Update config.py

config.py

@@ -4,6 +4,7 @@ import json
 from datetime import datetime
 import time
 import threading
+import leaderboard
 
 ARENA_NAME = "# 🏆 The GPU-Poor LLM Gladiator Arena 🏆 v25.10"
 
@@ -61,6 +62,10 @@ def load_approved_models():
 MODEL_REFRESH_INTERVAL = 3600  # Check every 1 hour
 _last_model_check = 0  # Track last check time
 
+# Model archiving configuration
+ARCHIVE_BATTLE_THRESHOLD = int(os.environ.get("ARCHIVE_BATTLE_THRESHOLD", 150))
+ARCHIVED_MODELS_PATH = os.environ.get("NEXTCLOUD_ARCHIVED_MODELS_PATH", "/gpu_poor_archived_models.json")
+
 def get_approved_models():
     """Get the current list of approved models with periodic refresh."""
     global _last_model_check
@@ -72,7 +77,7 @@ def get_approved_models():
         get_approved_models._models = load_approved_models()
         _last_model_check = current_time
     
-    return get_approved_models._models
+    return [model for model in get_approved_models._models if model[0] not in leaderboard.load_archived_models()]
 
 def refresh_approved_models():
     """Force refresh of the approved models list."""
@@ -80,6 +85,9 @@ def refresh_approved_models():
         delattr(get_approved_models, '_models')
     return get_approved_models()
 
+# Maximum tokens for model generation to prevent timeouts
+MAX_TOKENS = int(os.environ.get("MAX_TOKENS", 2048)) # Increased to 2048 based on feedback
+
 # Keep FALLBACK_MODELS as a safety net
 FALLBACK_MODELS = [
     ("hf.co/unsloth/Qwen3-4B-Instruct-2507-GGUF:Q8_K_XL", "Qwen 3 Instruct 2507 Unsloth (4B, 8-bit)")