Update app.py

# Limit concurrency (HF-compatible)
demo.queue(concurrency_count=1)

app.py

@@ -14,9 +14,16 @@ from huggingface_hub.errors import HfHubHTTPError
 
 HF_TOKEN = os.getenv("HF_TOKEN", None)
 
-# Recommended: 7B is more stable on free HF router
-# LLM_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
-LLM_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-14B"
+# Free model compatible with new HF API
+#LLM_MODEL = "meta-llama/Llama-3.2-1B-Instruct"
+#LLM_MODEL = "meta-llama/Llama-3.1-8B-Instruct"
+# LLM_MODEL = "mistralai/Mistral-7B-Instruct-v0.3" Does not work
+# LLM_MODEL ="HuggingFaceH4/zephyr-7b-beta"
+# LLM_MODEL = "NousResearch/Hermes-2-Pro-Mistral-7B" Does not work
+
+# Great LLM
+LLM_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
+#LLM_MODEL = "deepseek-ai/DeepSeek-R1-Distill-Qwen-14B"
 
 EMBEDDING_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
 
@@ -176,7 +183,7 @@ def answer_question(question, state):
 # GRADIO UI
 # ============================================================
 
-with gr.Blocks(title="PDF Q&A Chatbot", concurrency_limit=1) as demo:
+with gr.Blocks(title="PDF Q&A Chatbot") as demo:
     gr.Markdown("## 📄 PDF Q&A Chatbot\nUpload a PDF and ask questions about its content.")
 
     state = gr.State({"chunks": None, "index": None})
@@ -194,5 +201,8 @@ with gr.Blocks(title="PDF Q&A Chatbot", concurrency_limit=1) as demo:
     pdf_input.upload(load_pdf, inputs=[pdf_input, state], outputs=[pdf_status, state])
     ask_button.click(answer_question, inputs=[question, state], outputs=answer)
 
+# Limit concurrency (HF-compatible)
+demo.queue(concurrency_count=1)
+
 if __name__ == "__main__":
     demo.launch()