Update app.py

app.py

@@ -3,6 +3,7 @@ from datasets import Dataset
 from sentence_transformers import SentenceTransformer
 import faiss
 import time
+import json
 #import torch
 import pandas as pd
 from llama_cpp import Llama
@@ -152,10 +153,9 @@ def talk(prompt, history):
 # indicates the end of a sequence
     text = ""
     stream = model.create_chat_completion(messages, max_tokens=1000, stop=["</s>"], stream=True)
-    print(stream["choices"][0]["text"])
-#    for output in stream:
-#        text += output["choices"][0]["text"]
-#        yield text
+    for output in stream:
+        text += output["choices"][0]["text"]
+        yield text
 
 #    model_input = model.create_chat_completion(messages = messages)