Add application file and dependencies

app.py

@@ -0,0 +1,34 @@
+import os
+import spaces  # Activates Hugging Face's free ZeroGPU
+import gradio as gr
+from huggingface_hub import hf_hub_download
+from llama_cpp import Llama
+
+# 1. Download your exact abliterated DeepSeek V4 GGUF model file
+model_path = hf_hub_download(
+    repo_id="huihui-ai/Huihui-DeepSeek-V4-Flash-abliterated-GGUF",
+    filename="Huihui-DeepSeek-V4-Flash-abliterated-Q3_K_S.gguf"
+)
+
+# Initialize the model engine in memory
+llm = Llama(model_path=model_path, n_ctx=4096)
+
+# 2. Define the raw prediction function using the Cloud GPU
+@spaces.GPU
+def run_api(prompt, max_tokens=1024):
+    output = llm(
+        f"<｜User｜>{prompt}<｜Assistant｜>",
+        max_tokens=int(max_tokens),
+        stop=["<｜User｜>", "<｜Assistant｜>"]
+    )
+    return output["choices"][0]["text"]
+
+# 3. Create the API routing endpoint (Bypassing visual elements)
+demo = gr.Interface(
+    fn=run_api,
+    inputs=[gr.Textbox(label="prompt"), gr.Number(value=1024, label="max_tokens")],
+    outputs=gr.Textbox(label="response"),
+    api_name="predict"  # <--- This creates the raw web API endpoint
+)
+
+demo.launch()

requirements.txt

@@ -0,0 +1,2 @@
+llama-cpp-python
+huggingface_hub