From 4e446c1d7c764adcc009ca8091e302c31328ad7e Mon Sep 17 00:00:00 2001
From: javis-bot <claude@tkrmagid.kr>
Date: Sat, 13 Jun 2026 22:03:18 +0900
Subject: [PATCH] perf(stt): default Whisper model small -> medium for better
 Korean accuracy
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Whisper 'small' frequently mis-transcribes Korean (esp. the wake word 자비스:
'아비스', '자 비싸'). STT is only ~0.1s warm so there is ample latency headroom;
'medium' is far more accurate and fits VRAM alongside qwen2.5:3b on the 8GB GPU.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
---
 docker-compose.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/docker-compose.yml b/docker-compose.yml
index dbe60a0..f7b0916 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -64,7 +64,7 @@ services:
       OLLAMA_BASE_URL: http://ollama:11434
       OLLAMA_CHAT_MODEL: ${OLLAMA_CHAT_MODEL:-qwen2.5:3b}
       OLLAMA_EMBED_MODEL: ${OLLAMA_EMBED_MODEL:-nomic-embed-text}
-      WHISPER_MODEL: ${WHISPER_MODEL:-small}
+      WHISPER_MODEL: ${WHISPER_MODEL:-medium}
       WHISPER_DEVICE: ${WHISPER_DEVICE:-cuda}
       WHISPER_COMPUTE_TYPE: ${WHISPER_COMPUTE_TYPE:-float16}
       # Optional single-language lock for replies (empty = user's own language).