Update app.py

app.py

@@ -1,9 +1,16 @@
 import gradio as gr
-import requests
+from huggingface_hub import InferenceClient
 import os
 
 HF_API_TOKEN = os.getenv("HF_API_TOKEN")
 MODEL = "bigscience/bloom-560m" #"microsoft/Phi-3-mini-128k-instruct"
+# Initialize the InferenceClient with the token and model.
+# Using a lightweight model like google/flan-t5-small.
+client = InferenceClient(
+    "google/flan-t5-small",
+    token=HF_TOKEN
+)
+
 
 # -------------------------
 # Core function: call HF API
@@ -15,18 +22,16 @@ def query_llm(tweet, mode):
     else:
         prompt += "Task: Is the above TEXT toxic? First clearly state your decision. Then, provide specific reason(s) for your decision. If there is more than one reason, provide them in a numbered list. Your reason(s) must be non-redundant and jointly sufficient to justify your decision. In other words, there should not be any internal or external information unused in your explanation."
     
-    headers = {"Authorization": f"Bearer {HF_API_TOKEN}"}
-    response = requests.post(
-        f"https://api-inference.huggingface.co/models/{MODEL}",
-        headers=headers,
-        json={"inputs": prompt}
-    )
+    # Use the client to generate text from the full prompt.
     try:
-        out = response.json()
-        return out[0]["generated_text"] if isinstance(out, list) else str(out)
+        response = client.text_generation(
+            prompt=prompt,
+            max_new_tokens=250,  # Control the length of the generated output.
+            stream=False,       # Set to True for streaming responses.
+        )
+        return response
     except Exception as e:
-        return f"Error: {str(e)}"
-
+        return f"Error: {e}"
 
 # -------------------------
 # Preloaded tweets