feat: implement LLM provider client with automated fallback and Gemini model support

llm/client.py

@@ -4,7 +4,7 @@ from typing import List
 from langchain_core.messages import AIMessage
 from llm.providers import PROVIDERS
 
-PROVIDER_ORDER = os.getenv("LLM_PROVIDER_ORDER", "gemini_gemma, gemini, groq").split(",")
+PROVIDER_ORDER = os.getenv("LLM_PROVIDER_ORDER", "groq, gemini, gemini_gemma").split(",")
 
 _degraded_providers = {}
 
@@ -50,14 +50,14 @@ def invoke_llm(messages: List, tools: List, fallback_count: int = 0) -> AIMessag
     except Exception as e:
         error_msg = str(e).lower()
 
-        if "rate limit" in error_msg or "429" in error_msg:
+        if "rate limit" in error_msg or "429" in error_msg or "quota" in error_msg:
             print(f"{provider_name} rate limit hit. Waiting before retry...")
             import time
             wait_time = 10 * (fallback_count + 1)
             time.sleep(wait_time)
-
-        print(f"{provider_name} failed: {e}. Marking as degraded.")
-        _degraded_providers[provider_name] = True
+            _degraded_providers[provider_name] = True
+        else:
+            print(f"{provider_name} error: {e}. Trying next provider.")
 
         remaining = [n for n in PROVIDER_ORDER if n not in _degraded_providers]
         if remaining:

llm/providers/gemini.py

@@ -1,9 +1,14 @@
+import os
+from dotenv import load_dotenv
 from langchain_google_genai import ChatGoogleGenerativeAI
 
+load_dotenv()
+GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
+
 
 def invoke(messages, tools, model_name: str = "gemini-2.0-flash"):
     """Invoke Gemini models (free tier)."""
-    model = ChatGoogleGenerativeAI(model=model_name, temperature=0)
+    model = ChatGoogleGenerativeAI(model=model_name, temperature=0, google_api_key=GOOGLE_API_KEY)
     model_with_tools = model.bind_tools(tools)
     return model_with_tools.invoke(messages)
 

llm/providers/gemini_gemma.py

@@ -1,9 +1,14 @@
+import os
+from dotenv import load_dotenv
 from langchain_google_genai import ChatGoogleGenerativeAI
 
+load_dotenv()
+GOOGLE_API_KEY = os.getenv("GOOGLE_API_KEY")
+
 
 def invoke(messages, tools, model_name: str = "gemma-2-27b-it"):
     """Invoke Google Gemma models (free tier)."""
-    model = ChatGoogleGenerativeAI(model=model_name, temperature=0)
+    model = ChatGoogleGenerativeAI(model=model_name, temperature=0, google_api_key=GOOGLE_API_KEY)
     model_with_tools = model.bind_tools(tools)
     return model_with_tools.invoke(messages)