alembic/versions/0097_retrieval_log_best_available_id.py

@@ -0,0 +1,60 @@
+"""add retrieval_logs.best_available_id — WHICH record the bar refused (#3807)
+
+Revision ID: 0097
+Revises: 0096
+Create Date: 2026-09-09
+
+0096 added `best_available_score` so a threshold could be judged from what it
+rejected. It records how CLOSE the bar came to firing and not WHAT it turned
+away, and that turns out to be the half a decision actually needs.
+
+Live, `pre_tool_rule` shows a bar of ~0.72 with a near-miss p90 of 0.7071 —
+about 117 declines a day sitting within 0.013 of firing. Dropping the bar to
+0.707 would take that arm from 22 hits a day to roughly 139, a six-fold change
+on a surface that runs before every Bash call. The percentile says the mass is
+there. Nothing says whether it is worth showing.
+
+AND THE TWO OBVIOUS INSTRUMENTS DO NOT ANSWER IT. Pull-through cannot: the
+injected rule line already carries title and trigger, so a session can comply
+without ever calling `get_rule`, and rule pull-through therefore understates
+usefulness by construction. Reading the rejected records can — and `result_ids`
+holds only what was RETURNED, so on a zero-result call it is empty and the
+near-missed record has no name.
+
+So: the id, beside the score, from the SAME ranked candidate. The two must
+never be able to describe different records — a score paired with its
+neighbour's id would be worse than no id at all, because it invites a reader to
+judge the wrong record and conclude the bar is fine.
+
+NULLABLE AND UNBACKFILLED, for the reason 0095 and 0096 both spell out: a row
+written before this genuinely does not know, and inventing a value would put an
+artifact where a measurement belongs. Null here means "not measured", never
+"nothing was close".
+
+NOT A FOREIGN KEY, deliberately. `retrieval_logs` spans record types — the
+rules arms store rule ids, the note arms store note ids — and `source` is what
+says which table an id belongs to, exactly as `result_ids` has always worked.
+A constraint would have to point at one table and would be wrong for the other.
+
+Downgrade drops the column. Purely observational — nothing reads it for
+correctness.
+"""
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "0097"
+down_revision = "0096"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "retrieval_logs",
+        sa.Column("best_available_id", sa.Integer(), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("retrieval_logs", "best_available_id")

src/scribe/mcp/tools/search.py

@@ -128,6 +128,7 @@ async def search(
         project_id=project_id or None, is_task=is_task, results=raw,
         duration_ms=(time.perf_counter() - t0) * 1000.0,
         best_available=report.get("best_available_score"),
+        best_available_id=report.get("best_available_id"),
         searched=bool(report.get("searched", True)),
     )
     owners = await owner_names_for(
@@ -153,7 +154,9 @@ async def search(
     }
 
 
-async def retrieval_telemetry(days: int = 30) -> dict:
+async def retrieval_telemetry(
+    days: int = 30, near_miss_samples: int = 0,
+) -> dict:
     """What the retrieval telemetry says about YOUR surfaces, over a window.
 
     The read half of the loop the ranker's thresholds are meant to be tuned
@@ -181,6 +184,24 @@ async def retrieval_telemetry(days: int = 30) -> dict:
     render as a zero-result call, and nothing else in this readout tells them
     apart.
 
+    `near_miss_samples` (0-20, default 0) TURNS THE PERCENTILES INTO RECORDS
+    YOU CAN READ. Each source then carries `near_miss_records`: its highest
+    scoring declines, each with the `record_id` the bar refused and the `query`
+    that asked. Reach for it whenever you are about to move a threshold.
+
+    THE PERCENTILE CANNOT SETTLE A BAR ON ITS OWN, and this is the whole reason
+    the parameter exists. `near_misses.p90` says mass is sitting just under the
+    line; it says nothing about whether that mass is RELEVANT, and those are
+    different questions. Lowering a bar to where the mass is, without reading
+    what is there, is choosing a firing rate rather than a quality. Pull-through
+    cannot referee it either — the injected rule line already carries title and
+    trigger, so a session can comply without ever calling `get_rule`, which
+    makes rule pull-through understate usefulness by construction. Reading the
+    rejected records is the method that actually answers it.
+
+    Off by default because it is a LISTING, not a statistic: it is for the
+    moment you are making a decision, not for every readout.
+
     `near_misses` is `null` when no declining call in the window measured it —
     rows written before #3670 shipped cannot know. That is "not measured", not
     "nothing came close"; a 0.0 there would be a claim about the corpus
@@ -320,8 +341,15 @@ It is an UPPER BOUND per surface: a pull records the door it came
 
     Args:
         days: window size, default 30. Clamped to at least 1.
+        near_miss_samples: 0-20, default 0. How many of each source's highest
+            scoring DECLINES to list by record, with the query that asked.
+            Pass it when you are about to move a threshold; leave it off
+            otherwise. See the near-miss section above for why a percentile
+            alone cannot settle a bar.
     """
-    return await retrieval_summary(current_user_id(), days=days)
+    return await retrieval_summary(
+        current_user_id(), days=days, near_miss_samples=near_miss_samples,
+    )
 
 
 def register(mcp) -> None:

src/scribe/models/retrieval_log.py

@@ -62,6 +62,15 @@ class RetrievalLog(Base):
     # close": a 0.0 there would read as a corpus with no relevant records at
     # all, which is an artifact standing in for a measurement.
     best_available_score: Mapped[float | None] = mapped_column(Float, nullable=True)
+    # WHICH record scored that, so a reader can judge what the bar refused
+    # rather than only how close it came (#3807). Written from the same ranked
+    # candidate as the score above — the two describing different records would
+    # be worse than no id at all, because it invites judging the wrong one.
+    #
+    # Not a foreign key on purpose: this table spans record types (the rule arms
+    # store rule ids, the note arms store note ids) and `source` is what says
+    # which, exactly as `result_ids` has always worked.
+    best_available_id: Mapped[int | None] = mapped_column(Integer, nullable=True)
     # [{"id": int, "score": float, "rank": int}, ...], highest-first.
     result_ids: Mapped[list] = mapped_column(JSONB, nullable=False, default=list)
     duration_ms: Mapped[float | None] = mapped_column(Float, nullable=True)

src/scribe/routes/search.py

@@ -58,6 +58,7 @@ async def search_route():
         project_id=project_id, is_task=is_task, results=results,
         duration_ms=(time.perf_counter() - t0) * 1000.0,
         best_available=report.get("best_available_score"),
+        best_available_id=report.get("best_available_id"),
         searched=bool(report.get("searched", True)),
     )
     owners = await owner_names_for(

src/scribe/services/embeddings.py

@@ -658,7 +658,12 @@ async def semantic_search_notes(
         # looked" and from "the query failed". Set here, at the one point past
         # which a real result set exists.
         report["searched"] = True
-        report["best_available_score"] = scored[0][0] if scored else None
+        # ONE unpack, so the score and the id cannot describe different records
+        # (#3807). Splitting these into two expressions is how a later edit
+        # pairs a score with its neighbour's id.
+        best = scored[0] if scored else None
+        report["best_available_score"] = best[0] if best else None
+        report["best_available_id"] = int(best[1].id) if best else None
     scored = [pair for pair in scored if pair[0] >= threshold]
     if not demote_superseded:
         return scored[:limit]
@@ -909,7 +914,10 @@ async def semantic_search_rules(
     if report is not None:
         # See the sibling search: absent means the search never ran (#3765).
         report["searched"] = True
-        report["best_available_score"] = ranked[0][0] if ranked else None
+        # One unpack — see the sibling search (#3807).
+        best = ranked[0] if ranked else None
+        report["best_available_score"] = best[0] if best else None
+        report["best_available_id"] = int(best[1].id) if best else None
     return [pair for pair in ranked if pair[0] >= threshold][:limit]
 
 

src/scribe/services/plugin_context.py

@@ -484,6 +484,7 @@ async def _reserve_slot_for_reuse(
         threshold=cfg["threshold"], limit=1, project_id=project_id,
         is_task=None, results=reuse,
         best_available=_rep.get("best_available_score"),
+        best_available_id=_rep.get("best_available_id"),
         searched=bool(_rep.get("searched", True)),
         duration_ms=(time.perf_counter() - _t0) * 1000.0,
     )
@@ -553,6 +554,7 @@ async def build_autoinject_hint(
         threshold=cfg["threshold"], limit=cfg["top_k"],
         project_id=(project_id or None), is_task=None, results=hits,
         best_available=_rep_ai.get("best_available_score"),
+        best_available_id=_rep_ai.get("best_available_id"),
         searched=bool(_rep_ai.get("searched", True)),
         duration_ms=(time.perf_counter() - t0) * 1000.0,
     )
@@ -1048,6 +1050,12 @@ async def build_write_path_hint(
             best_available=(
                 None if withheld_here else _rep_wp.get("best_available_score")
             ),
+            # Withheld on the SAME condition as the score. A surviving id
+            # beside a null score would name a record without saying what it
+            # scored, which is the pair disagreeing in the other direction.
+            best_available_id=(
+                None if withheld_here else _rep_wp.get("best_available_id")
+            ),
             searched=bool(_rep_wp.get("searched", True)),
             duration_ms=(time.perf_counter() - t0) * 1000.0,
         )
@@ -1345,6 +1353,7 @@ async def build_write_path_hint(
             project_id=project_id,
             is_task=None, results=fresh, duration_ms=rule_ms,
             best_available=_rep_wpr.get("best_available_score"),
+            best_available_id=_rep_wpr.get("best_available_id"),
             searched=bool(_rep_wpr.get("searched", True)),
             # What the ranker found and this session had already been told.
             # Without it a zero row cannot say whether the bar was too high or
@@ -1453,6 +1462,7 @@ async def build_tool_rule_hint(
             project_id=project_id,
             is_task=None, results=fresh, duration_ms=duration_ms,
             best_available=_rep_ptr.get("best_available_score"),
+            best_available_id=_rep_ptr.get("best_available_id"),
             searched=bool(_rep_ptr.get("searched", True)),
             # See the sibling arm. It matters more here: this arm fires on every
             # Bash call, so a long session excludes its way to an all-zero row

src/scribe/services/retrieval_telemetry.py

@@ -57,6 +57,7 @@ def _build_payload(
     duration_ms: float | None,
     suppressed: int | None = None,
     best_available: float | None = None,
+    best_available_id: int | None = None,
 ) -> dict:
     """Reduce a retrieval call to a flat, JSON-safe RetrievalLog payload.
 
@@ -96,6 +97,10 @@ def _build_payload(
         "best_available_score": (
             None if best_available is None else round(float(best_available), 5)
         ),
+        # The record that scored it, so a reader can go and look (#3807).
+        "best_available_id": (
+            None if best_available_id is None else int(best_available_id)
+        ),
         "result_ids": items,
         "duration_ms": (round(duration_ms, 2) if duration_ms is not None else None),
     }
@@ -135,6 +140,7 @@ def record_retrieval(
     duration_ms: float | None = None,
     suppressed: int | None = None,
     best_available: float | None = None,
+    best_available_id: int | None = None,
     searched: bool = True,
 ) -> None:
     """Fire-and-forget: record one retrieval call.
@@ -189,6 +195,7 @@ def record_retrieval(
             duration_ms=duration_ms,
             suppressed=suppressed,
             best_available=best_available,
+            best_available_id=best_available_id,
         )
     except Exception:
         logger.debug("retrieval telemetry payload build failed", exc_info=True)
@@ -347,7 +354,9 @@ def _coverage(complete_from, since) -> dict:
     }
 
 
-async def retrieval_summary(user_id: int | None, *, days: int = 30) -> dict:
+async def retrieval_summary(
+    user_id: int | None, *, days: int = 30, near_miss_samples: int = 0,
+) -> dict:
     """What the retrieval telemetry says, per surface, over a window.
 
     Three aggregates side by side, each read from the table built for it — NOT
@@ -512,6 +521,54 @@ async def retrieval_summary(user_id: int | None, *, days: int = 30) -> dict:
                 quiet.update(_coverage(first_row, since))
                 out["sources"][src] = quiet
 
+            # WHAT THE BAR REFUSED, by name (#3807). Opt-in, because it is a
+            # LISTING and not a statistic: an id cannot be percentiled, and a
+            # reader tuning a threshold needs to go and read the records rather
+            # than see another number about them. Off by default so the
+            # ordinary readout keeps its size.
+            #
+            # Deliberately NOT a window function. This module's one production
+            # outage (#2663) was a grouped query the database rejected, swallowed
+            # by the broad except, every counter reading zero while the mocked
+            # tests passed — and the lesson recorded then was to group on a raw
+            # column and classify in Python rather than push cleverness into the
+            # SQL. So: one flat ordered query, overfetched, bucketed here.
+            if near_miss_samples > 0:
+                want = max(1, min(int(near_miss_samples), 20))
+                rows = (
+                    await session.execute(
+                        select(
+                            RetrievalLog.source,
+                            RetrievalLog.best_available_score,
+                            RetrievalLog.best_available_id,
+                            RetrievalLog.query,
+                        )
+                        .where(
+                            declined,
+                            RetrievalLog.created_at >= since,
+                            RetrievalLog.user_id == user_id,
+                            RetrievalLog.best_available_id.isnot(None),
+                        )
+                        .order_by(RetrievalLog.best_available_score.desc())
+                        # Overfetch so every source can fill its own quota even
+                        # when one of them holds all the highest scores.
+                        .limit(want * 40)
+                    )
+                ).all()
+                for src, score, rec_id, q in rows:
+                    bucket = out["sources"].get(src)
+                    if bucket is None:
+                        continue
+                    samples = bucket.setdefault("near_miss_records", [])
+                    if len(samples) >= want:
+                        continue
+                    samples.append({
+                        "score": _round(score),
+                        "record_id": int(rec_id),
+                        # Enough to recognise the ask, not the whole prompt.
+                        "query": (q or "")[:120],
+                    })
+
             # The corpus side, at its own grain. `ambient` mirrors
             # note_usage.usage_for_notes: an ambient surfacing was not a scored
             # CHOICE, so folding it into pull-through would understate it.

tests/test_services_retrieval_telemetry.py

@@ -1120,3 +1120,112 @@ def test_a_caller_that_never_asked_is_assumed_to_have_searched():
             limit=10, project_id=None, is_task=None, results=[],
         )
         assert build.called
+
+
+# ─── the bar's refusals have names now (#3807) ───────────────────────────────
+#
+# #3670 recorded how CLOSE the bar came to firing. That is the half a decision
+# does not need: live, pre_tool_rule sits at a ~0.72 bar with a near-miss p90 of
+# 0.7071, so dropping to 0.707 would take it from 22 hits a day to roughly 139.
+# The percentile says the mass is there and says nothing about whether it is
+# worth showing, and pull-through cannot referee it because the injected rule
+# line already carries title and trigger — a session can comply without ever
+# calling get_rule.
+#
+# Reading the rejected records is the method that answers it, and until now
+# `result_ids` held only what was RETURNED, so on a zero-result call the
+# near-missed record had no name.
+
+
+def test_the_rejected_record_is_named_beside_the_score_it_scored():
+    """Both halves, from one payload, because either alone is unusable.
+
+    A score with no id says a bar nearly fired and not what it nearly fired
+    ABOUT. An id with no score names a record without saying how close it came.
+    """
+    p = _build_payload(
+        user_id=1, source="pre_tool_rule", query="git push --force",
+        threshold=0.72, limit=1, project_id=None, is_task=None,
+        results=[], duration_ms=None, best_available=0.7104,
+        best_available_id=168,
+    )
+    assert p["result_count"] == 0
+    assert p["best_available_score"] == 0.7104
+    assert p["best_available_id"] == 168
+
+
+def test_an_unmeasured_near_miss_names_nothing():
+    """Null, on both halves, and for the reason its sibling is null: a row that
+    did not measure must not invent a record any more than it invents a score."""
+    p = _build_payload(
+        user_id=1, source="auto_inject", query="q", threshold=0.6,
+        limit=3, project_id=None, is_task=None, results=[], duration_ms=None,
+    )
+    assert p["best_available_score"] is None
+    assert p["best_available_id"] is None
+
+
+@pytest.mark.integration
+@pytest.mark.asyncio
+async def test_the_listing_names_the_highest_declines_per_source(_dispose_engine):
+    """Integration, because the listing is a second query against real Postgres
+    and this module's one outage was a query the database rejected in silence.
+
+    Also pins that the listing is PER SOURCE. One flat `ORDER BY score DESC`
+    with a global limit would let a noisy source consume the whole quota and
+    leave the surface you are actually tuning unrepresented — which reads as
+    "nothing was close" for that source, the exact misreading this milestone
+    has spent itself correcting.
+    """
+    from sqlalchemy import delete
+
+    from scribe.models import async_session
+    from scribe.models.retrieval_log import RetrievalLog
+    from scribe.services.retrieval_telemetry import (
+        _insert_retrieval_log, retrieval_summary,
+    )
+
+    UID = 990080
+    # A source whose declines score HIGH, and one whose declines score low.
+    for score, rid in ((0.71, 501), (0.70, 502), (0.69, 503)):
+        await _insert_retrieval_log(_build_payload(
+            user_id=UID, source="pre_tool_rule", query=f"cmd {rid}",
+            threshold=0.72, limit=1, project_id=None, is_task=None,
+            results=[], duration_ms=1.0,
+            best_available=score, best_available_id=rid,
+        ))
+    await _insert_retrieval_log(_build_payload(
+        user_id=UID, source="auto_inject", query="a quieter ask",
+        threshold=0.6, limit=3, project_id=None, is_task=None,
+        results=[], duration_ms=1.0,
+        best_available=0.31, best_available_id=901,
+    ))
+
+    try:
+        out = await retrieval_summary(UID, days=30, near_miss_samples=2)
+        assert out["read_failed"] is False, (
+            "the listing query did not execute — a rejected query here reads "
+            "as an absent listing, which is #2663 again"
+        )
+
+        top = out["sources"]["pre_tool_rule"]["near_miss_records"]
+        assert [r["record_id"] for r in top] == [501, 502], (
+            "the listing must be the HIGHEST declines, in order, capped at the "
+            "requested count"
+        )
+        assert top[0]["score"] == pytest.approx(0.71, abs=1e-4)
+        assert top[0]["query"] == "cmd 501", "the ask is what makes a hit judgeable"
+
+        # The low-scoring source keeps its own slot rather than being crowded
+        # out by the high scorer — this is what a global LIMIT would break.
+        quiet = out["sources"]["auto_inject"]["near_miss_records"]
+        assert [r["record_id"] for r in quiet] == [901]
+
+        off = await retrieval_summary(UID, days=30)
+        assert "near_miss_records" not in off["sources"]["pre_tool_rule"], (
+            "the listing is opt-in; the ordinary readout must not grow"
+        )
+    finally:
+        async with async_session() as s:
+            await s.execute(delete(RetrievalLog).where(RetrievalLog.user_id == UID))
+            await s.commit()