session-memory: Phase 4 Measure — baseline, effectiveness, trend (WP-0009)

Closes the loop. metrics.py: fleet metrics (infra-overhead share, error rate,
schema-thrash, token percentiles, success) + persisted baseline trend. effect.py:
before/after per-pattern effectiveness with an improved verdict per metric.
measure entrypoint with trend + --since effectiveness + JSON. Recorded pre-fix
baseline: 27 sessions, overhead median 11.7%, error rate 0.96, schema-thrash 8.
13 new tests; suite 139/139. Capture->Detect->Curate->Distribute->Measure complete.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

session_memory/measure/__main__.py

@@ -0,0 +1,101 @@
+"""Measure entrypoint (T03): fleet trend + per-pattern effectiveness.
+
+    python -m session_memory.measure [--config PATH] [--label L] [--since DATE]
+                                     [--no-save] [--json]
+
+Computes current fleet metrics over the real (quality-filtered) sessions, appends
+them to the baseline trend, and reports whether the fleet is getting cheaper /
+more reliable over time (FR-M3). With ``--since DATE`` it also reports before/after
+effectiveness around a change (FR-M1/FR-M2).
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+
+from ..core.store import Store
+from ..detect.quality import filter_real, quality_config
+from ..ingest import _expand, load_config
+from .effect import effectiveness
+from .metrics import load_baselines, save_baseline, snapshot
+
+_TREND_KEYS = ("infra_overhead_share_median", "error_rate", "schema_thrash_sessions",
+               "tokens_p50", "success_rate")
+
+
+def real_digests(config: dict) -> list[dict]:
+    s = config.get("store", {})
+    store = Store(_expand(s["db_path"]), _expand(s["blob_dir"]))
+    out = filter_real(store.list_digests(), quality_config(config))
+    store.close()
+    return out
+
+
+def _fmt_trend(baselines: list[dict]) -> str:
+    if not baselines:
+        return "  (no prior snapshots)"
+    lines = []
+    recent = baselines[-5:]
+    for b in recent:
+        when = (b.get("captured_at") or "")[:10]
+        lbl = f" {b['label']}" if b.get("label") else ""
+        lines.append(f"  {when}{lbl}: overhead_med={b.get('infra_overhead_share_median')} "
+                     f"err_rate={b.get('error_rate')} schema_thrash={b.get('schema_thrash_sessions')} "
+                     f"tok_p50={b.get('tokens_p50')} success={b.get('success_rate')} "
+                     f"(n={b.get('n_sessions')})")
+    return "\n".join(lines)
+
+
+def _report(current: dict, baselines: list[dict], eff: dict | None) -> str:
+    lines = [f"# Fleet metrics  (n={current.get('n_sessions')} real sessions)"]
+    for k in _TREND_KEYS:
+        lines.append(f"  {k} = {current.get(k)}")
+    lines.append("\n## Trend (recent snapshots)")
+    lines.append(_fmt_trend(baselines))
+    if eff is not None:
+        lines.append(f"\n## Effectiveness since {eff['applied_at']} "
+                     f"(before={eff['n_before']}, after={eff['n_after']})")
+        if eff["insufficient_data"]:
+            lines.append("  insufficient data on one side of the date")
+        else:
+            for k in _TREND_KEYS:
+                d = eff["deltas"].get(k, {})
+                mark = {True: "improved", False: "worse", None: "—"}[d.get("improved")]
+                lines.append(f"  {k}: {d.get('before')} -> {d.get('after')} "
+                             f"({d.get('change'):+}) {mark}")
+    return "\n".join(lines)
+
+
+def main(argv=None) -> int:
+    here = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+    ap = argparse.ArgumentParser(description="Measure fleet metrics + per-pattern effectiveness.")
+    ap.add_argument("--config", default=os.path.join(here, "config.toml"))
+    ap.add_argument("--label", default="")
+    ap.add_argument("--since", default=None, help="ISO date for before/after effectiveness")
+    ap.add_argument("--no-save", action="store_true", help="don't append to the baseline trend")
+    ap.add_argument("--json", action="store_true")
+    args = ap.parse_args(argv)
+
+    config = load_config(args.config)
+    digests = real_digests(config)
+    current = snapshot(digests, label=args.label)
+
+    path = _expand(config.get("measure", {}).get("baselines", "session_memory/measure/baselines.jsonl"))
+    prior = load_baselines(path)
+    if not args.no_save:
+        save_baseline(current, path)
+
+    eff = effectiveness(digests, args.since, label=args.label) if args.since else None
+
+    if args.json:
+        print(json.dumps({"current": current, "trend": prior + [current], "effectiveness": eff},
+                         indent=2))
+    else:
+        print(_report(current, prior + [current], eff))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())