Update app.py

app.py

@@ -1,38 +1,32 @@
-# build on your original chatbot from the previous lesson
-# a basic chatbot from the previous lesson is below -- edit it to incorporate the changes described above
-
 import gradio as gr
-from huggingface_hub import InferenceClient #imports huggingface models
+from huggingface_hub import InferenceClient
 
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 
 def respond(message, history):
     messages = [{"role": "system", "content": "I am a kind chatbot."}]
 
-# add all previous messages to the messages list
-    if history:
-        messages.extend(history)
+    # Add all previous messages to the messages list
+    for human, assistant in history:
+        messages.append({"role": "user", "content": human})
+        messages.append({"role": "assistant", "content": assistant})
 
-    # add the current user's message to the messages list
+    # Add the current user's message to the messages list
     messages.append({"role": "user", "content": message})
 
-    # makes the chat completion API call,
-    # sending the messages and other parameters to the model
-    # implements streaming, where one word/token appears at a time
     response = ""
-
-    # iterate through each message in the method
-    for message in client.chat_completion(
+    for chunk in client.chat_completion(
         messages,
         max_tokens=100,
-        temperature=.1,
-        stream=True):
-
-        # add the tokens to the output content
-          token = message.choices[0].delta.content # capture the most recent toke
-          response += token # Add it to the response
-          yield response # yield the response:
+        temperature=0.1,
+        stream=True
+    ):
+        token = chunk.choices[0].delta.content
+        response += token
+        # Filter out any unwanted tokens like </s> or <|endoftext|>, [USER], [/ASS]
+        cleaned_response = response.replace("</s>", "").replace("<|endoftext|>", "").replace("[USER]", "").replace("[/ASS]", "").strip()
+        yield cleaned_response
 
 chatbot = gr.ChatInterface(respond)
 
-chatbot.launch()
+chatbot.launch()