Switch to OpenAI client with Kimi-K3:fastest

- Use OpenAI router API (router.huggingface.co/v1)
- Drop InferenceClient + Together provider, model is 'moonshotai/Kimi-K3:fastest'

Co-Authored-By: Claude <noreply@anthropic.com>

app.py

@@ -5,10 +5,9 @@ import urllib.request
 
 from fastapi.responses import HTMLResponse
 from gradio import Server
-from huggingface_hub import InferenceClient
+from openai import OpenAI
 
-MODEL = "moonshotai/Kimi-K3"
-PROVIDER = "together"
+MODEL = "moonshotai/Kimi-K3:fastest"
 
 SYSTEM_PROMPT = (
     "You are Kimi K3, Moonshot AI's open-weight native multimodal agentic model. "
@@ -57,7 +56,11 @@ def chat(message: dict, history: list) -> dict:
         yield {"role": "assistant", "content": "HF_TOKEN is not configured on this Space. Set it in the Space settings → Secrets."}
         return
 
-    client = InferenceClient(api_key=token, provider=PROVIDER, timeout=600)
+    client = OpenAI(
+        base_url="https://router.huggingface.co/v1",
+        api_key=token,
+        timeout=600,
+    )
 
     messages = [{"role": "system", "content": SYSTEM_PROMPT}]
     for msg in history: