feat: deploy to hf spaces

backend/space/litellm_config.yaml

@@ -0,0 +1,43 @@
+litellm_settings:
+  drop_params: true
+model_list:
+  - model_name: 'HuggingFace: Mistral: Mistral 7B Instruct v0.1'
+    litellm_params:
+      model: huggingface/mistralai/Mistral-7B-Instruct-v0.1
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 1024
+  - model_name: 'HuggingFace: Mistral: Mistral 7B Instruct v0.2'
+    litellm_params:
+      model: huggingface/mistralai/Mistral-7B-Instruct-v0.2
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 1024
+  - model_name: 'HuggingFace: Meta: Llama 3 8B Instruct'
+    litellm_params:
+      model: huggingface/meta-llama/Meta-Llama-3-8B-Instruct
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 2047
+  - model_name: 'HuggingFace: Mistral: Mixtral 8x7B Instruct v0.1'
+    litellm_params:
+      model: huggingface/mistralai/Mixtral-8x7B-Instruct-v0.1
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 8192
+  - model_name: 'HuggingFace: Microsoft: Phi-3 Mini-4K-Instruct'
+    litellm_params:
+      model: huggingface/microsoft/Phi-3-mini-4k-instruct
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 1024
+  - model_name: 'HuggingFace: Google: Gemma 7B 1.1'
+    litellm_params:
+      model: huggingface/google/gemma-1.1-7b-it
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 1024
+  - model_name: 'HuggingFace: Yi-1.5 34B Chat'
+    litellm_params:
+      model: huggingface/01-ai/Yi-1.5-34B-Chat
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 1024
+  - model_name: 'HuggingFace: Nous Research: Nous Hermes 2 Mixtral 8x7B DPO'
+    litellm_params:
+      model: huggingface/NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO
+      api_key: os.environ/HF_TOKEN
+      max_tokens: 2048

backend/start.sh

@@ -30,4 +30,32 @@ if [ "$USE_CUDA_DOCKER" = "true" ]; then
   export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/usr/local/lib/python3.11/site-packages/torch/lib:/usr/local/lib/python3.11/site-packages/nvidia/cudnn/lib"
 fi
 
+
+# HFSPACE:START
+# Check if SPACE_ID is set, if so, configure for space
+if [ -n "$SPACE_ID" ]; then
+  echo "Configuring for HuggingFace Space deployment"
+  
+  # Copy litellm_config.yaml with specified ownership
+  echo "Copying litellm_config.yaml to the desired location with specified ownership..."
+  cp ./backend/space/litellm_config.yaml ./data/litellm/config.yaml
+
+  WEBUI_SECRET_KEY="$WEBUI_SECRET_KEY" uvicorn main:app --host "$HOST" --port "$PORT" --forwarded-allow-ips '*' &
+  webui_pid=$!
+  echo "Waiting for webui to start..."
+  while ! curl -s http://localhost:8080/health > /dev/null; do
+    sleep 1
+  done
+  echo "Creating admin user..."
+  curl \
+    -X POST "http://localhost:8080/api/v1/auths/signup" \
+    -H "accept: application/json" \
+    -H "Content-Type: application/json" \
+    -d "{ \"email\": \"${ADMIN_USER_EMAIL}\", \"password\": \"${ADMIN_USER_PASSWORD}\", \"name\": \"Admin\" }"
+  echo "Shutting down webui..."
+  kill $webui_pid
+  export WEBUI_URL=${SPACE_HOST}
+fi
+# HFSPACE:END
+
 WEBUI_SECRET_KEY="$WEBUI_SECRET_KEY" exec uvicorn main:app --host "$HOST" --port "$PORT" --forwarded-allow-ips '*'