Update app.py

app.py

@@ -3,9 +3,20 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 import requests
 
-url = "https://huggingface.co/models-json?num_parameters=min%3A76B&inference_provider=novita&sort=trending&withCount=true"
-res = requests.get(url).json()
-model_ids = [m['id'] for m in res.get('models', [])]
+def fetch_models():
+    models = []
+    page = 1
+    while True:
+        url = f"https://huggingface.co/models-json?num_parameters=min%3A0B&inference_provider=novita&p={page}&sort=modified&withCount=true"
+        res = requests.get(url).json()
+        page_models = [m['id'] for m in res.get('models', [])]
+        if not page_models:
+            break
+        models.extend(page_models)
+        page += 1
+    return models
+
+model_ids = fetch_models()
 
 client = InferenceClient(provider="novita", api_key=os.environ["HF_TOKEN"])
 
@@ -18,11 +29,12 @@ def respond(message, history, model):
         messages=[{"role": "user", "content": message}],
         stream=True
     ):
-        content = chunk.choices[0].delta.content
-        if content:
-            full_reply += content
-            history[-1] = (message, full_reply)
-            yield "", history
+        if chunk.choices and len(chunk.choices) > 0:
+            delta = chunk.choices[0].delta
+            if hasattr(delta, 'content') and delta.content:
+                full_reply += delta.content
+                history[-1] = (message, full_reply)
+                yield "", history
 
 with gr.Blocks() as demo:
     gr.Markdown("## AI")