Update app.py

app.py

@@ -46,7 +46,9 @@ Always respond with the chord name and a short explanation of the intervals.
 
 # Read token from Secrets (set in Space -> Settings -> Repository secrets)
 HF_TOKEN = os.environ.get("HF_TOKEN")
-client = InferenceClient(token=HF_TOKEN, model="google/flan-t5-small")
+
+# Create inference client
+client = InferenceClient(token=HF_TOKEN)
 
 def respond(message, history, system_message, max_tokens, temperature, top_p):
     if HF_TOKEN is None:
@@ -57,19 +59,17 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         prompt = f"{system_message}\nUser: {message}\nAnswer:"
         response = ""
 
-        for chunk in client.post(
-            json={
-                "inputs": prompt,
-                "parameters": {
-                    "max_new_tokens": max_tokens,
-                    "temperature": temperature,
-                    "top_p": top_p,
-                },
-                "stream": True,
-            }
+        for chunk in client.text_generation(
+            prompt,
+            model="google/flan-t5-small",   # ✅ ensure correct model is used
+            max_new_tokens=max_tokens,
+            temperature=temperature,
+            top_p=top_p,
+            stream=True,
         ):
-            if "token" in chunk:
-                token = chunk["token"]["text"]
+            # chunks are StreamingResponse objects
+            if hasattr(chunk, "token") and chunk.token and chunk.token.text:
+                token = chunk.token.text
                 response += token
                 yield response.strip()
 
@@ -81,7 +81,6 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
         yield f"⚠️ Error: {str(e)}\n\nTraceback:\n{tb}"
 
 
-
 chatbot = gr.ChatInterface(
     fn=respond,
     additional_inputs=[