Use AND strategy with lowercase embeddings for thought dedup (jaredlockhart#884)

OR strategy produced false positives from common short words ("2026",
"AI", "agent") matching via TCR on short titles. Switched to AND
(both TCR >= 0.6 AND embedding >= 0.6 required) which eliminates
all false positives while catching real duplicates.

Also lowercase titles before embedding so casing doesn't affect
similarity (e.g., "THE GHOST IN THE SHELL" vs "Ghost in the Shell"
was 0.381, now 0.652 after lowercasing).

Lowered THOUGHT_DEDUP_EMBEDDING_THRESHOLD default from 0.80 to 0.60
since title embeddings score lower than full-content embeddings.

Co-authored-by: Jared Lockhart <119884+jaredlockhart@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: 3 people

penny/penny/agents/thinking.py

@@ -186,7 +186,7 @@ async def after_run(self, user: str) -> bool:
         if report and not await self._is_duplicate_thought(user, report):
             title, content = self._parse_title(report)
             content_embedding = await self._embed_and_serialize(content)
-            title_embedding = await self._embed_and_serialize(title) if title else None
+            title_embedding = await self._embed_and_serialize(title.lower()) if title else None
             self.db.thoughts.add(
                 user,
                 content,
@@ -273,7 +273,7 @@ async def _is_duplicate_thought(self, user: str, report: str) -> bool:
         if not title:
             return False
         title_vec = (
-            await embed_text(self._embedding_model_client, title)
+            await embed_text(self._embedding_model_client, title.lower())
             if self._embedding_model_client
             else None
         )
@@ -284,7 +284,7 @@ async def _is_duplicate_thought(self, user: str, report: str) -> bool:
             title,
             title_vec,
             existing_items,
-            DedupStrategy.TCR_OR_EMBEDDING,
+            DedupStrategy.TCR_AND_EMBEDDING,
             embedding_threshold=self.config.runtime.THOUGHT_DEDUP_EMBEDDING_THRESHOLD,
             tcr_threshold=self.config.runtime.THOUGHT_DEDUP_TCR_THRESHOLD,
         )

penny/penny/config_params.py

@@ -232,9 +232,9 @@ def _validate_unit_float(value: str) -> float:
 
 ConfigParam(
     key="THOUGHT_DEDUP_EMBEDDING_THRESHOLD",
-    description="Embedding similarity threshold for thought deduplication (0-1)",
+    description="Embedding similarity threshold for thought title deduplication (0-1)",
     type=float,
-    default=0.80,
+    default=0.60,
     validator=_validate_unit_float,
     group=GROUP_INNER_MONOLOGUE,
 )