Update app.py

app.py

@@ -4,14 +4,12 @@ import pandas as pd
 from huggingface_hub import InferenceClient
 from threading import Timer
 from tqdm import tqdm
+import time
 
-HUGGINGFACE_TOKEN =os.environ.get("HUGGINGFACE_TOKEN")
-def get_available_free(use_cache = False):
-    if use_cache:
-        if os.path.exists(str(os.getcwd())+"/data.csv"):
-            # print("Loading data from file...")
-            return pd.read_csv("data.csv").to_dict(orient='list')
-            
+HUGGINGFACE_TOKEN = os.environ.get("HUGGINGFACE_TOKEN")
+
+def loop_query_data():
+    global all_models
     models_dict = InferenceClient(token=HUGGINGFACE_TOKEN).list_deployed_models("text-generation-inference")
     models = models_dict['text-generation'] + models_dict['text2text-generation']
     models_vision = models_dict['image-text-to-text']
@@ -25,7 +23,7 @@ def get_available_free(use_cache = False):
         "Vision": []
     }
 
-    all_models = list(set(models + models_vision + models_others))
+    all_models = list(set(all_models + models + models_vision + models_others))
     for m in tqdm(all_models):
         text_available = False
         chat_available = False
@@ -41,11 +39,8 @@ def get_available_free(use_cache = False):
             if e and "Model requires a Pro subscription" in str(e):
                 pro_sub = True
             if e and "Rate limit reached" in str(e):
-                # print("Rate Limited!!")
-                if os.path.exists(str(os.getcwd())+"/data.csv"):
-                    # print("Loading data from file...")
-                    return pd.read_csv(str(os.getcwd())+"/data.csv").to_dict(orient='list')
-                return []
+                print("Rate Limited, waiting 1 hour...")
+                time.sleep(60*60)
         try:
             InferenceClient(m, timeout=10).chat_completion(messages=[{'role': 'user', 'content': 'Hi.'}], max_tokens=1)
             chat_available = True
@@ -54,11 +49,8 @@ def get_available_free(use_cache = False):
             if e and "Model requires a Pro subscription" in str(e):
                 pro_sub = True
             if e and "Rate limit reached" in str(e):
-                # print("Rate Limited!!")
-                if os.path.exists("data.csv"):
-                    # print("Loading data from file...")
-                    return pd.read_csv(str(os.getcwd())+"/data.csv").to_dict(orient='list')
-                return []
+                print("Rate Limited, waiting 1 hour...")
+                time.sleep(60*60)
         models_conclusion["Model"].append(m)
         models_conclusion["API"].append("Free" if chat_available or text_available else ("Pro Subscription" if pro_sub else "Not Responding"))
         models_conclusion["Chat Completion"].append("---" if (pro_sub or (not chat_available and not text_available)) else ("✓" if chat_available else "⌀"))
@@ -67,6 +59,14 @@ def get_available_free(use_cache = False):
     pd.DataFrame(models_conclusion).to_csv(str(os.getcwd())+"/data.csv", index=False)
     return models_conclusion
 
+def get_available_free(use_cache = False):
+    if use_cache:
+        if os.path.exists(str(os.getcwd())+"/data.csv"):
+            # print("Loading data from file...")
+            return pd.read_csv("data.csv").to_dict(orient='list')
+    else:
+        return loop_query_data()
+
 def update_data(use_cache = False):
     data = get_available_free(use_cache)
     df = pd.DataFrame(data)
@@ -157,11 +157,11 @@ print(response)
 ```
 """
 first_run = True
+all_models = []
 with gr.Blocks() as demo:
     gr.Markdown("## HF Serverless LLM Inference API Status")
     gr.Markdown(description)
     search_box = gr.Textbox(label="Search for a model", placeholder="Type model name here...")
-    gr.Markdown("### Cached Endpoints")
     filter_box = gr.CheckboxGroup(choices=["Free", "Pro Subscription", "Not Responding", "Text Completion", "Chat Completion", "Vision"], label="Filters")
     table = gr.Dataframe(value=display_table(use_cache=True), headers="keys")