From 57bf30eca15e643909ef02496bb7c4a3003b641d Mon Sep 17 00:00:00 2001
From: Codex <codex@local>
Date: Wed, 15 Jul 2026 19:32:21 +0200
Subject: [PATCH] fix: normalize CosyVoice3 instructions

---
 deploy/selfmedia-voice/api.py | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/deploy/selfmedia-voice/api.py b/deploy/selfmedia-voice/api.py
index ae55a9e2..b889921a 100644
--- a/deploy/selfmedia-voice/api.py
+++ b/deploy/selfmedia-voice/api.py
@@ -89,6 +89,17 @@ async def lifespan(_: FastAPI):
 
 app = FastAPI(title="SelfMedia Voice", version="1", lifespan=lifespan)
 
+INSTRUCTION_END_TOKEN = "<|endofprompt|>"
+
+
+def normalize_instruction(value: str | None) -> str | None:
+    if value is None or not value.strip():
+        return None
+    normalized = value.strip()
+    if INSTRUCTION_END_TOKEN in normalized:
+        return normalized
+    return f"You are a helpful assistant. {normalized}{INSTRUCTION_END_TOKEN}"
+
 
 @app.get("/health")
 def health() -> dict:
@@ -125,7 +136,7 @@ def speech(request: SpeechRequest) -> Response:
         raise HTTPException(status_code=400, detail="voice must select the configured reference voice")
 
     started = time.perf_counter()
-    instruction = request.instructions or request.instruct
+    instruction = normalize_instruction(request.instructions or request.instruct)
     with runtime.lock:
         inference = (
             runtime.model.inference_instruct2(