+    r = await system(engine, "med_reminder")
+    [dose] = await doses()
+    assert buttons(r) == [f"med:taken:{dose.id}", f"med:skip:{dose.id}", f"med:side_effects:{dose.id}"]
+    assert "med_pending" not in (await get_state())[1]
+    skip = await press(engine, f"med:skip:{dose.id}")
+    assert f"med:skip_reason:side_effects:{dose.id}" in buttons(skip)
+    se = await press(engine, f"med:skip_reason:side_effects:{dose.id}")
+    await press(engine, f"med:se:nausea:{dose.id}")
+    [dose] = await doses()
+    assert dose.status == "skipped" and dose.side_effects == "nausea" and "med:se:nausea:" in buttons(se)[0]
+
+
+async def test_two_reminders_each_button_answers_its_own_dose(engine: Engine) -> None:
+    first = await system(engine, "med_reminder")
+    second = await system(engine, "med_reminder")
+    await press(engine, buttons(first)[0])  # «Принял» of the FIRST reminder, pressed after the second arrived
+    d1, d2 = await doses()
+    assert (d1.status, d2.status) == ("taken", "pending")
+    await press(engine, buttons(second)[1])
+    assert [d.status for d in await doses()] == ["taken", "skipped"]
+
+
+@pytest.mark.parametrize("data", ["med:taken", "med:skip", "med:se:nausea", "med:dose:yes", "med:skip_reason:quit"])
+async def test_legacy_payloadless_buttons_are_stale(engine: Engine, data: str) -> None:
+    await system(engine, "med_reminder")
+    r = await press(engine, data)
+    assert r[0].text == MEDS.stale and await checkin() is None
+    assert [d.status for d in await doses()] == ["pending"]
+
+
+async def test_foreign_unknown_and_old_dose_buttons_are_stale(engine: Engine) -> None:
+    async with session.get_clinical_session() as s:
+        rx = Prescription(token=OTHER, drug="сертралин", dose="50 мг", schedule={"times": ["08:00"]})
+        s.add(rx)
+        await s.flush()
+        foreign = Dose(token=OTHER, prescription_id=rx.id, planned_at=clock.now(), status="pending")
+        s.add(foreign)
+        await s.flush()
+        foreign_id = foreign.id
+    for data in (f"med:taken:{foreign_id}", "med:taken:9999", "med:taken:garbage", "med:taken:2026-10-06"):
+        assert (await press(engine, data))[0].text == MEDS.stale
+    old = await system(engine, "med_reminder")
+    clock.advance(days=1)
+    assert (await press(engine, buttons(old)[0]))[0].text == MEDS.stale
+    assert await checkin() is None
+
+
+async def test_reminder_without_dose_row_uses_day_payload(engine: Engine) -> None:
+    async with session.get_clinical_session() as s:
+        for row in await s.scalars(select(Prescription)):
+            await s.delete(row)
+    r = await system(engine, "med_reminder")
+    assert buttons(r)[0] == f"med:taken:{NOW.date().isoformat()}"
+    r = await press(engine, buttons(r)[0])
+    assert r[0].text == MEDS.taken_ack and (await checkin()).med_taken is True and await doses() == []
+
+
+async def test_checkin_med_step_answers_todays_dose(engine: Engine) -> None:
+    await system(engine, "med_reminder")
+    await system(engine, "daily_checkin")
+    await press(engine, "daily:mood:5")
+    await press(engine, "daily:med:yes")
+    assert [d.status for d in await doses()] == ["taken"]
+
+
+# ----------------------------------------------------------------------------------------------- 4. abandon timeout
+
+
+async def test_abandoned_checkin_returns_to_idle(engine: Engine) -> None:
+    await system(engine, "daily_checkin")
+    await press(engine, "daily:mood:3")
+    assert (await get_state())[0].startswith("daily.")
+    clock.advance(hours=TEXTS.settings.abandon_after_hours + 1)
+    r = await system(engine, "med_reminder")
+    state, data = await get_state()
+    assert state == "idle" and "checkin" not in data
+    assert len(r) == 1 and r[0].text.startswith(MEDS.reminder.text.split("{")[0])  # no check-in question repeated
+    assert (await checkin()).mood == 3  # stored answer survives
+
+
+async def test_short_pause_keeps_checkin(engine: Engine) -> None:
+    await system(engine, "daily_checkin")
+    clock.advance(hours=TEXTS.settings.abandon_after_hours - 1)
+    r = await press(engine, "daily:mood:6")
+    assert (await get_state())[0] != "idle" and r
+
+
+async def test_abandoned_scale_lets_queued_scale_start(engine: Engine) -> None:
+    await system(engine, "scale_due", scale="phq9", scheduled=True)
+    await answer(engine, "phq9", [1, 2])
+    clock.advance(hours=4)
+    await system(engine, "scale_due", scale="pcl5", scheduled=True)  # queued; system events don't extend the step
+    clock.advance(hours=3)
+    r = await text(engine, "добрый вечер")
+    state, data = await get_state()
+    assert state == "scale.pcl5" and "phq9" not in data["scales"] and not data.get("pending_scales")
+    assert DAILY.scale_intros.scheduled["pcl5"] in r[-1].text
+    assert await results("phq9") == []  # unfinished: no total, no severity
+    async with session.get_clinical_session() as s:
+        stored = (await s.scalars(select(ItemResponse).where(ItemResponse.scale == "phq9"))).all()
+    assert [x.score for x in stored] == [1, 2]
+
+
+async def test_timeout_never_applies_in_crisis(engine: Engine) -> None:
+    await system(engine, "scale_due", scale="phq9", scheduled=False)
+    await answer(engine, "phq9", [0] * 8 + [2])  # item 9 → crisis + C-SSRS (none done before)
+    state, data = await get_state()
+    assert state == f"scale.{SCREEN}" and data["crisis"]["open"] is True
+    clock.advance(hours=TEXTS.settings.abandon_after_hours * 3)
+    await press(engine, f"scale:{SCREEN}:cssrs_1:0")
+    state, data = await get_state()
+    assert state == f"scale.{SCREEN}" and data["scales"][SCREEN]["idx"] == 1
+
+
+async def test_timeout_never_skips_onboarding_scale(engine: Engine) -> None:
+    item = load_scale("pcl5").items[0]
+    old = (clock.now() - timedelta(hours=48)).isoformat()
+    session_data = {
+        "idx": 0,
+        "answers": {},
+        "return_state": "veteran.onboarding.phq9",
+        "scheduled": True,
+        "intro": None,
+        "intro_shown": True,
+    }
+    async with session.get_clinical_session() as s:
+        await repo.set_dialog_state(
+            s,
+            token=TOKEN,
+            state="scale.pcl5",
+            data={"consent": True, "scales": {"pcl5": session_data}, STEP_AT: old},
+            now=clock.now(),
+        )
+    await press(engine, f"scale:pcl5:{item.id}:1")
+    state, data = await get_state()
+    assert state == "scale.pcl5" and data["scales"]["pcl5"]["idx"] == 1
+
+
+async def test_timeout_disabled_same_day(engine: Engine, monkeypatch: pytest.MonkeyPatch) -> None:
+    clock.set_now(NOW.replace(hour=1))
+    off = TEXTS.model_copy(update={"settings": TEXTS.settings.model_copy(update={"abandon_after_hours": 0})})
+    monkeypatch.setattr(engine_module, "load_dialog_texts", lambda: off)
+    await system(engine, "daily_checkin")
+    clock.advance(hours=20)
+    await press(engine, "daily:mood:4")
+    assert (await get_state())[0] != "idle" and (await checkin()).mood == 4