Merge pull request #3 from morethanair/fix/korean-embedding-model

app.py

@@ -84,7 +84,7 @@ def generate_youtube_embed_html(youtube_url: str, timestamp: str) -> str | None:
 PINECONE_API_KEY = os.getenv("PINECONE_API_KEY","pcsk_PZHLK_TRAvMCyNmJM4FKGCX7rbbY22a58fhnWYasx1mf3WL6sRasoASZXfsbnJYvCQ13w") # Load from environment variable
 PINECONE_ENV = os.getenv("PINECONE_ENV", "us-east-1")
 INDEX_NAME = "video-embeddings"
-EMBEDDING_MODEL = SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2')
+EMBEDDING_MODEL_ID = 'jhgan/ko-sroberta-multitask' # 모델 ID로 변경
 
 # OpenAI 설정
 OPENAI_API_KEY = "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"
@@ -109,8 +109,8 @@ def init_pinecone():
 def load_embedding_model():
     """Sentence Transformer 모델을 로드합니다."""
     try:
-        model = SentenceTransformer(EMBEDDING_MODEL)
-        logger.info(f"Successfully loaded embedding model: {EMBEDDING_MODEL}")
+        model = SentenceTransformer(EMBEDDING_MODEL_ID) # 수정된 모델 ID 사용
+        logger.info(f"Successfully loaded embedding model: {EMBEDDING_MODEL_ID}") # 수정된 모델 ID 사용
         return model
     except Exception as e:
         st.error(f"임베딩 모델 로딩 중 오류 발생: {e}")

requirements.txt

@@ -5,3 +5,4 @@ pandas
 openai 
 sentence-transformers==2.2.2
 torch==2.2.0
+huggingface-hub==0.0.8