Fix Ollama connection issues with enhanced error handling and updated ngrok URL

app.py

@@ -1,4 +1,4 @@
-# Force redeploy trigger - version 1.9
+# Force redeploy trigger - version 2.0
 import streamlit as st
 from utils.config import config
 import requests
@@ -87,13 +87,13 @@ st.sidebar.markdown("---")
 BASE_URL = os.environ.get("SPACE_ID", "")
 IS_HF_SPACE = bool(BASE_URL)
 
-# Fetch Ollama status
+# Fetch Ollama status with enhanced error handling
 def get_ollama_status(ngrok_url):
     try:
         response = requests.get(
             f"{ngrok_url}/api/tags",
             headers=NGROK_HEADERS,
-            timeout=10
+            timeout=15  # Increased timeout
         )
         if response.status_code == 200:
             models = response.json().get("models", [])
@@ -112,11 +112,30 @@ def get_ollama_status(ngrok_url):
             else:
                 st.session_state.model_status = "no_models"
                 return {
-                    "running": False,
+                    "running": True,  # Server is running but no models
                     "model_loaded": None,
                     "remote_host": ngrok_url,
                     "message": "Connected to Ollama but no models found"
                 }
+        elif response.status_code == 404:
+            # Server might be running but endpoint not available
+            response2 = requests.get(f"{ngrok_url}", headers=NGROK_HEADERS, timeout=10)
+            if response2.status_code == 200:
+                st.session_state.model_status = "checking"
+                return {
+                    "running": True,
+                    "model_loaded": "unknown",
+                    "remote_host": ngrok_url,
+                    "message": "Server running, endpoint check inconclusive"
+                }
+            else:
+                st.session_state.model_status = "unreachable"
+                return {
+                    "running": False,
+                    "model_loaded": None,
+                    "error": f"HTTP {response.status_code}",
+                    "remote_host": ngrok_url
+                }
         else:
             st.session_state.model_status = "unreachable"
             return {
@@ -125,6 +144,14 @@ def get_ollama_status(ngrok_url):
                 "error": f"HTTP {response.status_code}",
                 "remote_host": ngrok_url
             }
+    except requests.exceptions.Timeout:
+        st.session_state.model_status = "unreachable"
+        return {
+            "running": False,
+            "model_loaded": None,
+            "error": "Timeout - server not responding",
+            "remote_host": ngrok_url
+        }
     except Exception as e:
         st.session_state.model_status = "unreachable"
         return {
@@ -156,10 +183,12 @@ if ollama_status is None:
         "remote_host": st.session_state.ngrok_url
     }
 
-# Update model status
-if ollama_status.get("running", False):
-    if ollama_status.get("available_models"):
+# Update model status with better logic
+if ollama_status and ollama_status.get("running", False):
+    if ollama_status.get("available_models") and len(ollama_status.get("available_models", [])) > 0:
         st.session_state.model_status = "ready"
+    elif ollama_status.get("model_loaded") == "unknown":
+        st.session_state.model_status = "ready"  # Assume ready if server responds
     else:
         st.session_state.model_status = "no_models"
 else:

test_ollama_connection.py

@@ -5,8 +5,9 @@ from dotenv import load_dotenv
 # Load environment variables
 load_dotenv()
 
-OLLAMA_HOST = os.getenv("OLLAMA_HOST", "https://ace32bd59aef.ngrok-free.app")
-MODEL_NAME = os.getenv("LOCAL_MODEL_NAME", "mistral:latest")
+# Use the currently active ngrok URL from your logs
+OLLAMA_HOST = os.getenv("OLLAMA_HOST", "https://f943b91f0a0c.ngrok-free.app")
+MODEL_NAME = os.getenv("LOCAL_MODEL_NAME", "mistral")
 
 print(f"Testing Ollama connection to: {OLLAMA_HOST}")
 print(f"Using model: {MODEL_NAME}")
@@ -30,6 +31,14 @@ try:
         print(f"Found {len(models)} models:")
         for model in models:
             print(f"  - {model['name']} ({model.get('size', 'Unknown size')})")
+    elif response.status_code == 404:
+        print("⚠️  Endpoint not found - checking root endpoint...")
+        # Try basic connectivity
+        response2 = requests.get(f"{OLLAMA_HOST}", headers=headers, timeout=10)
+        if response2.status_code == 200:
+            print("✓ Server is running but /api/tags endpoint not available")
+        else:
+            print(f"✗ Server returned: {response2.status_code}")
     else:
         print(f"Error: {response.text}")
 except Exception as e: