fix(start.sh): robust wait for Ollama, retry model creation, don't hard-exit if Modelfile missing

start.sh

@@ -1,39 +1,73 @@
 #!/usr/bin/env bash
-set -e
+set -euo pipefail
 
-# start.sh - ensures model exists, starts ollama, then runs gunicorn for API
+echo "===== Application Startup at $(date) ====="
+
+# start.sh - robust startup flow: start Ollama, wait for readiness, create model (if provided), then run gunicorn
 APP_DIR=/app
 MODEL_NAME=aj-mini
 
-cd $APP_DIR
+cd "$APP_DIR"
 
-# If ollama is missing, print message and continue to attempt fallback
+# Check if ollama is installed
 if ! command -v ollama >/dev/null 2>&1; then
-  echo "Warning: ollama binary not found in container. Make sure Ollama was installed."
+  echo "Error: ollama binary not found in container. Make sure Ollama was installed."
+  exit 1
 fi
 
-# Create model if Modelfile exists and model not present
-if command -v ollama >/dev/null 2>&1; then
-  if ! ollama list | grep -q "$MODEL_NAME"; then
-    if [ -f "$APP_DIR/Modelfile-aj-mini-v2" ]; then
-      echo "Creating model $MODEL_NAME from Modelfile..."
-      ollama create $MODEL_NAME -f Modelfile-aj-mini-v2 || true
-    else
-      echo "Modelfile not found. Skipping model creation."
-    fi
-  else
-    echo "Model $MODEL_NAME already exists."
+# Start ollama serve in background
+echo "Starting ollama serve..."
+ollama serve > /tmp/ollama.log 2>&1 &
+OLLAMA_PID=$!
+echo "Ollama PID: $OLLAMA_PID"
+
+# Tail Ollama logs in background for visibility
+tail -n +1 -f /tmp/ollama.log &
+
+# Wait for ollama to be ready (with retries)
+echo "Waiting for Ollama to be ready (http://localhost:11434)..."
+for i in {1..60}; do
+  if curl -s http://127.0.0.1:11434/api/tags > /dev/null 2>&1; then
+    echo "Ollama is ready (after $i attempts)!"
+    break
   fi
+  echo "Waiting for Ollama to start... ($i/60)"
+  sleep 2
+  if [ $i -eq 60 ]; then
+    echo "Warning: Ollama did not become ready after retries. Check /tmp/ollama.log for details but continuing startup." >&2
+  fi
+done
+
+# Function to create model with retries
+create_model() {
+  local tries=0
+  local max=5
+  until [ $tries -ge $max ]; do
+    if ollama list | grep -q "^$MODEL_NAME\b"; then
+      echo "Model $MODEL_NAME already exists."
+      return 0
+    fi
+    echo "Attempting to create model $MODEL_NAME (try $((tries+1))/$max)..."
+    if ollama create "$MODEL_NAME" -f Modelfile-aj-mini-v2 >> /tmp/ollama.log 2>&1; then
+      echo "Model $MODEL_NAME created successfully."
+      return 0
+    fi
+    tries=$((tries+1))
+    sleep 3
+  done
+  echo "Failed to create model $MODEL_NAME after $max attempts. Continuing; API may return errors until model is available." >&2
+  return 1
+}
 
-  # Start ollama serve in background
-  echo "Starting ollama serve..."
-  ollama serve &
-  OLLAMA_PID=$!
-  echo "Ollama PID: $OLLAMA_PID"
+# Create model if Modelfile exists
+if [ -f "$APP_DIR/Modelfile-aj-mini-v2" ]; then
+  echo "Found Modelfile at $APP_DIR/Modelfile-aj-mini-v2"
+  create_model || true
 else
-  echo "ollama not available, continuing. Model calls will fail unless ollama is installed."
+  echo "Modelfile not found at $APP_DIR/Modelfile-aj-mini-v2 - skipping model creation. If you expect a local model, add the Modelfile and push again." >&2
 fi
 
 # Start Gunicorn for Flask app
 echo "Starting gunicorn for API..."
-exec gunicorn api_server:app --bind 0.0.0.0:5000 --workers 2
+echo "===== API Server Ready ====="
+exec gunicorn api_server:app --bind 0.0.0.0:5000 --workers 2 --timeout 120