Update app.py

app.py

@@ -16,12 +16,15 @@ import bm25s
 import aiohttp
 from bs4 import BeautifulSoup
 
-# 🔥 CRITICAL HOTFIX: Prevents Optimum from crashing on newer Transformers versions
+# 🔥 CRITICAL HOTFIX FOR HUGGING FACE SPACES
+# Bypasses the Optimum crash with the latest Transformers version natively on Python 3.13
 import transformers.utils
-if not hasattr(transformers.utils, 'is_tf_available'):
-    transformers.utils.is_tf_available = lambda: False
-if not hasattr(transformers.utils, 'is_torch_available'):
-    transformers.utils.is_torch_available = lambda: True
+import transformers.utils.import_utils
+for mod in [transformers.utils, transformers.utils.import_utils]:
+    if not hasattr(mod, 'is_tf_available'):
+        mod.is_tf_available = lambda: False
+    if not hasattr(mod, 'is_torch_available'):
+        mod.is_torch_available = lambda: True
 
 from transformers import AutoTokenizer, pipeline
 from optimum.onnxruntime import ORTModelForCausalLM
@@ -31,13 +34,13 @@ import onnxruntime as ort
 TOTAL_CORES = multiprocessing.cpu_count()
 AI_CORES = max(1, int(TOTAL_CORES * 0.4))
 
-# Strictly enforce CPU Affinity (Locks the process to specific cores)
+# Try to strictly enforce CPU Affinity, fallback safely if Docker/Space restricts permissions
 try:
     p = psutil.Process()
     p.cpu_affinity(list(range(AI_CORES)))
     print(f"🖥️ Total Cores: {TOTAL_CORES} | 🔒 Process strictly bound to {AI_CORES} cores.")
-except AttributeError:
-    print(f"🖥️ Total Cores: {TOTAL_CORES} | Allocating {AI_CORES} cores to AI.")
+except Exception as e:
+    print(f"🖥️ Total Cores: {TOTAL_CORES} | Allocating {AI_CORES} cores to AI. (Strict affinity skipped: {e})")
 
 raw_data = []
 suggestion_list = []
@@ -386,5 +389,6 @@ async def ai_mode_chat_get(q: str = Query(...)):
 async def ai_mode_chat_post(q: str = Form(...)):
     return await process_rag_chat(q)
 
+# Using `app` variable directly prevents Spaces routing errors if the file is named differently
 if __name__ == "__main__":
-    uvicorn.run("main:app", host="0.0.0.0", port=7860, workers=1)
+    uvicorn.run(app, host="0.0.0.0", port=7860, workers=1)