Update entrypoint.sh

entrypoint.sh

@@ -1,22 +1,37 @@
 #!/bin/bash
 
-# Make sure `/data/db` directory exists even with persistent storage
+# Ensure `/data/db` directory exists even with persistent storage
 mkdir -p /data/db
-# If app crashed, mongo didn't stop gracefully. Remove all the old *.lock files
-find /data/db -name "*.lock" -type f -exec rm -f {} \; 
+# If app crashed, remove old *.lock files to allow MongoDB to restart
+find /data/db -name "*.lock" -type f -exec rm -f {} \;
+
 # Start the local Mongo database
 mongod &
 
-# Start the text-generation-inference process
-text-generation-launcher --model-id ${MODEL_NAME} --num-shard 1 --port 8080 --trust-remote-code &
+# Wait for MongoDB to be ready
+until nc -z localhost 27017; do
+    echo "Waiting for MongoDB to start..."
+    sleep 2
+done
+
+# Set the Hugging Face API URL and Token
+HF_API_URL="https://api-inference.huggingface.co/models/${MODEL_NAME}"
+HF_TOKEN=${HF_TOKEN}
 
-# Wait for text-generation-inference to start
-curl --retry 60 --retry-delay 10 --retry-connrefused http://127.0.0.1:8080/health
+# Ensure API token is set
+if [ -z "$HF_TOKEN" ]; then
+    echo "Error: HF_TOKEN is not set. Exiting."
+    exit 1
+fi
+
+# Configure Chat-UI to use the Hugging Face Inference API
+export HF_API_URL HF_TOKEN
 
 # Start the chat-ui process
 dotenv -e /app/.env -c -- node /app/build/index.js -- --host 0.0.0.0 --port 3000
+
 # Wait for any process to exit
 wait -n
 
-# Exit with status of process that exited first
-exit $?
+# Exit with the status of the first process that exits
+exit $?