fix vectorstore qdrant

deploy.sh

@@ -2,15 +2,15 @@
 
 if [ "$1" = "--no-cache" ]; then
     echo "📦️ Building without cache"
-    ssh idsg1 'cd /mnt/um-share-drive/vemonet/libre-chat ; git pull ; docker-compose build --no-cache ; docker-compose down ; docker-compose up --force-recreate -d'
+    ssh idsg1 'cd /mnt/um-share-drive/vemonet/libre-chat ; git pull ; docker compose build --no-cache ; docker compose down ; docker compose up --force-recreate -d'
 else
     echo "♻️ Building with cache"
-    ssh ids2 'cd /mnt/um-share-drive/vemonet/libre-chat ; git pull ; docker-compose up --force-recreate --build -d'
+    ssh idsg1 'cd /mnt/um-share-drive/vemonet/libre-chat ; git pull ; docker compose up --force-recreate --build -d'
 fi
 
 
 # Build with cache:
-# ssh ids2 'cd /data/deploy-services/knowledge-collaboratory ; git pull ; docker-compose -f docker-compose.yml -f docker-compose.prod.yml up --force-recreate --build -d'
+# ssh ids2 'cd /data/deploy-services/knowledge-collaboratory ; git pull ; docker compose -f docker compose.yml -f docker compose.prod.yml up --force-recreate --build -d'
 
 # Build without cache:
-# ssh ids2 'cd /data/deploy-services/knowledge-collaboratory ; git pull ; docker-compose -f docker-compose.yml -f docker-compose.prod.yml build ; docker-compose -f docker-compose.yml -f docker-compose.prod.yml down ; docker-compose -f docker-compose.yml -f docker-compose.prod.yml up --force-recreate -d'
+# ssh ids2 'cd /data/deploy-services/knowledge-collaboratory ; git pull ; docker compose -f docker compose.yml -f docker compose.prod.yml build ; docker compose -f docker compose.yml -f docker compose.prod.yml down ; docker compose -f docker compose.yml -f docker compose.prod.yml up --force-recreate -d'

docs/docs/use-docker.md

@@ -1,10 +1,6 @@
 [![Image size](https://ghcr-badge.egpl.dev/vemonet/libre-chat/size)](https://github.com/vemonet/libre-chat/pkgs/container/libre-chat)
 
-Libre Chat is available as a [docker image](https://github.com/vemonet/libre-chat/pkgs/container/libre-chat), it is recommended to use docker for deploying in production as it uses gunicorn to run multiple workers.
-
-!!! Warning "Shared memory for multiple users"
-
-    Memory of the chatbot is shared betweem the users that are on the same worker.
+Libre Chat is available as a [docker image](https://github.com/vemonet/libre-chat/pkgs/container/libre-chat) that will use CUDA when available. It is recommended to use docker for deploying in production as it uses gunicorn to run multiple workers.
 
 ## ⚡ Quickstart
 
@@ -14,6 +10,11 @@ If you just want deploy it using the pre-trained Mixtral model, you can use dock
 docker run -it -p 8000:8000 ghcr.io/vemonet/libre-chat:main
 ```
 
+!!! Warning "Loading the model takes time"
+
+    Downloading the model will take time the first time, you can also pre-download it manually.
+    If you are using GPU, loading the model when the application starts also takes some time (can take a few minutes)
+
 ## ⚙️ Configure with docker compose
 
 1. Create a `chat.yml` file with your chat web service configuration.

src/libre_chat/llm.py

@@ -226,6 +226,7 @@ def build_vectorstore(self, documents_path: Optional[str] = None) -> Optional[Qd
             embeddings = HuggingFaceEmbeddings(
                 model_name=self.conf.vector.embeddings_path, model_kwargs={"device": self.device}
             )
+            os.makedirs(self.conf.vector.vector_path, exist_ok=True)
             vectorstore = Qdrant.from_documents(
                 splitted_texts,
                 embeddings,