Mert Koseoglu 46796db1b0 feat(stats/B3b-slices3.2-3.6): wire multi-adapter renderer into formatReport ...

formatReport now accepts an optional `multiAdapter: MultiAdapterLifetimeStats`
opt (Engineer #74's getMultiAdapterLifetimeStats output). When supplied:

  3.2 — emits "Where it came from (tools you actually used —
        fixtures + probes filtered):" sub-block with per-adapter rows
        showing Captures / Indexed / Total kept out, marketing labels
        instead of raw IDs, sorted by total bytes desc, em-dash for
        zero captures (matches Mert-approved demo design).
  3.3 — adds a "Skipped (N): name1, name2, ..." disclosure line under
        the table for adapters that have DBs on disk but failed the
        isReal filter (event count / project count / recency / avg
        bytes), with the "no real chat activity" rationale.
  3.4 — opening tagline switches to "across N AI tools" when >= 2 real
        adapters are present, falls back to "in {marketing-label}"
        when only one. Days-since-first-event prefix when known.
  3.5 — backward compat preserved: when `multiAdapter` is omitted the
        renderer's output is byte-identical to before. No load-bearing
        display strings touched in tests/session/stats-output-format.test.ts.
  3.6 — "All your work" header is promoted to "All your work everywhere"
        and pulls totalEvents / totalSessions from the multi-adapter
        sums when supplied, so the receipt scope matches the rendered
        per-adapter breakdown.

Wiring lives in all three formatReport paths (active session, fresh
session, per-conversation) so behaviour is consistent regardless of
which branch executes.

Test coverage: tests/session/multi-adapter-render.test.ts (8 new tests,
all GREEN). Engineer #74's tests/session/multi-adapter-stats.test.ts
keeps passing untouched (16 tests). Full suite: 2546 passing
(+28 vs baseline 2518), 12 failing (same 12 pre-existing baseline),
20 skipped.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

2026-05-10 13:35:16 +03:00

9.3 KiB

Raw Permalink Blame History

View Raw