feat: FC links a conclusive pair itself instead of asking (4392)

Operator, 2026-09-24: *"I don't want this to be manual that defeats the
convenience that I'm going for."*

Confirm-only was right while every signal was circumstantial. Time proximity
and a body that mentions Discord can never be more than suggestive, so asking
was the honest response to what FC actually knew. A shared working name is
different in kind: when the creator's own name for a piece appears in exactly
these two posts and nowhere else in their library, there is nothing left for
the operator to adjudicate, and asking is a chore FC invented for them.

AUTO_LINK_FLOOR is 1.0 and sits deliberately above IDENTITY_FLOOR's 0.75. The
gap between them IS the review queue — real evidence, not certain enough for
FC to act on alone. Measured on artist 8, of 15 name-sharing pairs: 11 are
conclusive, 2 more propose, 2 fall short of both.

Three refusals, because an auto-link is FC asserting something the operator
never saw:

  * Exactly one candidate may be conclusive. Two is not a tie to be broken by
    score — one post can tease two pieces dropped separately, and then each
    drop carries a different name from the same teaser, each individually
    conclusive. Two conclusive answers to "which drop is this" means the
    question was wrong, so both queue and FC says nothing.
  * Neither end may already be claimed by an accepted link. That link is the
    operator's decision and reassigning its other end would overrule them
    silently.
  * The whole thing is one setting, defaulting on, reversible in the UI —
    an accepted link is a row they can dismiss.

`match_post` now returns (proposed, linked) so a sweep can report what it did
on its own rather than only what it queued.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LVjrnpQjRgHdvq95rASoiR

backend/app/api/settings.py

@@ -44,6 +44,7 @@ _EDITABLE_FIELDS = (
     "discord_link_enabled",
     "discord_link_threshold",
     "discord_link_window_hours",
+    "discord_link_auto",
     "extdl_mega_enabled",
     "extdl_gdrive_enabled",
     "extdl_mediafire_enabled",
@@ -165,6 +166,10 @@ async def update_import_settings():
         body["discord_link_enabled"], bool
     ):
         return jsonify({"error": "discord_link_enabled must be a boolean"}), 400
+    if "discord_link_auto" in body and not isinstance(
+        body["discord_link_auto"], bool
+    ):
+        return jsonify({"error": "discord_link_auto must be a boolean"}), 400
     if "discord_link_threshold" in body:
         v = body["discord_link_threshold"]
         if not isinstance(v, (int, float)) or isinstance(v, bool) or v < 0 or v > 1:

backend/app/models/import_settings.py

@@ -139,6 +139,25 @@ class ImportSettings(Base):
         server_default="24",
     )
 
+    # Whether FC links a CONCLUSIVE pair without asking.
+    #
+    # Operator, 2026-09-24: *"I don't want this to be manual that defeats the
+    # convenience that I'm going for."* Confirm-only was the right default
+    # while the only signals were circumstantial — proximity and a body that
+    # mentions Discord can never be more than suggestive, and asking was the
+    # honest response to that. A shared working name is different in kind: when
+    # the name appears in exactly these two posts and nowhere else in the
+    # artist's library, there is nothing for the operator to adjudicate.
+    #
+    # Only the conclusive band is affected (post_association_service.
+    # AUTO_LINK_FLOOR). Everything weaker still queues, and an accepted link is
+    # a row the operator can dismiss, so this is reversible in the UI rather
+    # than only in the database.
+    discord_link_auto: Mapped[bool] = mapped_column(
+        Boolean, nullable=False, default=True,
+        server_default="true",
+    )
+
     # #830 off-platform file-host downloads — per-host enable lever (default on,
     # rule #26). Column names are extdl_<host>_enabled so the worker reads them
     # via getattr(settings, f"extdl_{host}_enabled", True).

backend/app/services/post_association_service.py

@@ -129,6 +129,20 @@ MAX_CANDIDATES = 25
 # the manual button's job.
 MAX_RECENT_DROPS = 200
 
+# What a shared name must reach before FC links a pair WITHOUT asking.
+#
+# 1.0, which under post_naming's post-span counting means the name appears in
+# exactly these two posts and nowhere else in the artist's library. That is not
+# "strong evidence" — within the library it is conclusive, and the remaining
+# ways to be wrong are a mis-parse or the creator reusing a name for a genuinely
+# different piece on the same day.
+#
+# Deliberately above IDENTITY_FLOOR, which is what a name needs to PROPOSE.
+# The gap between them is the review queue: real evidence, not certain enough
+# for FC to act on by itself. Measured on artist 8, 15 name-sharing pairs: 11
+# are conclusive, 2 more propose, 2 fall short of both.
+AUTO_LINK_FLOOR = 1.0
+
 
 @dataclass(frozen=True)
 class _Corpus:
@@ -272,14 +286,38 @@ class PostAssociationService:
             .limit(MAX_CANDIDATES)
         )).scalars().all()
 
+    async def _claimed(self, announcement_id: int, payload_id: int) -> bool:
+        """Is either end of this pair already spoken for by an accepted link?
+
+        An auto-link is FC asserting something the operator never saw, so it
+        only happens where there is nothing to contradict. A drop already
+        linked to a different announcement is exactly such a contradiction, and
+        resolving it is a judgement about which one is right — which is the
+        operator's, not FC's.
+        """
+        return (await self.session.execute(
+            select(PostAssociation.id).where(
+                PostAssociation.status == "linked",
+                or_(
+                    PostAssociation.payload_post_id == payload_id,
+                    PostAssociation.announcement_post_id == announcement_id,
+                ),
+            ).limit(1)
+        )).scalar() is not None
+
     async def match_post(
         self, announcement_id: int, *, threshold: float, window_hours: float,
-    ) -> int:
-        """Score one announcement against nearby groupings. Returns proposals made."""
+        auto_link: bool = False,
+    ) -> tuple[int, int]:
+        """Score one announcement against nearby groupings.
+
+        Returns `(proposed, linked)` — how many pairs were written, and how
+        many of those were linked outright rather than queued.
+        """
         announcement = await self.session.get(Post, announcement_id)
         if announcement is None or announcement.synthesized_by is not None:
             # A synthetic post cannot announce anything — FC wrote it.
-            return 0
+            return 0, 0
 
         window = timedelta(hours=window_hours)
         declared = declared_signal(announcement.description)
@@ -289,6 +327,7 @@ class PostAssociationService:
         here_text = corpus.text_by_post.get(announcement.id, "")
 
         made = 0
+        scored: list[tuple[Post, float, dict, float]] = []
         for group in await self._candidate_groups(announcement, window=window):
             if group.id in already:
                 continue
@@ -335,15 +374,35 @@ class PostAssociationService:
                 # Carried so the queue can say WHY. A review queue that cannot
                 # explain itself is one the operator learns to click through.
                 signals["identity_token"] = token
+            scored.append((group, score, signals, identity))
+
+        # Who, if anyone, FC links without asking.
+        #
+        # EXACTLY ONE candidate may be conclusive. Two drops sharing a name
+        # with one teaser at full strength is not a tie to be broken by score —
+        # it means the name identifies something other than what FC thinks it
+        # does, and the right response is to queue both and say nothing.
+        auto_id = None
+        if auto_link:
+            conclusive = [c for c in scored if c[3] >= AUTO_LINK_FLOOR]
+            if len(conclusive) == 1 and not await self._claimed(
+                announcement.id, conclusive[0][0].id
+            ):
+                auto_id = conclusive[0][0].id
+
+        linked = 0
+        for group, score, signals, identity in scored:
+            status = "linked" if group.id == auto_id else "pending"
+            linked += status == "linked"
             self.session.add(PostAssociation(
                 announcement_post_id=announcement.id,
                 payload_post_id=group.id,
                 score=score,
                 signals=signals,
-                status="pending",
+                status=status,
             ))
             made += 1
-        return made
+        return made, linked
 
     async def list_pending(self) -> list[dict]:
         rows = (await self.session.execute(
@@ -412,7 +471,7 @@ async def rescan(session: AsyncSession, *, now: datetime | None = None) -> dict:
     """Score every recent non-synthetic post against nearby groupings."""
     settings = await ImportSettings.load(session)
     if not settings.discord_link_enabled:
-        return {"enabled": False, "scanned": 0, "proposed": 0}
+        return {"enabled": False, "scanned": 0, "proposed": 0, "linked": 0}
 
     now = now or datetime.now(UTC)
     window_hours = float(settings.discord_link_window_hours)
@@ -466,17 +525,25 @@ async def rescan(session: AsyncSession, *, now: datetime | None = None) -> dict:
 
     svc = PostAssociationService(session)
     proposed = 0
+    linked = 0
     # Sorted because `ids` is now a union of two queries: set iteration order
     # is arbitrary, and a sweep that visits posts in a different order each
     # run is one whose failures cannot be reproduced.
     for pid in sorted(ids):
-        proposed += await svc.match_post(
+        made, auto = await svc.match_post(
             pid,
             threshold=float(settings.discord_link_threshold),
             window_hours=window_hours,
+            auto_link=bool(settings.discord_link_auto),
         )
+        proposed += made
+        linked += auto
     log.info(
-        "discord announcement matcher: scanned %d post(s), proposed %d pair(s)",
-        len(ids), proposed,
+        "discord announcement matcher: scanned %d post(s), proposed %d pair(s), "
+        "linked %d outright",
+        len(ids), proposed, linked,
     )
-    return {"enabled": True, "scanned": len(ids), "proposed": proposed}
+    return {
+        "enabled": True, "scanned": len(ids), "proposed": proposed,
+        "linked": linked,
+    }

backend/app/tasks/maintenance.py

@@ -1184,12 +1184,16 @@ def group_discord_drops() -> str:
     soft_time_limit=900, time_limit=1200,
 )
 def match_post_associations() -> str:
-    """Milestone 388 E5: propose which Patreon post announced which Discord drop.
+    """Milestone 388 E5: which Patreon post announced which Discord drop.
 
-    Proposes only — every pair lands in a review queue and nothing is linked
-    until the operator accepts. Maintenance lane for the same reason as the
-    grouper: no inference, no ML library, and it must not depend on the
-    optional ml-worker being present.
+    A CONCLUSIVE pair — one the creator's own working name identifies, where
+    that name appears in these two posts and nowhere else in their library — is
+    linked outright when `discord_link_auto` is on, because there is nothing
+    there for the operator to adjudicate. Everything weaker lands in the review
+    queue and stays unlinked until they accept it.
+
+    Maintenance lane for the same reason as the grouper: no inference, no ML
+    library, and it must not depend on the optional ml-worker being present.
     """
     import asyncio