fix HF Gradio error

app.py

@@ -31,15 +31,13 @@ else:
 # --- Inference API client ---
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta", token=os.getenv("HF_TOKEN"))
 
-def respond(message, history: list[dict]):
-    # 1. encode query and retrieve context
+def respond(message, history):
     query_embedding = embedding_model.encode([message], convert_to_numpy=True)
     faiss.normalize_L2(query_embedding)
     _, indices = index.search(query_embedding, k=5)
 
     context = "\n".join([f"{codes[i]}: {descriptions[i]}" for i in indices[0]])
 
-    # 2. prepare system prompt with role + retrieved context
     system_prompt = f"""You are an expert assistant specialized in tariff classification.
 Your job is to help users find the most appropriate tariff codes based on their description.
 Use only the provided context below to answer.
@@ -48,23 +46,22 @@ Context:
 {context}
 """
 
-    # 3. insert system message at the beginning
     messages = [{"role": "system", "content": system_prompt}]
     messages += history + [{"role": "user", "content": message}]
 
-    response = {"role": "assistant", "content": ""}
-
-    for message in client.chat_completion(
+    full_response = ""
+    for chunk in client.chat_completion(
         messages,
         max_tokens=512,
         stream=True,
         temperature=0.7,
         top_p=0.95,
     ):
-        token = message.choices[0].delta.content
-        response["content"] += token
-        formatted = response["content"].replace("\n", "\n\n")
-        yield {"role": "assistant", "content": formatted}
+        token = chunk.choices[0].delta.content
+        if token:
+            full_response += token
+            yield full_response.replace("\n", "\n\n")
+
 
 demo = gr.ChatInterface(
     respond, type="messages",