Update system to use remote Ollama host with graceful fallback handling

.env

@@ -1,20 +1,18 @@
-# Hugging Face API
-HF_TOKEN=your_hf_token
-HF_API_ENDPOINT_URL=https://zxzbfrlg3ssrk7d9.us-east-1.aws.endpoints.huggingface.cloud/v1/
+# Hugging Face Settings
+HF_TOKEN=your_huggingface_token_here
+HF_API_ENDPOINT_URL=https://api-inference.huggingface.co/v1/
 
-# External APIs
-TAVILY_API_KEY=your_tavily_key
-OPENWEATHER_API_KEY=your_weather_key
-NASA_API_KEY=your_nasa_key
+# API Keys
+TAVILY_API_KEY=your_tavily_api_key_here
+OPENWEATHER_API_KEY=your_openweather_api_key_here
+NASA_API_KEY=your_nasa_api_key_here
 
-# Redis
+# Redis Configuration
 REDIS_HOST=localhost
 REDIS_PORT=6379
-REDIS_USERNAME=your_redis_user
-REDIS_PASSWORD=your_redis_pass
+REDIS_USERNAME=
+REDIS_PASSWORD=
 
-# Local Model
+# Model Configuration
 LOCAL_MODEL_NAME=mistral-7b
-
-# Ollama
-OLLAMA_HOST=http://localhost:11434
+OLLAMA_HOST=https://a877ef1aa487.ngrok-free.app

README.md

@@ -21,3 +21,9 @@ A personal development assistant powered by LLMs.
 - User sessions (Rob & Sarah)
 - FastAPI backend
 - Streamlit UI
+
+## Deployment
+
+The application is designed to work in Hugging Face Spaces environment. For local LLM inference, it connects to a remote Ollama instance via ngrok tunnel at . This allows the application to access powerful local models without requiring them to be installed directly in the Space.
+
+In case the remote Ollama instance is unavailable, the system gracefully falls back to checking a local instance, and handles unavailability by showing appropriate status messages in the UI.

services/ollama_monitor.py

@@ -1,4 +1,5 @@
 import requests
+import time
 from utils.config import config
 
 def check_ollama_status():
@@ -14,23 +15,31 @@ def check_ollama_status():
         }
     """
     ngrok_url = "https://a877ef1aa487.ngrok-free.app/"
-    local_url = config.ollama_host  # From .env
+    local_url = "http://localhost:11434/"  # Always check localhost as fallback
 
-    def _get_model_from_url(base_url):
-        try:
-            response = requests.get(f"{base_url}/api/tags", timeout=3)
-            if response.status_code == 200:
-                models = response.json().get("models", [])
-                if models:
-                    return models[0].get("name")
-        except Exception:
-            return None
+    def _get_model_from_url(base_url, retries=3, delay=1):
+        """Try to get model info with retry logic"""
+        for attempt in range(retries):
+            try:
+                response = requests.get(f"{base_url}/api/tags", timeout=5)
+                if response.status_code == 200:
+                    models = response.json().get("models", [])
+                    if models:
+                        return models[0].get("name")
+            except Exception as e:
+                if attempt < retries - 1:  # Don't sleep on last attempt
+                    time.sleep(delay * (2 ** attempt))  # Exponential backoff
+                continue
         return None
 
-    local_model = _get_model_from_url(local_url)
+    # First try remote ngrok URL, then fall back to local
     remote_model = _get_model_from_url(ngrok_url)
-
-    model_loaded = local_model or remote_model
+    local_model = None
+    
+    if not remote_model:  # Only check local if remote failed
+        local_model = _get_model_from_url(local_url)
+    
+    model_loaded = remote_model or local_model
     running = bool(model_loaded)
 
     return {

utils/config.py

@@ -14,6 +14,7 @@ class Config:
         self.redis_username = os.getenv("REDIS_USERNAME")
         self.redis_password = os.getenv("REDIS_PASSWORD")
         self.local_model_name = os.getenv("LOCAL_MODEL_NAME", "mistral-7b")
-        self.ollama_host = os.getenv("OLLAMA_HOST", "http://localhost:11434")
+        # Use remote Ollama host as default, falling back to environment variable
+        self.ollama_host = os.getenv("OLLAMA_HOST", "https://a877ef1aa487.ngrok-free.app")
 
 config = Config()