get available model

app.py

@@ -44,11 +44,49 @@ final_answer = FinalAnswerTool()
 # model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud' 
 # model_id='Qwen/Qwen2.5-Coder-32B-Instruct',# it is possible that this model may be overloaded
 
+MODEL_IDS = [
+    #'https://wxknx1kg971u7k1n.us-east-1.aws.endpoints.huggingface.cloud/',
+    #'https://jc26mwg228mkj8dw.us-east-1.aws.endpoints.huggingface.cloud/', 
+    #'meta-llama/Llama-3.2-1B-Instruct', ## Does a poor job of interpreting my questions and matching them to the tools
+    'Qwen/Qwen2.5-Coder-32B-Instruct',
+    'Qwen/Qwen2.5-Coder-14B-Instruct',
+    'Qwen/Qwen2.5-Coder-7B-Instruct',
+    'Qwen/Qwen2.5-Coder-3B-Instruct',
+    'Qwen/Qwen2.5-Coder-1.5B-Instruct'
+    # Add here wherever model is working for you
+]
+
+def is_model_overloaded(model_url):
+    """Verify if the model is overloaded doing a test call."""
+    try:
+        response = requests.post(model_url, json={"inputs": "Test"})
+        if verbose: 
+            print(response.status_code)
+        if response.status_code == 503:  # 503 Service Unavailable = Overloaded
+            return True
+        if response.status_code == 404:  # 404 Client Error: Not Found 
+            return True
+        if response.status_code == 424:  # 424 Client Error: Failed Dependency for url:
+            return True
+        return False
+    except requests.RequestException:
+        return True  # if there are an error is overloaded
+
+def get_available_model():
+    """Select the first model available from the list."""
+    for model_url in MODEL_IDS:
+        print("trying",model_url)
+        if not is_model_overloaded(model_url):
+            return model_url
+    return MODEL_IDS[0]  # if all are failing, use the first model by dfault
+
+selected_model_id = get_available_model()
+
 model = HfApiModel(
-max_tokens=2096,
-temperature=0.5,
-model_id='https://pflgm2locj2t89co.us-east-1.aws.endpoints.huggingface.cloud',
-custom_role_conversions=None,
+    max_tokens=2096,
+    temperature=0.5,
+    model_id=selected_model_id,
+    custom_role_conversions=None,
 )
 
 
@@ -60,7 +98,7 @@ with open("prompts.yaml", 'r') as stream:
     
 agent = CodeAgent(
     model=model,
-    tools=[image_generation_tool,final_answer], ## add your tools here (don't remove final answer)
+    tools=[final_answer], ## add your tools here (don't remove final answer)
     max_steps=6,
     verbosity_level=1,
     grammar=None,