Fix critical LLM provider bugs, Redis config, and Streamlit null safety issues

README.md

@@ -1,39 +0,0 @@
----
-title: AI Life Coach
-emoji: 🧘
-colorFrom: blue
-colorTo: green
-sdk: streamlit
-sdk_version: 1.24.0
-app_file: app.py
-pinned: false
-license: apache-2.0
----
-
-# AI Life Coach
-
-A personal development assistant powered by LLMs.
-
-## Features
-
-- Ollama & Hugging Face models
-- Redis-based memory
-- 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 `https://ace32bd59aef.ngrok-free.app`. 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.
-
-## Troubleshooting
-
-If you're experiencing connection issues with Ollama, you can run the diagnostic script:
-
-```
-python diagnose_ollama.py
-```
-
-This will test connectivity to your configured Ollama host and provide detailed information about any connection problems.

api/main.py

@@ -1,6 +1,7 @@
 from fastapi import FastAPI
 from api.status import router as status_router
 from api.chat import router as chat_router
+from core.memory import check_redis_health
 
 app = FastAPI()
 
@@ -11,3 +12,12 @@ app.include_router(chat_router, prefix="/api")
 @app.get("/")
 async def root():
     return {"message": "AI Life Coach API is running"}
+
+@app.get("/health")
+async def health_check():
+    """Health check endpoint"""
+    redis_healthy = check_redis_health()
+    return {
+        "status": "healthy" if redis_healthy else "degraded",
+        "redis": "healthy" if redis_healthy else "unhealthy"
+    }

app.py

@@ -1,4 +1,4 @@
-# Force redeploy trigger - version 1.8
+# Force redeploy trigger - version 1.9
 import streamlit as st
 from utils.config import config
 import requests
@@ -12,13 +12,10 @@ st.set_page_config(page_title="AI Life Coach", page_icon="🧘", layout="centere
 # Initialize session state
 if 'ngrok_url' not in st.session_state:
     st.session_state.ngrok_url = config.ollama_host
-
 if 'model_status' not in st.session_state:
     st.session_state.model_status = "checking"
-
 if 'available_models' not in st.session_state:
     st.session_state.available_models = []
-
 if 'selected_model' not in st.session_state:
     st.session_state.selected_model = config.local_model_name
 
@@ -47,7 +44,7 @@ NGROK_HEADERS = {
 def fetch_available_models(ngrok_url):
     try:
         response = requests.get(
-            f"{ngrok_url}/api/tags", 
+            f"{ngrok_url}/api/tags",
             headers=NGROK_HEADERS,
             timeout=5
         )
@@ -72,11 +69,11 @@ st.sidebar.markdown("---")
 st.sidebar.subheader("Model Selection")
 if st.session_state.available_models:
     selected_model = st.sidebar.selectbox(
-        "Select Model", 
+        "Select Model",
         st.session_state.available_models,
-        index=st.session_state.available_models.index(st.session_state.selected_model) 
-              if st.session_state.selected_model in st.session_state.available_models 
-              else 0
+        index=st.session_state.available_models.index(st.session_state.selected_model)
+        if st.session_state.selected_model in st.session_state.available_models
+        else 0
     )
     st.session_state.selected_model = selected_model
 else:
@@ -94,7 +91,7 @@ IS_HF_SPACE = bool(BASE_URL)
 def get_ollama_status(ngrok_url):
     try:
         response = requests.get(
-            f"{ngrok_url}/api/tags", 
+            f"{ngrok_url}/api/tags",
             headers=NGROK_HEADERS,
             timeout=10
         )
@@ -106,7 +103,7 @@ def get_ollama_status(ngrok_url):
             if models:
                 selected_model_available = st.session_state.selected_model in model_names
                 return {
-                    "running": True, 
+                    "running": True,
                     "model_loaded": st.session_state.selected_model if selected_model_available else model_names[0],
                     "remote_host": ngrok_url,
                     "available_models": model_names,
@@ -115,15 +112,23 @@ def get_ollama_status(ngrok_url):
             else:
                 st.session_state.model_status = "no_models"
                 return {
-                    "running": False, 
+                    "running": False,
                     "model_loaded": None,
                     "remote_host": ngrok_url,
                     "message": "Connected to Ollama but no models found"
                 }
+        else:
+            st.session_state.model_status = "unreachable"
+            return {
+                "running": False,
+                "model_loaded": None,
+                "error": f"HTTP {response.status_code}",
+                "remote_host": ngrok_url
+            }
     except Exception as e:
         st.session_state.model_status = "unreachable"
         return {
-            "running": False, 
+            "running": False,
             "model_loaded": None,
             "error": str(e),
             "remote_host": ngrok_url
@@ -139,9 +144,18 @@ def get_conversation_history(user_id):
         st.warning(f"Could not load conversation history: {e}")
     return []
 
-# Check Ollama status
+# Get Ollama status with null safety
 ollama_status = get_ollama_status(st.session_state.ngrok_url)
 
+# Add null safety check
+if ollama_status is None:
+    ollama_status = {
+        "running": False,
+        "model_loaded": None,
+        "error": "Failed to get Ollama status",
+        "remote_host": st.session_state.ngrok_url
+    }
+
 # Update model status
 if ollama_status.get("running", False):
     if ollama_status.get("available_models"):
@@ -151,9 +165,13 @@ if ollama_status.get("running", False):
 else:
     st.session_state.model_status = "unreachable"
 
-# Display Ollama status
+# Ensure ollama_status is a dict even if None
+ollama_status = ollama_status or {}
+
+# Determine if we should use fallback
 use_fallback = not ollama_status.get("running", False) or config.use_fallback
 
+# Display Ollama status
 if use_fallback:
     st.sidebar.warning("🌐 Using Hugging Face fallback (Ollama not available)")
     if "error" in ollama_status:
@@ -213,14 +231,12 @@ def send_to_ollama(user_input, conversation_history, ngrok_url, model_name):
                 "top_p": 0.9
             }
         }
-        
         response = requests.post(
             f"{ngrok_url}/api/chat",
             json=payload,
             headers=NGROK_HEADERS,
             timeout=60
         )
-        
         if response.status_code == 200:
             response_data = response.json()
             return response_data.get("message", {}).get("content", "")
@@ -237,7 +253,6 @@ def send_to_hf(user_input, conversation_history):
     try:
         from core.llm import LLMClient
         llm_client = LLMClient(provider="huggingface")
-        
         # Format for HF
         prompt = "You are a helpful life coach. "
         for msg in conversation_history:
@@ -246,7 +261,6 @@ def send_to_hf(user_input, conversation_history):
             elif msg["role"] == "assistant":
                 prompt += f"Assistant: {msg['content']} "
         prompt += "Assistant:"
-        
         response = llm_client.generate(prompt, max_tokens=500, stream=False)
         return response
     except Exception as e:
@@ -268,10 +282,9 @@ if st.button("Send"):
     else:
         # Display user message
         st.markdown(f"**You:** {user_input}")
-
+        
         # Prepare conversation history
-        conversation_history = [{"role": msg["role"], "content": msg["content"]} 
-                              for msg in conversation[-5:]]
+        conversation_history = [{"role": msg["role"], "content": msg["content"]} for msg in conversation[-5:]]
         conversation_history.append({"role": "user", "content": user_input})
         
         # Send to appropriate backend
@@ -281,8 +294,8 @@ if st.button("Send"):
                 backend_used = "Hugging Face"
             else:
                 ai_response = send_to_ollama(
-                    user_input, 
-                    conversation_history, 
+                    user_input,
+                    conversation_history,
                     st.session_state.ngrok_url,
                     st.session_state.selected_model
                 )

core/llm.py

@@ -8,7 +8,6 @@ from utils.config import config
 
 class LLMProvider(ABC):
     """Abstract base class for all LLM providers"""
-    
     def __init__(self, model_name: str, timeout: int = 30, retries: int = 3):
         self.model_name = model_name
         self.timeout = timeout
@@ -29,10 +28,9 @@ class LLMProvider(ABC):
                 last_exception = e
                 if attempt < self.retries:
                     time.sleep(1 * (2 ** attempt))  # Exponential backoff
-                continue
+                    continue
         raise last_exception
 
-
 class OllamaProvider(LLMProvider):
     def __init__(self, model_name: str, host: str = None, timeout: int = 30, retries: int = 3):
         super().__init__(model_name, timeout, retries)
@@ -78,10 +76,10 @@ class OllamaProvider(LLMProvider):
                 return stream_response()
             else:
                 return response.json()["response"]
-                
+        
+        # Fixed: Moved return outside the _make_request function
         return self._retry_request(_make_request)
 
-
 class HuggingFaceProvider(LLMProvider):
     def __init__(self, model_name: str, timeout: int = 30, retries: int = 3):
         super().__init__(model_name, timeout, retries)
@@ -108,10 +106,10 @@ class HuggingFaceProvider(LLMProvider):
                 return stream_response()
             else:
                 return response.choices[0].message.content
-                
+        
+        # Fixed: Moved return outside the _make_request function
         return self._retry_request(_make_request)
 
-
 class OpenAIProvider(LLMProvider):
     def __init__(self, model_name: str, api_key: str = None, timeout: int = 30, retries: int = 3):
         super().__init__(model_name, timeout, retries)
@@ -135,17 +133,17 @@ class OpenAIProvider(LLMProvider):
                 return stream_response()
             else:
                 return response.choices[0].message.content
-                
+        
+        # Fixed: Moved return outside the _make_request function
         return self._retry_request(_make_request)
 
-
 class LLMClient:
     PROVIDER_MAP = {
         "ollama": OllamaProvider,
         "huggingface": HuggingFaceProvider,
         "openai": OpenAIProvider
     }
-    
+
     def __init__(self, provider: str = "ollama", model_name: str = None, **provider_kwargs):
         self.provider_name = provider.lower()
         self.model_name = model_name or self._get_default_model()
@@ -155,7 +153,7 @@ class LLMClient:
             
         provider_class = self.PROVIDER_MAP[self.provider_name]
         self.provider = provider_class(self.model_name, **provider_kwargs)
-    
+
     def _get_default_model(self) -> str:
         """Get default model based on provider"""
         defaults = {
@@ -164,11 +162,11 @@ class LLMClient:
             "openai": "gpt-3.5-turbo"
         }
         return defaults.get(self.provider_name, "mistral")
-    
+
     def generate(self, prompt: str, max_tokens: int = 500, stream: bool = False) -> Union[str, Generator[str, None, None]]:
         """Unified generate method that delegates to provider"""
         return self.provider.generate(prompt, max_tokens, stream)
-    
+
     @classmethod
     def get_available_providers(cls) -> list:
         """Return list of supported providers"""

ngrok.yml.txt

@@ -0,0 +1,9 @@
+version: "2"
+authtoken: 32HaXMF3tuRxfas1siT3CIhLjH4_2AXbGGma38NnCF1tjyJNZ
+tunnels:
+  ai-coach-api:
+    addr: 8000
+    proto: http
+  ai-coach-ui:
+    addr: 8501
+    proto: http

test_setup.py

@@ -0,0 +1,33 @@
+import requests
+import os
+from dotenv import load_dotenv
+
+load_dotenv()
+
+# Test Ollama
+ollama_host = os.getenv("OLLAMA_HOST", "http://localhost:11434")
+model_name = os.getenv("LOCAL_MODEL_NAME", "mistral:latest")
+
+print(f"Testing Ollama at: {ollama_host}")
+try:
+    response = requests.get(f"{ollama_host}/api/tags")
+    print(f"Ollama Status: {response.status_code}")
+    print(f"Models available: {response.json()}")
+except Exception as e:
+    print(f"Ollama Error: {e}")
+
+# Test model generation
+print(f"\nTesting model: {model_name}")
+try:
+    response = requests.post(f"{ollama_host}/api/generate", json={
+        "model": model_name,
+        "prompt": "Hello, world!",
+        "stream": False
+    })
+    print(f"Model Test Status: {response.status_code}")
+    if response.status_code == 200:
+        print("✅ Ollama and model are working correctly!")
+    else:
+        print(f"❌ Model test failed: {response.text}")
+except Exception as e:
+    print(f"Model Test Error: {e}")

utils/config.py

@@ -6,15 +6,20 @@ class Config:
         load_dotenv()
         self.hf_token = os.getenv("HF_TOKEN")
         self.hf_api_url = os.getenv("HF_API_ENDPOINT_URL")
+        self.use_fallback = os.getenv("USE_FALLBACK", "false").lower() == "true"
         self.tavily_api_key = os.getenv("TAVILY_API_KEY")
         self.openweather_api_key = os.getenv("OPENWEATHER_API_KEY")
         self.nasa_api_key = os.getenv("NASA_API_KEY")
-        self.redis_host = os.getenv("REDIS_HOST")
+        
+        # Redis configuration with proper defaults
+        self.redis_host = os.getenv("REDIS_HOST", "localhost")
         self.redis_port = int(os.getenv("REDIS_PORT", "6379"))
         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")
-        # Use the correct remote Ollama host from the ngrok warning page
+        self.redis_retries = int(os.getenv("REDIS_RETRIES", "3"))
+        self.redis_retry_delay = int(os.getenv("REDIS_RETRY_DELAY", "1"))
+        
+        self.local_model_name = os.getenv("LOCAL_MODEL_NAME", "mistral")
         self.ollama_host = os.getenv("OLLAMA_HOST", "https://ace32bd59aef.ngrok-free.app")
 
 config = Config()