refac

Dockerfile

@@ -13,6 +13,7 @@ ARG USE_CUDA_VER=cu128
 # IMPORTANT: If you change the embedding model (sentence-transformers/all-MiniLM-L6-v2) and vice versa, you aren't able to use RAG Chat with your previous documents loaded in the WebUI! You need to re-embed them.
 ARG USE_EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2
 ARG USE_RERANKING_MODEL=""
+ARG USE_AUXILIARY_EMBEDDING_MODEL=TaylorAI/bge-micro-v2
 
 # Tiktoken encoding name; models to use can be found at https://huggingface.co/models?library=tiktoken
 ARG USE_TIKTOKEN_ENCODING_NAME="cl100k_base"
@@ -52,6 +53,7 @@ ARG USE_SLIM
 ARG USE_PERMISSION_HARDENING
 ARG USE_EMBEDDING_MODEL
 ARG USE_RERANKING_MODEL
+ARG USE_AUXILIARY_EMBEDDING_MODEL
 ARG UID
 ARG GID
 
@@ -67,7 +69,8 @@ ENV ENV=prod \
     USE_SLIM_DOCKER=${USE_SLIM} \
     USE_CUDA_DOCKER_VER=${USE_CUDA_VER} \
     USE_EMBEDDING_MODEL_DOCKER=${USE_EMBEDDING_MODEL} \
-    USE_RERANKING_MODEL_DOCKER=${USE_RERANKING_MODEL}
+    USE_RERANKING_MODEL_DOCKER=${USE_RERANKING_MODEL} \
+    USE_AUXILIARY_EMBEDDING_MODEL_DOCKER=${USE_AUXILIARY_EMBEDDING_MODEL}
 
 ## Basis URL Config ##
 ENV OLLAMA_BASE_URL="/ollama" \
@@ -88,6 +91,7 @@ ENV WHISPER_MODEL="base" \
 ## RAG Embedding model settings ##
 ENV RAG_EMBEDDING_MODEL="$USE_EMBEDDING_MODEL_DOCKER" \
     RAG_RERANKING_MODEL="$USE_RERANKING_MODEL_DOCKER" \
+    AUXILIARY_EMBEDDING_MODEL="$USE_AUXILIARY_EMBEDDING_MODEL_DOCKER" \
     SENTENCE_TRANSFORMERS_HOME="/app/backend/data/cache/embedding/models"
 
 ## Tiktoken model settings ##
@@ -136,6 +140,7 @@ RUN pip3 install --no-cache-dir uv && \
     pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/$USE_CUDA_DOCKER_VER --no-cache-dir && \
     uv pip install --system -r requirements.txt --no-cache-dir && \
     python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \
+    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ.get('AUXILIARY_EMBEDDING_MODEL', 'TaylorAI/bge-micro-v2'), device='cpu')" && \
     python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \
     python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \
     else \
@@ -143,6 +148,7 @@ RUN pip3 install --no-cache-dir uv && \
     uv pip install --system -r requirements.txt --no-cache-dir && \
     if [ "$USE_SLIM" != "true" ]; then \
     python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ['RAG_EMBEDDING_MODEL'], device='cpu')" && \
+    python -c "import os; from sentence_transformers import SentenceTransformer; SentenceTransformer(os.environ.get('AUXILIARY_EMBEDDING_MODEL', 'TaylorAI/bge-micro-v2'), device='cpu')" && \
     python -c "import os; from faster_whisper import WhisperModel; WhisperModel(os.environ['WHISPER_MODEL'], device='cpu', compute_type='int8', download_root=os.environ['WHISPER_MODEL_DIR'])"; \
     python -c "import os; import tiktoken; tiktoken.get_encoding(os.environ['TIKTOKEN_ENCODING_NAME'])"; \
     fi; \

backend/open_webui/routers/evaluations.py

@@ -49,7 +49,9 @@ router = APIRouter()
 #    4. Feedbacks about unrelated topics (e.g., "cooking") contribute less
 #    This gives topic-specific leaderboards without needing separate data.
 
-EMBEDDING_MODEL_NAME = "TaylorAI/bge-micro-v2"
+import os
+
+EMBEDDING_MODEL_NAME = os.environ.get("AUXILIARY_EMBEDDING_MODEL", "TaylorAI/bge-micro-v2")
 _embedding_model = None