Fix Ollama connection issues and improve error handling for Hugging Face Spaces

app.py

@@ -1,8 +1,9 @@
-# Force redeploy trigger - version 1.1
+# Force redeploy trigger - version 1.2
 import streamlit as st
 from utils.config import config
 import requests
 import json
+import os
 from core.memory import load_user_state
 
 # Set page config
@@ -13,19 +14,50 @@ st.sidebar.title("🧘 AI Life Coach")
 user = st.sidebar.selectbox("Select User", ["Rob", "Sarah"])
 st.sidebar.markdown("---")
 
+# Get the base URL for API calls (works in Hugging Face Spaces)
+# In HF Spaces, we need to use the same port for both frontend and backend
+# or properly configure the backend service
+BASE_URL = os.environ.get("SPACE_ID", "")  # Will be set in HF Spaces
+IS_HF_SPACE = bool(BASE_URL)
+
+# Headers to skip ngrok browser warning
+NGROK_HEADERS = {
+    "ngrok-skip-browser-warning": "true",
+    "User-Agent": "AI-Life-Coach-App"
+}
+
 # Fetch Ollama status
 def get_ollama_status():
     try:
-        # Add headers to skip ngrok browser warning
-        headers = {
-            "ngrok-skip-browser-warning": "true",
-            "User-Agent": "AI-Life-Coach-App"
-        }
-        response = requests.get("http://localhost:8000/api/ollama-status", headers=headers)
+        # Try to connect to the remote Ollama service directly
+        response = requests.get(
+            f"{config.ollama_host}/api/tags", 
+            headers=NGROK_HEADERS,
+            timeout=10
+        )
         if response.status_code == 200:
-            return response.json()
-    except Exception:
-        return {"running": False, "model_loaded": None}
+            models = response.json().get("models", [])
+            if models:
+                return {
+                    "running": True, 
+                    "model_loaded": models[0].get("name"),
+                    "remote_host": config.ollama_host
+                }
+    except Exception as e:
+        # If direct connection fails, show error
+        return {
+            "running": False, 
+            "model_loaded": None,
+            "error": str(e),
+            "remote_host": config.ollama_host
+        }
+    
+    # If we get here, connection worked but no models
+    return {
+        "running": False, 
+        "model_loaded": None,
+        "remote_host": config.ollama_host
+    }
 
 # After user selects name, load conversation history
 def get_conversation_history(user_id):
@@ -34,20 +66,40 @@ def get_conversation_history(user_id):
         return json.loads(user_state["conversation"])
     return []
 
+# Check Ollama status
 ollama_status = get_ollama_status()
 
 # Display Ollama status
 if ollama_status["running"]:
     st.sidebar.success(f"🧠 Model Running: {ollama_status['model_loaded']}")
+    st.sidebar.info(f"Connected to: {ollama_status['remote_host']}")
 else:
-    st.sidebar.error("🧠 Ollama is not running or no model loaded.")
+    st.sidebar.error("🧠 Ollama is not accessible")
+    st.sidebar.info(f"Configured host: {ollama_status['remote_host']}")
+    if "error" in ollama_status:
+        st.sidebar.caption(f"Error: {ollama_status['error']}")
 
 # Main chat interface
 st.title("🧘 AI Life Coach")
 st.markdown("Talk to your personal development assistant.")
 
+# Show detailed status
+with st.expander("🔍 Connection Status"):
+    st.write("Ollama Status:", ollama_status)
+    st.write("Environment Info:")
+    st.write("- Is HF Space:", IS_HF_SPACE)
+    st.write("- Base URL:", BASE_URL or "Not in HF Space")
+    st.write("- Configured Ollama Host:", config.ollama_host)
+
 if not ollama_status["running"]:
-    st.warning("⚠️ Ollama is not running. Please start Ollama to use the AI Life Coach.")
+    st.warning("⚠️ Ollama is not accessible. Please check your Ollama/ngrok setup.")
+    st.info("""
+    Troubleshooting tips:
+    1. Ensure your Ollama service is running locally
+    2. Verify your ngrok tunnel is active and pointing to Ollama (port 11434)
+    3. Check that the ngrok URL in your .env file matches your active tunnel
+    4. Confirm that your ngrok account allows connections from Hugging Face Spaces
+    """)
 else:
     # Display conversation history
     conversation = get_conversation_history(user)
@@ -65,24 +117,35 @@ else:
             # Display user message
             st.markdown(f"**You:** {user_input}")
 
-            # Send to backend
+            # Send to Ollama directly (bypassing backend for simplicity)
             with st.spinner("AI Coach is thinking..."):
                 try:
-                    # Add headers to skip ngrok browser warning
-                    headers = {
-                        "ngrok-skip-browser-warning": "true",
-                        "User-Agent": "AI-Life-Coach-App"
+                    # Prepare the prompt with conversation history
+                    conversation_history = [{"role": msg["role"], "content": msg["content"]} 
+                                          for msg in conversation[-5:]]  # Last 5 messages
+                    conversation_history.append({"role": "user", "content": user_input})
+                    
+                    payload = {
+                        "model": config.local_model_name,
+                        "messages": conversation_history,
+                        "stream": False
                     }
+                    
                     response = requests.post(
-                        "http://localhost:8000/api/chat",
-                        json={"user_id": user, "message": user_input},
-                        headers=headers
+                        f"{config.ollama_host}/api/chat",
+                        json=payload,
+                        headers=NGROK_HEADERS,
+                        timeout=60
                     )
+                    
                     if response.status_code == 200:
                         response_data = response.json()
-                        ai_response = response_data.get("response", "")
+                        ai_response = response_data.get("message", {}).get("content", "")
                         st.markdown(f"**AI Coach:** {ai_response}")
+                        
+                        # Note: In a production app, we'd save the conversation to Redis here
                     else:
-                        st.error("Failed to get response from AI Coach.")
+                        st.error(f"Failed to get response from Ollama: {response.status_code}")
+                        st.error(response.text[:200])
                 except Exception as e:
                     st.error(f"Connection error: {e}")

services/ollama_monitor.py

@@ -14,7 +14,7 @@ def check_ollama_status():
             "local_url": "http://localhost:11434/"
         }
     """
-    ngrok_url = "https://ace32bd59aef.ngrok-free.app/"
+    ngrok_url = config.ollama_host  # Use configured host
     local_url = "http://localhost:11434/"  # Always check localhost as fallback
 
     def _get_model_from_url(base_url, retries=3, delay=1):
@@ -26,7 +26,7 @@ def check_ollama_status():
                     "ngrok-skip-browser-warning": "true",
                     "User-Agent": "AI-Life-Coach-App"
                 }
-                response = requests.get(f"{base_url}/api/tags", timeout=5, headers=headers)
+                response = requests.get(f"{base_url}/api/tags", timeout=10, headers=headers)
                 if response.status_code == 200:
                     models = response.json().get("models", [])
                     if models:

start.sh

@@ -0,0 +1,18 @@
+#!/bin/bash
+
+echo "Starting AI Life Coach..."
+
+# Start FastAPI backend in background
+echo "Starting FastAPI backend..."
+uvicorn api.main:app --host 0.0.0.0 --port 8000 &
+BACKEND_PID=0
+
+# Give backend a moment to start
+sleep 3
+
+# Start Streamlit frontend
+echo "Starting Streamlit frontend..."
+streamlit run app.py --server.port 8501 --server.address 0.0.0.0
+
+# Kill backend when Streamlit exits
+kill