Testing HF InterfaceClient

app.py

@@ -178,6 +178,11 @@ if __name__ == "__main__":
         print("ℹ️  SPACE_ID environment variable not found (running locally?). Repo URL cannot be determined.")
 
     print("-"*(60 + len(" App Starting ")) + "\n")
+    print("\n" + "*"*30 + " Debug area " + "*"*30)
+    from utils import callHfInferenceClientLLM
+    response = callHfInferenceClientLLM("What is the capital of France?")
+    print(response)
+    print("\n" + "*"*30 + " Debug area " + "*"*30)
 
     print("Launching Gradio Interface for Basic Agent Evaluation...")
     demo.launch(debug=True, share=False)

config.py

@@ -3,9 +3,7 @@ baseApiUrl = "https://agents-course-unit4-scoring.hf.space"
 questionsUrl = f"{baseApiUrl}/questions"
 submitUrl = f"{baseApiUrl}/submit"
 
-runLocal = False
-localModel = "llama3.2"
-hfMoldel = "(tbd)"
+hfMoldel = "meta-llama/Llama-3.3-70B-Instruct"
 
 logLevel = 'DEBUG'
 logFile = 'log.txt'

utils.py

@@ -1,19 +1,19 @@
 
-import config
-if config.runLocal:
-    from ollama import chat as OllamaChat
+import os
 
+from huggingface_hub import InferenceClient
 from langchain_community.tools import DuckDuckGoSearchRun
+import config
 
 
 def callWebSearch(query):
     return DuckDuckGo(query)
 
 def callLLM(query):
-    if config.runLocal:
+    if "LOCALLLM" in os.environ:
         return callLocalLLM(query)
     else:
-        return callHfLLM(query)
+        return callHfInferenceClientLLM(query)
 
 def DuckDuckGo(query):
     search_tool = DuckDuckGoSearchRun()
@@ -21,12 +21,22 @@ def DuckDuckGo(query):
     return results
 
 def callLocalLLM(query):
-    response = OllamaChat(model=config.localModel, messages=[ { 'role': 'user', 'content': query } ])
+    response = OllamaChat(model=os.environ["LOCALLLM"], messages=[ { 'role': 'user', 'content': query } ])
     return response['message']['content']
 
-def callHfLLM(query):
-    return f"No way to call {config.hfMoldel} yet"
+def callHfInferenceClientLLM(query):
+    client = InferenceClient(config.hfMoldel)
+    response = client.chat.completions.create(
+        messages = [ {"role": "user", "content": query } ],
+        stream=False, max_tokens=1024 )
+    return response.choices[0].message.content
 
 if __name__ == "__main__":
+    os.environ["LOCALLLM"] = "llama3.2"
+    from ollama import chat as OllamaChat
+    response = callLLM("What is the capital of France?")
+    print(response)
     response = callWebSearch("who is the president of France")
+    print(response)
+    response = callHfInferenceClientLLM("What is the capital of France?")
     print(response)