From 74c3fac2cc569369a2235e493c95832c77815f40 Mon Sep 17 00:00:00 2001
From: jze9 <jze9programer@gmail.com>
Date: Thu, 28 May 2026 20:39:28 +0500
Subject: [PATCH] Add deduplication to store_memory

Before inserting, check if a near-identical memory exists (cosine > 0.85).
If found, skip insert. Prevents flooding the DB with repeated facts
from the same conversation.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 memory.py | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/memory.py b/memory.py
index acd1267..e90393a 100644
--- a/memory.py
+++ b/memory.py
@@ -25,6 +25,16 @@ async def store_memory(content: str, user_id: str = "default",
     pool = await get_pool()
     vec = await embed(content)
     async with pool.acquire() as conn:
+        # skip if near-duplicate already exists (cosine > 0.85)
+        existing = await conn.fetchval("""
+            SELECT id FROM memories
+            WHERE user_id = $1
+              AND (metadata->>'archived') IS NULL
+              AND 1 - (embedding <=> $2::vector) > 0.85
+            LIMIT 1
+        """, user_id, _vec_str(vec))
+        if existing:
+            return
         await conn.execute("""
             INSERT INTO memories (user_id, content, embedding, importance, memory_type, metadata)
             VALUES ($1, $2, $3::vector, $4, $5, $6)