diff --git a/backend/api.py b/backend/api.py
index 5cad18e..4ad106a 100644
--- a/backend/api.py
+++ b/backend/api.py
@@ -213,7 +213,6 @@ def get_audio(UUID: UUID4):
         raise HTTPException(status_code=404, detail="Audio not found")
     if isinstance(PENDING_AUDIO[UUID], str):
         audio_text = PENDING_AUDIO[UUID]
-        del PENDING_AUDIO[UUID]
         response = openai_client.audio.speech.create(
             model="tts-1",
             voice="alloy",
diff --git a/frontend/API.md b/frontend/API.md
deleted file mode 100644
index c7ab717..0000000
--- a/frontend/API.md
+++ /dev/null
@@ -1,33 +0,0 @@
-
-> For the demo, language ∈ {EN_JA, EN_ZH-CN}
-
-POST /ai-mark {
-	question: str, 
-	user_answer: str, 
-	expected: str, 
-	chapter: str,
-	language: str
-}
--> {correct: bool, reason: str}
-
-POST /human-chat {
-	target_name: str,
-	target_hobbies: list\[str],
-	history: list\[{from\_me: bool, msg: str}],
-	language: str
-}
--> {msg: str}
-
-> Chat histories will be stored in localStorage
-
-POST /recognize (Body is an audio file)
--> {text: str}
-
-POST /character-chat {
-	character: str
-	history: list\[{from_me: bool, msg: str}]
-	language: str
-}
--> {msg: str, audio: UUID}
-
-GET /audio/{UUID}
\ No newline at end of file