works now.

app.py

@@ -42,7 +42,7 @@ ADAPTER_SUBFOLDER = os.getenv("ADAPTER_SUBFOLDER") or None
 ATTN_IMPL         = os.getenv("ATTN_IMPL", "eager")
 DTYPE             = DTYPE_MAP.get(os.getenv("DTYPE", "bf16").lower(), torch.bfloat16)
 SYSTEM_DEF        = os.getenv("SYSTEM_PROMPT", "You are Mirel, a memory-stable symbolic assistant.")
-MAX_DEF           = int(os.getenv("MAX_NEW_TOKENS", "1024"))
+MAX_DEF           = int(os.getenv("MAX_NEW_TOKENS", "256"))
 ZEROGPU           = os.getenv("ZEROGPU", os.getenv("ZERO_GPU", "0")) == "1"
 LOAD_4BIT         = os.getenv("LOAD_4BIT", "0") == "1"
 
@@ -528,5 +528,5 @@ if __name__ == "__main__":
     demo.queue(max_size=8 if ZEROGPU else 32).launch(
         server_name="0.0.0.0", 
         server_port=7860,
-        share=True
-    )
+        share=False
+    )

requirements.txt

@@ -2,7 +2,6 @@ huggingface_hub>=0.34.0
 transformers>=4.55.0
 accelerate>=0.33.0
 peft>=0.11.0
-gradio==4.19.2
 torch>=2.4.0   # ZeroGPU-supported (2.3.x is NOT supported)
 bitsandbytes>=0.43.1
 openai_harmony