feat(ai): bge over litellm

ai/compose.yaml

@@ -105,6 +105,7 @@ services:
       LITELLM_MASTER_KEY: ${AI_LITELLM_API_KEY}
       DATABASE_URL: postgresql://${AI_DB_USER}:${AI_DB_PASSWORD}@ai_postgres:5432/litellm
       GPU_VLLM_LLAMA_URL: ${GPU_VLLM_LLAMA_URL}
+      GPU_VLLM_BGE_URL: ${GPU_VLLM_BGE_URL}
       # LITELLM_DROP_PARAMS: 'true'  # DISABLED: Was breaking streaming
       NO_DOCS: "true"
       NO_REDOC: "true"

ai/litellm-config.yaml

@@ -42,6 +42,15 @@ model_list:
       supports_system_messages: true # Llama supports system messages
       stream: true # Enable streaming by default
 
+  # Embeddings - BGE Large (Port 8002)
+  - model_name: bge-large-en
+    litellm_params:
+      model: openai/BAAI/bge-large-en-v1.5
+      api_base: os.environ/GPU_VLLM_BGE_URL
+      api_key: "EMPTY"
+      rpm: 1000
+      tpm: 500000
+
 litellm_settings:
   drop_params: false # DISABLED: Was breaking streaming
   set_verbose: true # Enable verbose logging for debugging streaming issues