Backend

I have removed the hardcoded part for the all-mpnet-base-v2 and substituted it with a separate EMBEDDINGS_PATH variable.
Now the user is able to pre-download the embeddings (similarly to the recomended way with all-mpnet-base-v2) and also leave it to sentence-transformers without needing to touch vectorstore/base.py

application/core/settings.py

@@ -16,6 +16,7 @@ class Settings(BaseSettings):
         None  # if LLM_PROVIDER is openai, LLM_NAME can be gpt-4 or gpt-3.5-turbo
     )
     EMBEDDINGS_NAME: str = "huggingface_sentence-transformers/all-mpnet-base-v2"
+    EMBEDDINGS_PATH: Optional[str] = "./models/all-mpnet-base-v2" # Set None for SentenceTransformer to manage download
     CELERY_BROKER_URL: str = "redis://localhost:6379/0"
     CELERY_RESULT_BACKEND: str = "redis://localhost:6379/1"
     MONGO_URI: str = "mongodb://localhost:27017/docsgpt"

application/vectorstore/base.py

@@ -74,17 +74,11 @@ class BaseVectorStore(ABC):
                     embeddings_name,
                     openai_api_key=embeddings_key
                 )
-        elif embeddings_name == "huggingface_sentence-transformers/all-mpnet-base-v2":
-            if os.path.exists("./models/all-mpnet-base-v2"):
-                embedding_instance = EmbeddingsSingleton.get_instance(
-                    embeddings_name = "./models/all-mpnet-base-v2",
-                )
-            else:
-                embedding_instance = EmbeddingsSingleton.get_instance(
-                    embeddings_name,
-                )
         else:
-            embedding_instance = EmbeddingsSingleton.get_instance(embeddings_name)
+            model_identifier = embeddings_name
+            if settings.EMBEDDINGS_PATH and os.path.exists(settings.EMBEDDINGS_PATH):
+                model_identifier = settings.EMBEDDINGS_PATH
+            embedding_instance = EmbeddingsSingleton.get_instance(model_identifier)
 
         return embedding_instance