Remove rate-limited Groq models (llama-3.1 variants, gemma) and use Mixtral as default

app/groq_client.py

@@ -10,7 +10,7 @@ load_dotenv()
 _GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
 _client = AsyncGroq(api_key=_GROQ_API_KEY) if _GROQ_API_KEY else None
 
-DEFAULT_DECISION_MODEL = "llama-3.1-8b-instant"
+DEFAULT_DECISION_MODEL = "mixtral-8x7b-32768"
 MAX_AGENT_SPEED = 80
 
 

app/hf_spaces.py

@@ -43,10 +43,8 @@ KNOWN_SPACES_MODELS = [
 
 # Groq models (built-in)
 GROQ_MODELS = [
-    {"id": "llama-3.1-8b-instant", "name": "Llama 3.1 8B", "backend": "groq"},
-    {"id": "llama-3.1-70b-versatile", "name": "Llama 3.1 70B", "backend": "groq"},
     {"id": "mixtral-8x7b-32768", "name": "Mixtral 8x7B", "backend": "groq"},
-    {"id": "gemma-7b-it", "name": "Gemma 7B", "backend": "groq"},
+    {"id": "llama2-70b-4096", "name": "Llama 2 70B", "backend": "groq"},
 ]