From 0fbf69e359232f3b14ab0be4453306926346f552 Mon Sep 17 00:00:00 2001
From: q <q@think>
Date: Sat, 18 Apr 2026 19:38:55 +0300
Subject: [PATCH] =?UTF-8?q?Increase=20CHUNK=5FSIZE=20256=E2=86=92512,=20OV?=
 =?UTF-8?q?ERLAP=20128=E2=86=92192=20for=20better=20retrieval=20context?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Larger chunks give the reranker more context per candidate and reduce
chunk count (~2x fewer), so each message_id is better represented.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 index/main.py | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/index/main.py b/index/main.py
index d6083ea..f310c34 100644
--- a/index/main.py
+++ b/index/main.py
@@ -73,8 +73,8 @@ class SparseVector(BaseModel):
     values: list[float]
 
 
-CHUNK_SIZE = 256
-OVERLAP_SIZE = 128
+CHUNK_SIZE = 512
+OVERLAP_SIZE = 192
 SPARSE_MODEL_NAME = "Qdrant/bm25"
 FASTEMBED_CACHE_PATH = "/models/fastembed"