cctally 1.7.4 → 1.8.0

package/bin/_lib_view_models.py

@@ -0,0 +1,993 @@
+"""View-model kernel for CLI / dashboard / share consumers.
+
+This module owns the per-domain row dataclasses and the ``*View``
+wrappers that carry rows + data-plane aggregates (totals, averages).
+
+Bundle 1 (landed):
+
+Row dataclasses moved verbatim from ``bin/_cctally_tui.py`` — no field
+changes:
+
+- ``DailyPanelRow``, ``MonthlyPeriodRow``, ``WeeklyPeriodRow``,
+  ``TuiSessionRow``.
+
+``TuiTrendRow`` moved with the 10 nullable fields added per spec §4.1
+(``week_start_date``, ``week_end_date``, ``week_end_at``,
+``weekly_cost_usd``, ``usage_captured_at``, ``cost_captured_at``,
+``as_of``, ``range_start_iso``, ``range_end_iso``, ``freshness``). All
+defaults are ``None`` so existing TUI / dashboard fixtures that
+construct ``TuiTrendRow`` positionally stay byte-stable.
+
+Frozen ``*View`` dataclasses + builders:
+
+- ``DailyView`` + ``build_daily_view(entries, *, now_utc, display_tz)``
+- ``MonthlyView`` + ``build_monthly_view(entries, *, now_utc, n,
+  display_tz)``
+- ``WeeklyView`` + ``build_weekly_view(conn, entries, *, weeks, now_utc,
+  display_tz, as_of_utc)``
+- ``TrendView`` + ``build_trend_view(conn, *, now_utc, n, display_tz)``
+- ``SessionsView`` + ``build_sessions_view(entries, *, now_utc, limit,
+  display_tz)``
+
+Each ``*View`` carries ``rows`` (typed row tuple) plus a parallel
+``aggregated`` ``BucketUsage`` tuple where CLI byte-stable JSON requires
+it (the weekly view also carries an ``overlay`` tuple of ``(used_pct,
+dpp)`` pairs aligned with ``aggregated``). All builders return totals
+(``total_cost_usd`` / ``total_tokens``) so the dashboard envelope
+adapter (in ``bin/_cctally_dashboard.py``) and the React panel layer
+share a single source of truth.
+
+Helpers:
+
+- ``_load_lib(name)`` — late-import a sibling under ``bin/`` (matches
+  ``_lib_aggregators._load_lib``; keeps the import-time graph acyclic).
+- ``_cctally()`` — accessor for the ``cctally`` module at call-time
+  (spec §5.5; lets builders touch top-level helpers without binding at
+  module load).
+- ``_display_tz_label`` / ``_model_breakdowns_to_models_late`` —
+  presentation-side helpers shared across builders.
+
+``bin/_cctally_tui.py`` re-exports each name so historical imports
+(``from _cctally_tui import DailyPanelRow``, ``ns["DailyPanelRow"]``
+direct-dict reads in tests) keep resolving.
+
+Spec: docs/superpowers/specs/2026-05-17-view-model-unification-design.md
+"""
+from __future__ import annotations
+
+import datetime as dt
+import pathlib
+import sys
+from dataclasses import dataclass
+from typing import Any
+
+
+def _cctally():
+    """Resolve the current ``cctally`` module at call-time (spec §5.5)."""
+    return sys.modules["cctally"]
+
+
+def _load_lib(name: str):
+    """Late-import a sibling under ``bin/`` (same recipe as
+    ``_lib_aggregators._load_lib`` — keeps the import-time graph acyclic
+    even when builders need access to ``_cctally_dashboard`` /
+    ``_lib_aggregators`` / ``_lib_share``).
+    """
+    cached = sys.modules.get(name)
+    if cached is not None:
+        return cached
+    import importlib.util as _ilu
+    p = pathlib.Path(__file__).resolve().parent / f"{name}.py"
+    spec = _ilu.spec_from_file_location(name, p)
+    mod = _ilu.module_from_spec(spec)
+    sys.modules[name] = mod
+    spec.loader.exec_module(mod)
+    return mod
+
+
+# === Row dataclasses (Task 2: moved verbatim from _cctally_tui.py) =========
+# Field order, types, and defaults match the originals byte-stable.
+
+
+@dataclass
+class TuiTrendRow:
+    """Trend row used by CLI ``report``, TUI trend panel, dashboard
+    trend panel, and share ``_build_report_snapshot``.
+
+    The first 7 fields (``week_label`` through ``is_current``) are the
+    historical TUI/dashboard surface — preserved verbatim. The 10
+    nullable fields below were added by spec §4.1 so ``cmd_report`` can
+    consume a single typed shape (eliminates the camelCase dict
+    workaround documented at ``_build_report_snapshot:~12299``).
+
+    JSON serialization sites (``cmd_report --json``, dashboard envelope
+    ``trend.weeks[]``) map field-by-field to today's keys; this typed
+    in-memory shape is internal.
+    """
+    # ---- existing TUI/dashboard fields (verbatim) ----
+    week_label: str              # e.g. "Apr 14"
+    week_start_at: dt.datetime
+    used_pct: float | None       # None when the week has a cost snapshot
+                                 # but no usage snapshot (phantom weeks)
+    dollars_per_percent: float | None
+    delta_dpp: float | None      # vs prior week
+    spark_height: int            # 1..8 normalized
+    is_current: bool
+
+    # ---- NEW (spec §4.1): required by cmd_report JSON contract; nullable ----
+    week_start_date: dt.date | None = None
+    week_end_date: dt.date | None = None
+    week_end_at: dt.datetime | None = None
+    weekly_cost_usd: float | None = None
+    usage_captured_at: str | None = None      # ISO-8601 or None
+    cost_captured_at: str | None = None       # ISO-8601 or None
+    as_of: str | None = None                  # ISO-8601 or None
+    range_start_iso: str | None = None
+    range_end_iso: str | None = None
+    freshness: dict | None = None             # {label, captured_at, age_seconds}
+
+
+@dataclass
+class WeeklyPeriodRow:
+    """One subscription-week row for the dashboard's Weekly panel/modal.
+
+    `models` is a list of `{model, display, chip, cost_usd, cost_pct}`
+    dicts sorted by `cost_usd` descending. Pre-bucketed in Python so
+    the React layer never re-derives per-model coloring.
+    """
+    label: str                          # "04-23" — MM-DD of the week start
+    cost_usd: float
+    total_tokens: int
+    input_tokens: int
+    output_tokens: int
+    cache_creation_tokens: int
+    cache_read_tokens: int
+    used_pct: float | None              # from weekly_usage_snapshots overlay
+    dollar_per_pct: float | None        # cost / used_pct when used_pct > 0
+    delta_cost_pct: float | None        # (cost - prev_cost) / prev_cost
+    is_current: bool
+    models: list[dict[str, Any]]
+    week_start_at: str                  # ISO-8601 with tz, from SubWeek.start_ts
+    week_end_at: str                    # ISO-8601 with tz, from SubWeek.end_ts
+
+
+@dataclass
+class MonthlyPeriodRow:
+    """One calendar-month row for the dashboard's Monthly panel/modal."""
+    label: str                          # "YYYY-MM"
+    cost_usd: float
+    total_tokens: int
+    input_tokens: int
+    output_tokens: int
+    cache_creation_tokens: int
+    cache_read_tokens: int
+    delta_cost_pct: float | None
+    is_current: bool
+    models: list[dict[str, Any]]
+
+
+@dataclass
+class DailyPanelRow:
+    """One row of the dashboard's Daily heatmap panel.
+
+    `intensity_bucket` is the server-computed quintile bucket (0..5) —
+    bucket 0 is reserved for zero-cost days; buckets 1..5 are quintiles
+    over non-zero days.
+
+    v2.3: Added per-day token rollup + `cache_hit_pct` so the Daily
+    detail modal can surface the same fields the CLI's `daily` command
+    shows. Defaults preserve compatibility with `_empty_dashboard_snapshot`
+    and any pre-v2.3 fixture that omits the new fields.
+    """
+    date: str              # local-tz YYYY-MM-DD
+    label: str             # "MM-DD" — pre-formatted, mirrors Weekly/Monthly idiom
+    cost_usd: float
+    is_today: bool
+    intensity_bucket: int  # 0..5
+    models: list[dict[str, Any]]   # ModelCostRow shape, sorted desc by cost
+    # ---- v2.3 additions: Daily modal token + cache rollup ----
+    input_tokens: int = 0
+    output_tokens: int = 0
+    cache_creation_tokens: int = 0
+    cache_read_tokens: int = 0
+    total_tokens: int = 0
+    cache_hit_pct: float | None = None
+
+
+@dataclass
+class TuiSessionRow:
+    started_at: dt.datetime
+    duration_minutes: float
+    model_primary: str           # first model used in the session
+    cost_usd: float
+    cache_hit_pct: float | None
+    project_label: str           # basename of project_path
+    session_id: str              # full session UUID (v2: needed for session-detail modal)
+
+
+# === Internal helpers ======================================================
+
+
+def _display_tz_label(display_tz) -> str:
+    """Mirror of cctally._share_display_tz_label.
+
+    Kept here so builders don't depend on the cctally namespace at
+    build time. ``ZoneInfo`` -> ``zone.key``; ``None`` -> ``"local"``
+    (per resolve_display_tz's convention).
+    """
+    return display_tz.key if display_tz is not None else "local"
+
+
+def _model_breakdowns_to_models_late(model_breakdowns, cost_usd):
+    """Late-bound shim for ``_model_breakdowns_to_models`` in
+    ``_cctally_dashboard``.
+
+    Cannot eagerly import at module load (``_cctally_dashboard`` is a
+    heavier sibling and creating an import-time edge would force its
+    side-effects on every builder load). Resolved at first call.
+    """
+    mod = _load_lib("_cctally_dashboard")
+    return mod._model_breakdowns_to_models(model_breakdowns, cost_usd)
+
+
+# === DailyView + build_daily_view (Task 3) =================================
+
+
+@dataclass(frozen=True)
+class DailyView:
+    """Daily domain view — entries-driven, newest-first.
+
+    ``rows`` carries one ``DailyPanelRow`` per *non-empty* day (NO
+    gap-fill — the dashboard envelope adapter materializes the
+    contiguous heatmap window post-builder; CLI / share consume rows
+    as-is to preserve byte-stable ``cctally daily --json``).
+
+    ``aggregated`` is the parallel ``BucketUsage`` tuple from
+    ``_aggregate_daily`` (same order as ``rows``). CLI's
+    ``_bucket_to_json`` and ``_render_bucket_table`` plus the share
+    ``_build_daily_snapshot`` consume this shape; the dashboard
+    envelope adapter consumes ``rows``.
+
+    Carrying both shapes (BucketUsage + DailyPanelRow) mirrors the
+    SessionsView pattern in Bundle 2 (spec §6.5) — CLI/share renderers
+    today depend on BucketUsage fields (``bucket``, ``model_breakdowns``,
+    ``models: list[str]``) that aren't present on ``DailyPanelRow``;
+    forcing the rename onto the renderer would break byte-stable
+    ``cctally daily --json``.
+    """
+    rows: "tuple[DailyPanelRow, ...]" = ()
+    aggregated: tuple = ()                    # tuple[BucketUsage, ...] — forward-ref kept untyped to avoid an import-time edge into the aggregator's BucketUsage shape.
+    total_cost_usd: float = 0.0
+    total_tokens: int = 0
+    period_start: "dt.datetime | None" = None
+    period_end: "dt.datetime | None" = None
+    display_tz_label: str = ""
+
+
+def build_daily_view(entries, *, now_utc, display_tz=None):
+    """Build a ``DailyView`` from raw ``UsageEntry`` list (spec §5.1).
+
+    Gap-free: only days with entries appear in ``view.rows`` /
+    ``view.aggregated`` (newest-first). The contiguous-window
+    materialization the dashboard heatmap needs is presentation logic
+    and stays at the dashboard envelope adapter.
+
+    Per-row derivations: ``cache_hit_pct`` (cache_read / (input +
+    cache_creation + cache_read) * 100), ``is_today`` (date == today
+    in display_tz), ``models[]`` via
+    ``_model_breakdowns_to_models``.
+
+    Leaves ``DailyPanelRow.label`` and ``intensity_bucket`` at dataclass
+    defaults — the dashboard envelope adapter populates them. CLI /
+    share consumers ignore them and read ``view.aggregated`` instead.
+    """
+    _agg = _load_lib("_lib_aggregators")
+    buckets = _agg._aggregate_daily(entries, mode="auto", tz=display_tz)
+    if not buckets:
+        return DailyView(
+            rows=(),
+            aggregated=(),
+            total_cost_usd=0.0,
+            total_tokens=0,
+            period_start=None,
+            period_end=now_utc,
+            display_tz_label=_display_tz_label(display_tz),
+        )
+
+    today_local = (
+        now_utc.astimezone(display_tz) if display_tz is not None
+        # internal fallback: host-local intentional
+        else now_utc.astimezone()
+    ).date()
+
+    rows = []
+    # buckets come oldest-first from _aggregate_daily; reverse for newest-first.
+    reversed_buckets = list(reversed(buckets))
+    total_cost = 0.0
+    total_tok = 0
+    for b in reversed_buckets:
+        denom = b.input_tokens + b.cache_creation_tokens + b.cache_read_tokens
+        cache_hit = (b.cache_read_tokens / denom * 100.0) if denom > 0 else None
+        d = dt.date.fromisoformat(b.bucket)
+        row = DailyPanelRow(
+            date=b.bucket,
+            label="",                  # adapter fills
+            cost_usd=b.cost_usd,
+            is_today=(d == today_local),
+            intensity_bucket=0,        # adapter fills
+            models=_model_breakdowns_to_models_late(
+                b.model_breakdowns, b.cost_usd,
+            ),
+            input_tokens=b.input_tokens,
+            output_tokens=b.output_tokens,
+            cache_creation_tokens=b.cache_creation_tokens,
+            cache_read_tokens=b.cache_read_tokens,
+            total_tokens=b.total_tokens,
+            cache_hit_pct=cache_hit,
+        )
+        rows.append(row)
+        total_cost += b.cost_usd
+        total_tok += b.total_tokens
+
+    earliest = dt.date.fromisoformat(buckets[0].bucket)
+    period_start = dt.datetime.combine(
+        earliest, dt.time.min, tzinfo=dt.timezone.utc,
+    )
+    return DailyView(
+        rows=tuple(rows),
+        aggregated=tuple(reversed_buckets),
+        total_cost_usd=total_cost,
+        total_tokens=total_tok,
+        period_start=period_start,
+        period_end=now_utc,
+        display_tz_label=_display_tz_label(display_tz),
+    )
+
+
+# === MonthlyView + build_monthly_view (Task 8) =============================
+
+
+@dataclass(frozen=True)
+class MonthlyView:
+    """Monthly domain view — entries-driven, newest-first.
+
+    Like ``DailyView``: ``rows`` carries the typed ``MonthlyPeriodRow``
+    tuple for dashboard/share, ``aggregated`` carries the parallel
+    ``BucketUsage`` tuple for CLI byte-stability.
+
+    Boundary-spillover bucket is dropped (mirrors the existing dashboard
+    builder at ``_dashboard_build_monthly_periods:1752``):
+    in tzs west of UTC, the bucket builder emits an extra ``YYYY-MM``
+    row for entries that straddle the UTC range start into the prior
+    local month. Slicing to ``n`` after reversal drops it.
+
+    ``delta_cost_pct`` is computed per row vs the next-older row; the
+    oldest row's value is ``None`` (no prior to compare against).
+    """
+    rows: "tuple[MonthlyPeriodRow, ...]" = ()
+    aggregated: tuple = ()                    # tuple[BucketUsage, ...] — forward-ref kept untyped to avoid an import-time edge into the aggregator's BucketUsage shape.
+    total_cost_usd: float = 0.0
+    total_tokens: int = 0
+    period_start: "dt.datetime | None" = None
+    period_end: "dt.datetime | None" = None
+    display_tz_label: str = ""
+
+
+def build_monthly_view(entries, *, now_utc, n=12, display_tz=None):
+    """Build a ``MonthlyView`` for the trailing ``n`` calendar months
+    (spec §5.2).
+
+    Calls ``_aggregate_monthly``. Drops the boundary-spillover bucket
+    (mirrors ``_dashboard_build_monthly_periods``). Computes
+    ``delta_cost_pct`` per row vs the next-older row. Newest-first.
+
+    Totals (``total_cost_usd`` / ``total_tokens``) sum over the
+    truncated row set so the React panel sees the same number as the
+    CLI table footer would.
+    """
+    _agg = _load_lib("_lib_aggregators")
+    buckets = _agg._aggregate_monthly(entries, mode="auto", tz=display_tz)
+    if not buckets:
+        return MonthlyView(
+            rows=(), aggregated=(),
+            total_cost_usd=0.0, total_tokens=0,
+            period_start=None, period_end=now_utc,
+            display_tz_label=_display_tz_label(display_tz),
+        )
+
+    # Reverse for newest-first AND cap to n BEFORE the delta loop —
+    # boundary-spillover drop (see MonthlyView docstring).
+    buckets = list(reversed(buckets))[:n]
+    cur_label = (
+        now_utc.astimezone(display_tz) if display_tz is not None
+        # internal fallback: host-local intentional
+        else now_utc.astimezone()
+    ).strftime("%Y-%m")
+
+    rows = []
+    total_cost = 0.0
+    total_tok = 0
+    for i, b in enumerate(buckets):
+        prev = buckets[i + 1] if i + 1 < len(buckets) else None
+        delta = None
+        if prev is not None and prev.cost_usd > 0:
+            delta = (b.cost_usd - prev.cost_usd) / prev.cost_usd
+        rows.append(MonthlyPeriodRow(
+            label=b.bucket,                          # "YYYY-MM"
+            cost_usd=b.cost_usd,
+            total_tokens=b.total_tokens,
+            input_tokens=b.input_tokens,
+            output_tokens=b.output_tokens,
+            cache_creation_tokens=b.cache_creation_tokens,
+            cache_read_tokens=b.cache_read_tokens,
+            delta_cost_pct=delta,
+            is_current=(b.bucket == cur_label),
+            models=_model_breakdowns_to_models_late(
+                b.model_breakdowns, b.cost_usd,
+            ),
+        ))
+        total_cost += b.cost_usd
+        total_tok += b.total_tokens
+
+    # period_start = first day of the oldest visible month, UTC.
+    earliest_label = buckets[-1].bucket
+    yr, mo = earliest_label.split("-")
+    period_start = dt.datetime(
+        int(yr), int(mo), 1, tzinfo=dt.timezone.utc,
+    )
+    return MonthlyView(
+        rows=tuple(rows),
+        aggregated=tuple(buckets),
+        total_cost_usd=total_cost,
+        total_tokens=total_tok,
+        period_start=period_start,
+        period_end=now_utc,
+        display_tz_label=_display_tz_label(display_tz),
+    )
+
+
+# === WeeklyView + build_weekly_view (Task 9) ===============================
+
+
+@dataclass(frozen=True)
+class WeeklyView:
+    """Weekly domain view — subscription-week aligned, newest-first.
+
+    ``rows`` carries the typed ``WeeklyPeriodRow`` tuple (already
+    overlaid with ``weekly_usage_snapshots``); ``aggregated`` carries
+    the parallel ``BucketUsage`` tuple for CLI byte-stability;
+    ``overlay`` carries ``(used_pct, dollar_per_pct)`` tuples in the
+    same order as ``aggregated`` (drives the CLI's
+    ``_render_weekly_table`` / ``_weekly_to_json``).
+
+    ``rows`` and ``aggregated`` are both newest-first. ``overlay``
+    aligns with ``aggregated``.
+    """
+    rows: "tuple[WeeklyPeriodRow, ...]" = ()
+    aggregated: tuple = ()                    # tuple[BucketUsage, ...] — forward-ref kept untyped to avoid an import-time edge into the aggregator's BucketUsage shape.
+    overlay: "tuple[tuple[float | None, float | None], ...]" = ()
+    total_cost_usd: float = 0.0
+    total_tokens: int = 0
+    period_start: "dt.datetime | None" = None
+    period_end: "dt.datetime | None" = None
+    display_tz_label: str = ""
+
+
+def build_weekly_view(conn, entries, *, weeks, now_utc, display_tz=None,
+                      as_of_utc=None):
+    """Build a ``WeeklyView`` from subscription-week boundaries
+    (spec §5.3).
+
+    ``weeks`` is the ``list[SubWeek]`` computed by the caller via
+    ``_compute_subscription_weeks(conn, range_start, range_end)``.
+
+    Calls ``_aggregate_weekly(entries, weeks)`` and overlays
+    ``weekly_usage_snapshots`` per ``WeekRef`` via
+    ``get_latest_usage_for_week`` (clamped to ``as_of_utc`` when
+    provided — the CLI passes ``range_end``-Z so historical
+    ``--until <past>`` queries pick the period-relevant snapshot, not
+    today's). Derives ``dollar_per_pct``, ``delta_cost_pct``,
+    ``is_current``.
+
+    Output is newest-first across ``rows`` / ``aggregated`` /
+    ``overlay`` — the CLI re-reverses for asc rendering.
+    """
+    _agg = _load_lib("_lib_aggregators")
+    _cct_core = _load_lib("_cctally_core")
+    buckets_asc = _agg._aggregate_weekly(entries, weeks)
+    if not buckets_asc:
+        return WeeklyView(
+            rows=(), aggregated=(), overlay=(),
+            total_cost_usd=0.0, total_tokens=0,
+            period_start=None, period_end=now_utc,
+            display_tz_label=_display_tz_label(display_tz),
+        )
+
+    # Index SubWeek by `start_date.isoformat()` — the invariant
+    # _aggregate_weekly enforces is one-to-one between bucket key and
+    # SubWeek.
+    week_by_key = {w.start_date.isoformat(): w for w in weeks}
+    parse_iso = _cct_core.parse_iso_datetime
+    make_ref = _cct_core.make_week_ref
+    get_usage = _cct_core.get_latest_usage_for_week
+
+    # Build asc overlay + asc WeeklyPeriodRow list first; reverse later.
+    asc_overlay: list = []
+    asc_rows: list = []
+    total_cost = 0.0
+    total_tok = 0
+    for i, b in enumerate(buckets_asc):
+        sw = week_by_key.get(b.bucket)
+        if sw is None:
+            # _aggregate_weekly invariant: every emitted bucket key maps
+            # 1:1 to a SubWeek in ``weeks``. Surface the invariant
+            # violation loudly rather than silently desynchronizing the
+            # three parallel lists (``asc_rows`` vs ``asc_overlay`` vs
+            # ``buckets_asc``) — under the prior defensive ``continue``
+            # branch, ``asc_overlay`` would advance one slot while
+            # ``asc_rows`` stayed put, creating a latent index-misalign
+            # for any consumer reading ``view.rows`` parallel to
+            # ``view.aggregated`` / ``view.overlay``.
+            raise AssertionError(
+                f"_aggregate_weekly emitted bucket without matching "
+                f"SubWeek (invariant violation): bucket={b.bucket!r}"
+            )
+        ref = make_ref(
+            week_start_date=sw.start_date.isoformat(),
+            week_end_date=sw.end_date.isoformat(),
+            week_start_at=sw.start_ts,
+            week_end_at=sw.end_ts,
+        )
+        usage_row = get_usage(conn, ref, as_of_utc=as_of_utc)
+        used_pct = None
+        if usage_row is not None and usage_row["weekly_percent"] is not None:
+            used_pct = float(usage_row["weekly_percent"])
+        dpp = (b.cost_usd / used_pct) if (used_pct and used_pct > 0) else None
+        asc_overlay.append((used_pct, dpp))
+
+        # delta_cost_pct vs the prior (older) bucket. asc order: prior
+        # is at index i - 1.
+        prev = buckets_asc[i - 1] if i > 0 else None
+        delta = None
+        if prev is not None and prev.cost_usd > 0:
+            delta = (b.cost_usd - prev.cost_usd) / prev.cost_usd
+
+        # is_current: now_utc falls inside [start_ts, end_ts).
+        try:
+            sw_start = parse_iso(sw.start_ts, "week.start_ts")
+            sw_end = parse_iso(sw.end_ts, "week.end_ts")
+            is_current = sw_start <= now_utc < sw_end
+        except ValueError:
+            # parse_iso_datetime raises ValueError on malformed ISO; any
+            # other exception is a genuine bug — let it propagate.
+            is_current = False
+
+        asc_rows.append(WeeklyPeriodRow(
+            label=sw.start_date.strftime("%m-%d"),
+            cost_usd=b.cost_usd,
+            total_tokens=b.total_tokens,
+            input_tokens=b.input_tokens,
+            output_tokens=b.output_tokens,
+            cache_creation_tokens=b.cache_creation_tokens,
+            cache_read_tokens=b.cache_read_tokens,
+            used_pct=used_pct,
+            dollar_per_pct=dpp,
+            delta_cost_pct=delta,
+            is_current=is_current,
+            models=_model_breakdowns_to_models_late(
+                b.model_breakdowns, b.cost_usd,
+            ),
+            week_start_at=sw.start_ts,
+            week_end_at=sw.end_ts,
+        ))
+        total_cost += b.cost_usd
+        total_tok += b.total_tokens
+
+    # Reverse to newest-first across all three parallel lists.
+    rows = list(reversed(asc_rows))
+    aggregated = list(reversed(buckets_asc))
+    overlay = list(reversed(asc_overlay))
+
+    period_start_dt = None
+    if weeks:
+        try:
+            period_start_dt = parse_iso(weeks[0].start_ts, "weeks[0].start_ts")
+        except ValueError:
+            # parse_iso_datetime raises ValueError on malformed ISO; any
+            # other exception is a genuine bug — let it propagate.
+            period_start_dt = None
+    return WeeklyView(
+        rows=tuple(rows),
+        aggregated=tuple(aggregated),
+        overlay=tuple(overlay),
+        total_cost_usd=total_cost,
+        total_tokens=total_tok,
+        period_start=period_start_dt,
+        period_end=now_utc,
+        display_tz_label=_display_tz_label(display_tz),
+    )
+
+
+# === TrendView + build_trend_view (Task 10) ================================
+
+
+@dataclass(frozen=True)
+class TrendView:
+    """Trend view — last n subscription weeks for cmd_report / TUI / dashboard
+    trend panel / share `_build_report_snapshot` (spec §4.2, §5.4).
+
+    Rows are typed ``TuiTrendRow`` (extended per spec §4.1 with 10
+    nullable fields so the same shape serves both the TUI's 7-field
+    surface and ``cmd_report``'s 11-field JSON contract).
+
+    ``avg_dollars_per_pct`` is the 3-sample-rule mean (None when fewer
+    than 3 rows carry a non-None ``dollars_per_percent``). The
+    dashboard envelope adapter emits it as ``trend.avg_dollars_per_pct``
+    so the React layer doesn't re-derive.
+
+    Row ordering matches ``cmd_report`` / ``_tui_build_trend``:
+    chronological (oldest first), suitable for the TUI sparkline left-
+    to-right walk + cmd_report's `--json` trend list (which is then
+    sorted by recency at the render site).
+    """
+    rows: "tuple[TuiTrendRow, ...]" = ()        # oldest-first
+    avg_dollars_per_pct: "float | None" = None
+    period_start: "dt.datetime | None" = None
+    period_end: "dt.datetime | None" = None
+    display_tz_label: str = ""
+
+
+def build_trend_view(conn, *, now_utc, n=8, display_tz=None):
+    """Build a ``TrendView`` of the last ``n`` subscription weeks
+    (spec §5.4).
+
+    Reads ``weekly_usage_snapshots`` for usage% per ``WeekRef``. Reads
+    ``weekly_cost_snapshots`` for cost — EXCEPT for weeks touched by a
+    reset event, where the builder bypasses the cache and live-
+    computes cost from ``session_entries`` via
+    ``_compute_cost_for_weekref(week_ref)``. This matches the existing
+    cmd_report path (``bin/cctally:~7969-7979``) and ``_tui_build_trend``
+    (``bin/_cctally_tui.py:~1515-1524``).
+
+    Rows are emitted oldest-first (chronological); each row populates
+    all 17 ``TuiTrendRow`` fields (7 historical + 10 extended).
+    ``delta_dpp`` is computed against the prior chronological row's
+    non-None dpp. ``spark_height`` is normalized 1..8 across the
+    window's valid dpp samples.
+
+    ``avg_dollars_per_pct`` follows the 3-sample rule (spec §4.3):
+    mean over non-None ``dollars_per_percent`` values iff at least 3
+    rows qualify; else None. Mirrors
+    ``_build_report_snapshot``'s historical behavior.
+    """
+    _cct_core = _load_lib("_cctally_core")
+    parse_iso = _cct_core.parse_iso_datetime
+    make_ref = _cct_core.make_week_ref
+    get_usage = _cct_core.get_latest_usage_for_week
+    c = _cctally()
+
+    week_refs = c.get_recent_weeks(conn, max(1, n))
+    if not week_refs:
+        return TrendView(
+            rows=(), avg_dollars_per_pct=None,
+            period_start=None, period_end=now_utc,
+            display_tz_label=_display_tz_label(display_tz),
+        )
+
+    # Determine current_key + current_week_start_at — same pattern as
+    # _tui_build_trend / cmd_report's Bug D handling.
+    latest_usage = conn.execute(
+        "SELECT week_start_date, week_end_date "
+        "FROM weekly_usage_snapshots "
+        "ORDER BY captured_at_utc DESC, id DESC LIMIT 1"
+    ).fetchone()
+    current_key = None
+    current_week_start_at = None
+    if latest_usage is not None and latest_usage["week_start_date"] is not None:
+        current_key = latest_usage["week_start_date"]
+        try:
+            canon_start, canon_end = c._get_canonical_boundary_for_date(
+                conn, latest_usage["week_start_date"]
+            )
+            current_ref = make_ref(
+                week_start_date=latest_usage["week_start_date"],
+                week_end_date=latest_usage["week_end_date"],
+                week_start_at=canon_start,
+                week_end_at=canon_end,
+            )
+            _adjusted = c._apply_reset_events_to_weekrefs(conn, [current_ref])
+            if _adjusted:
+                current_week_start_at = _adjusted[0].week_start_at
+        except Exception:
+            current_week_start_at = None
+
+    # Build a chronological (oldest-first) intermediate over week_refs.
+    # week_refs come newest-first from get_recent_weeks; reverse.
+    chrono = list(reversed(week_refs))
+
+    # Split-key set (Bug D): credited weeks appear twice in week_refs
+    # with identical WeekRef.key. Pin as_of_utc=week_end_at for those
+    # so each segment finds its own latest snapshot.
+    split_keys = {
+        r.key for r in week_refs
+        if sum(1 for x in week_refs if x.key == r.key) > 1
+    }
+
+    try:
+        _fresh_cfg = c._get_oauth_usage_config(c.load_config())
+    except Exception:
+        _fresh_cfg = c._get_oauth_usage_config({})
+
+    intermediate: list = []
+    for week_ref in chrono:
+        usage = get_usage(
+            conn, week_ref,
+            as_of_utc=(
+                week_ref.week_end_at if week_ref.key in split_keys else None
+            ),
+        )
+        usage_captured_at = usage["captured_at_utc"] if usage else None
+        if c._week_ref_has_reset_event(conn, week_ref):
+            cost_usd = c._compute_cost_for_weekref(week_ref)
+            cost_captured_at = (
+                usage_captured_at if cost_usd is not None else None
+            )
+            range_start_iso = week_ref.week_start_at
+            range_end_iso = week_ref.week_end_at
+        else:
+            cost = c.get_latest_cost_for_week(conn, week_ref)
+            cost_usd = float(cost["cost_usd"]) if cost else None
+            cost_captured_at = cost["captured_at_utc"] if cost else None
+            range_start_iso = (
+                cost["range_start_iso"] if cost and cost["range_start_iso"]
+                else None
+            )
+            range_end_iso = (
+                cost["range_end_iso"] if cost and cost["range_end_iso"]
+                else None
+            )
+        percent = float(usage["weekly_percent"]) if usage else None
+        ratio = (
+            cost_usd / percent
+            if (cost_usd is not None and percent and percent > 0)
+            else None
+        )
+        intermediate.append({
+            "week_ref": week_ref,
+            "used_pct": percent,
+            "cost_usd": cost_usd,
+            "dpp": ratio,
+            "usage_captured_at": usage_captured_at,
+            "cost_captured_at": cost_captured_at,
+            "range_start_iso": range_start_iso,
+            "range_end_iso": range_end_iso,
+        })
+
+    # Normalize dpp into spark heights 1..8 across the chrono window.
+    dpps = [d["dpp"] for d in intermediate if d["dpp"] is not None]
+    if dpps:
+        lo, hi = min(dpps), max(dpps)
+        span = (hi - lo) or 1e-9
+    else:
+        lo, hi, span = 0.0, 1.0, 1e-9
+
+    rows: list = []
+    prev_dpp: float | None = None
+    for d in intermediate:
+        week_ref = d["week_ref"]
+        percent = d["used_pct"]
+        cost_usd = d["cost_usd"]
+        dpp = d["dpp"]
+        usage_captured_at = d["usage_captured_at"]
+        cost_captured_at = d["cost_captured_at"]
+        range_start_iso = d["range_start_iso"]
+        range_end_iso = d["range_end_iso"]
+
+        delta = (
+            (dpp - prev_dpp)
+            if (dpp is not None and prev_dpp is not None) else None
+        )
+        spark = 1
+        if dpp is not None:
+            spark = int(round((dpp - lo) / span * 7)) + 1
+            spark = max(1, min(8, spark))
+
+        # WeekRef.week_start is a date; build a tz-aware datetime.
+        if week_ref.week_start_at:
+            week_start_dt = parse_iso(week_ref.week_start_at, "week_start_at")
+            _format_dt = c.format_display_dt
+            week_label = _format_dt(
+                week_start_dt, display_tz, fmt="%b %d", suffix=False,
+            )
+        else:
+            week_start_dt = dt.datetime.combine(
+                week_ref.week_start, dt.time(0, 0), dt.timezone.utc,
+            )
+            week_label = week_ref.week_start.strftime("%b %d")
+
+        is_cur = (
+            current_key is not None
+            and week_ref.key == current_key
+            and (
+                current_week_start_at is None
+                or week_ref.week_start_at == current_week_start_at
+            )
+        )
+
+        # as_of = max(usage_captured_at, cost_captured_at).
+        usage_dt = c._parse_iso_datetime_optional(usage_captured_at)
+        cost_dt = c._parse_iso_datetime_optional(cost_captured_at)
+        if usage_dt and cost_dt:
+            as_of_dt = usage_dt if usage_dt >= cost_dt else cost_dt
+        else:
+            as_of_dt = usage_dt or cost_dt
+        as_of = as_of_dt.isoformat(timespec="seconds") if as_of_dt else None
+
+        freshness = None
+        if usage_captured_at:
+            age_s = c._seconds_since_iso(usage_captured_at)
+            if age_s is not None and age_s <= 86400:
+                freshness = {
+                    "label": c._freshness_label(age_s, _fresh_cfg),
+                    "captured_at": usage_captured_at,
+                    "age_seconds": int(age_s),
+                }
+
+        rows.append(TuiTrendRow(
+            week_label=week_label,
+            week_start_at=week_start_dt,
+            used_pct=percent,
+            dollars_per_percent=dpp,
+            delta_dpp=delta,
+            spark_height=spark,
+            is_current=is_cur,
+            # Extended fields (spec §4.1)
+            week_start_date=week_ref.week_start,
+            week_end_date=week_ref.week_end,
+            week_end_at=(
+                parse_iso(week_ref.week_end_at, "week_end_at")
+                if week_ref.week_end_at else None
+            ),
+            weekly_cost_usd=cost_usd,
+            usage_captured_at=usage_captured_at,
+            cost_captured_at=cost_captured_at,
+            as_of=as_of,
+            range_start_iso=range_start_iso,
+            range_end_iso=range_end_iso,
+            freshness=freshness,
+        ))
+        if dpp is not None:
+            prev_dpp = dpp
+
+    # 3-sample average rule (spec §4.3): mean of non-None dpps iff at
+    # least 3 samples qualify.
+    valid_dpps = [r.dollars_per_percent for r in rows
+                  if r.dollars_per_percent is not None]
+    avg = (sum(valid_dpps) / len(valid_dpps)) if len(valid_dpps) >= 3 else None
+
+    return TrendView(
+        rows=tuple(rows),
+        avg_dollars_per_pct=avg,
+        period_start=None,
+        period_end=now_utc,
+        display_tz_label=_display_tz_label(display_tz),
+    )
+
+
+# === SessionsView + build_sessions_view (Task 13) ==========================
+
+
+@dataclass(frozen=True)
+class SessionsView:
+    """Sessions domain view — Claude sessions (merged across resumes),
+    last-activity descending.
+
+    Dual-shape per the spec §6.5 pattern that the daily/monthly/weekly
+    views established:
+
+    - ``rows`` carries the typed ``TuiSessionRow`` tuple consumed by the
+      TUI sessions panel and the dashboard session-detail surface.
+    - ``aggregated`` carries the parallel ``ClaudeSessionUsage`` tuple
+      consumed by ``cmd_session`` (CLI table + ``--json``) and the
+      share ``_build_session_snapshot`` (needs ``source_paths``,
+      ``model_breakdowns``, ``last_activity`` — fields ``TuiSessionRow``
+      doesn't carry).
+
+    Both shapes derive from the SAME ``_aggregate_claude_sessions``
+    call so the resumed-session merge invariant (``CLAUDE.md`` "Cost /
+    weekly / session" gotcha block — a ``sessionId`` across multiple
+    JSONL files collapses into ONE row) is preserved end-to-end:
+    ``rows[i]`` and ``aggregated[i]`` describe the same merged
+    sessionId.
+
+    ``total_sessions == len(rows) == len(aggregated)`` always (spec
+    §4.3). Empty entries → ``rows=()``, ``aggregated=()``, totals
+    zero — no exceptions on empty input.
+
+    ``limit=None`` keeps the full aggregator output (CLI use case —
+    ``cctally session`` has no ``--limit`` flag and emits everything in
+    the date range). ``limit`` ≥ 1 truncates BOTH parallel tuples to
+    the leading ``limit`` rows (TUI / dashboard use case — the
+    sessions pane promises "last N sessions"). The aggregator's
+    descending-by-last-activity sort means the leading rows are the
+    most recent; the TUI's ``[:100]`` cap stays semantic-stable.
+    """
+    rows: "tuple[TuiSessionRow, ...]" = ()
+    aggregated: tuple = ()              # tuple[ClaudeSessionUsage, ...] — forward-ref kept untyped to avoid an import-time edge into the aggregator's ClaudeSessionUsage shape.
+    total_sessions: int = 0
+    total_cost_usd: float = 0.0
+    period_start: "dt.datetime | None" = None
+    period_end: "dt.datetime | None" = None
+    display_tz_label: str = ""
+
+
+def build_sessions_view(entries, *, now_utc, limit=None, display_tz=None):
+    """Build a ``SessionsView`` from joined Claude session entries
+    (spec §5.5).
+
+    ``entries`` is the ``list[_JoinedClaudeEntry]`` from
+    ``get_claude_session_entries(range_start, range_end)``. The caller
+    controls the date window + ``skip_sync`` semantics; the builder
+    does no I/O of its own beyond what ``_aggregate_claude_sessions``
+    already does (cost recompute from ``CLAUDE_MODEL_PRICING``).
+
+    Per-row derivations (mirror today's ``_tui_build_sessions``
+    inline body):
+      - ``duration_minutes`` = ``(last_activity - first_activity)`` in
+        minutes (float).
+      - ``cache_hit_pct`` = ``cache_read / (input + cache_creation +
+        cache_read) * 100`` when the denominator is positive; ``None``
+        otherwise.
+      - ``model_primary`` = first model in the session's first-seen
+        order; ``"—"`` if the session somehow has no models (defensive
+        — the aggregator only emits sessions with at least one entry).
+      - ``project_label`` = ``os.path.basename(project_path)`` or
+        ``project_path`` itself when the basename is empty (root paths).
+
+    ``period_start`` is set to ``now_utc - 365d`` to match
+    ``_tui_build_sessions``' bounded scan window — strictly cosmetic
+    metadata (the caller's actual date range owns the entries fetched).
+    Share / CLI consumers prefer the caller-supplied range; this field
+    is informational only.
+    """
+    import os as _os                # late: keep top-level imports lean.
+    _agg = _load_lib("_lib_aggregators")
+    aggregated = _agg._aggregate_claude_sessions(entries)
+    # Apply limit truncation up front so `rows` and `aggregated` stay
+    # in lockstep (spec §4.3 invariant: `total_sessions == len(rows)
+    # == len(aggregated)`). limit=None → keep everything.
+    if limit is not None:
+        aggregated = aggregated[:limit]
+
+    rows = []
+    total_cost = 0.0
+    for s in aggregated:
+        duration_min = (
+            (s.last_activity - s.first_activity).total_seconds() / 60.0
+        )
+        denom = s.input_tokens + s.cache_creation_tokens + s.cache_read_tokens
+        cache_pct = (
+            (s.cache_read_tokens / denom * 100.0) if denom > 0 else None
+        )
+        rows.append(TuiSessionRow(
+            started_at=s.first_activity,
+            duration_minutes=duration_min,
+            model_primary=(s.models[0] if s.models else "—"),
+            cost_usd=s.cost_usd,
+            cache_hit_pct=cache_pct,
+            project_label=(
+                _os.path.basename(s.project_path) or s.project_path
+            ),
+            session_id=s.session_id,
+        ))
+        total_cost += s.cost_usd
+
+    return SessionsView(
+        rows=tuple(rows),
+        aggregated=tuple(aggregated),
+        total_sessions=len(rows),
+        total_cost_usd=total_cost,
+        period_start=(now_utc - dt.timedelta(days=365)),
+        period_end=now_utc,
+        display_tz_label=_display_tz_label(display_tz),
+    )