Update to_md function for illustrating code blocks

app.py

@@ -1,3 +1,69 @@
+import openai
+import tiktoken
+
+import datetime
+import json
+import os
+
+openai.api_key = os.getenv('API_KEY')
+openai.request_times = 0
+
+def ask(question, history, behavior):
+    openai.request_times += 1
+    print(f"request times {openai.request_times}: {datetime.datetime.now()}: {question}")
+    try:
+        response = openai.ChatCompletion.create(
+            model="gpt-3.5-turbo",
+            messages=forget_long_term(
+                [
+                    {"role":"system", "content":content}
+                    for content in behavior
+                ] + [
+                    {"role":"user" if i%2==0 else "assistant", "content":content}
+                    for i,content in enumerate(history + [question])
+                ]
+            )
+        )["choices"][0]["message"]["content"]
+        while response.startswith("\n"):
+            response = response[1:]
+    except Exception as e:
+        print(e)
+        response = 'Timeout! Please wait a few minutes and retry'
+    history = history + [question, response]
+    return history
+
+def num_tokens_from_messages(messages, model="gpt-3.5-turbo"):
+    """Returns the number of tokens used by a list of messages."""
+    try:
+        encoding = tiktoken.encoding_for_model(model)
+    except KeyError:
+        encoding = tiktoken.get_encoding("cl100k_base")
+    if model == "gpt-3.5-turbo":  # note: future models may deviate from this
+        num_tokens = 0
+        for message in messages:
+            num_tokens += 4  # every message follows <im_start>{role/name}\n{content}<im_end>\n
+            for key, value in message.items():
+                num_tokens += len(encoding.encode(value))
+                if key == "name":  # if there's a name, the role is omitted
+                    num_tokens += -1  # role is always required and always 1 token
+        num_tokens += 2  # every reply is primed with <im_start>assistant
+        return num_tokens
+    else:
+        raise NotImplementedError(f"""num_tokens_from_messages() is not presently implemented for model {model}.
+See https://github.com/openai/openai-python/blob/main/chatml.md for information on how messages are converted to tokens.""")
+
+def forget_long_term(messages, max_num_tokens=4000):
+    while num_tokens_from_messages(messages)>max_num_tokens:
+        if messages[0]["role"]=="system" and not len(messages[0]["content"]>=max_num_tokens):
+            messages = messages[:1] + messages[2:]
+        else:
+            messages = messages[1:]
+    return messages
+
+
+import gradio as gr
+
+
 def to_md(content):
     is_inside_code_block = False
     output_spans = []
@@ -35,4 +101,65 @@ def to_md(content):
                         output_spans.append("\n")
         else:
             output_spans.append(content[i])
-    return "".join(output_spans)
+    return "".join(output_spans)
+
+
+def predict(question, history=[], behavior=[]):
+    history = ask(question, history, behavior)
+    response = [(to_md(history[i]),to_md(history[i+1])) for i in range(0,len(history)-1,2)]
+    return "", history, response
+
+
+with gr.Blocks() as demo:
+    
+    examples_txt = [
+        ['帮我写一个python脚本实现快排'],
+        ['如何用numpy提取数组的分位数？'],
+        ['how to match the code block in markdown such like ```def foo():\n    pass``` through regex in python?'],
+        ['how to load a pre-trained language model and generate sentences?'],
+    ]
+    
+    examples_bhv = [
+        f"You are a helpful assistant. You will answer all the questions step-by-step.",
+        f"You are a helpful assistant. Today is {datetime.date.today()}.",
+    ]
+    
+    gr.Markdown(
+        """
+        朋友你好，
+        
+        这是我利用[gradio](https://gradio.app/creating-a-chatbot/)编写的一个小网页，用于以网页的形式给大家分享ChatGPT请求服务，希望你玩的开心。关于使用技巧或学术研讨，欢迎在[Community](https://huggingface.co/spaces/zhangjf/chatbot/discussions)中和我交流。
+        
+        这一版相比于原版的[chatbot](https://huggingface.co/spaces/zhangjf/chatbot)，用了较低版本的gradio==3.16.2，因而能更好地展示markdown中的源代码
+        
+        p.s. 响应时间和聊天内容长度正相关，一般能在5秒~30秒内响应。
+        """)
+    
+    behavior = gr.State([])
+    
+    with gr.Column(variant="panel"):
+        with gr.Row().style(equal_height=True):
+            with gr.Column(scale=0.85):
+                bhv = gr.Textbox(show_label=False, placeholder="输入你想让ChatGPT扮演的人设").style(container=False)
+            with gr.Column(scale=0.15, min_width=0):
+                button_set = gr.Button("Set")
+    bhv.submit(fn=lambda x:(x,[x]), inputs=[bhv], outputs=[bhv, behavior])
+    button_set.click(fn=lambda x:(x,[x]), inputs=[bhv], outputs=[bhv, behavior])
+    
+
+    state = gr.State([])
+    
+    with gr.Column(variant="panel"):
+        chatbot = gr.Chatbot()
+        txt = gr.Textbox(show_label=False, placeholder="输入你想让ChatGPT回答的问题").style(container=False)
+        with gr.Row():
+            button_gen = gr.Button("Submit")
+            button_clr = gr.Button("Clear")
+        
+    gr.Examples(examples=examples_bhv, inputs=bhv, label="Examples for setting behavior")
+    gr.Examples(examples=examples_txt, inputs=txt, label="Examples for asking question")
+    txt.submit(predict, [txt, state, behavior], [txt, state, chatbot])
+    button_gen.click(fn=predict, inputs=[txt, state, behavior], outputs=[txt, state, chatbot])
+    button_clr.click(fn=lambda :([],[]), inputs=None, outputs=[chatbot, state])
+
+demo.launch()