Update app.py

app.py

@@ -1,11 +1,56 @@
 import os
-from huggingface_hub import InferenceApi
+from fastapi import FastAPI
+from pydantic import BaseModel
+from huggingface_hub import InferenceClient
+from smolagent import BasicAgent
+from crewai_tools import BaseTool
 
-inference_api = InferenceApi(
-    repo_id="Qwen/Qwen1.5-1.8B-Chat",
-    token=os.environ["HUGGINGFACE_API_KEY"]
+# Use your token from HF secrets
+HF_TOKEN = os.environ.get("HF_TOKEN")  # Must be set in HF Space secrets
+
+# Define a wrapper that conforms to SmolAgent’s LLM expectations
+class HuggingFaceInferenceWrapper:
+    def __init__(self, client: InferenceClient):
+        self.client = client
+
+    def __call__(self, prompt: str, **kwargs):
+        # Use text_generation endpoint
+        response = self.client.text_generation(prompt=prompt, max_new_tokens=512)
+        return response
+
+# Initialize InferenceClient
+client = InferenceClient(
+    model="Qwen/Qwen1.5-1.8B-Chat",  # Or another model you have access to
+    token=HF_TOKEN
 )
 
+# Wrap the client
+llm = HuggingFaceInferenceWrapper(client)
+
+# Define a dummy tool
+class DuckDuckGoTool(BaseTool):
+    name: str = "DuckDuckGo Search"
+    description: str = "Use this tool to search for real-time facts and current events."
+
+    def _run(self, query: str) -> str:
+        return f"Search results for '{query}' (this is a placeholder)."
+
+# Create the agent
+agent = BasicAgent(
+    role="Helpful AI",
+    goal="Answer tough questions using search and reasoning",
+    backstory="An expert at finding information and answering questions.",
+    tools=[DuckDuckGoTool()],
+    llm=llm
+)
+
+# Define FastAPI app
+app = FastAPI()
+
+class QuestionInput(BaseModel):
+    question: str
 
-llm = HuggingFaceInferenceWrapper(inference_api)
-print(llm.generate("What is the capital of France?"))
+@app.post("/ask")
+def ask_question(payload: QuestionInput):
+    result = agent.run(payload.question)
+    return {"answer": result}