import gradio as gr from gradio import ChatInterface, Request from gradio.helpers import special_args import anyio import os import threading import sys from itertools import chain import autogen from autogen.code_utils import extract_code from autogen import UserProxyAgent, AssistantAgent, Agent, OpenAIWrapper LOG_LEVEL = "INFO" TIMEOUT = 60 class myChatInterface(ChatInterface): async def _submit_fn( self, message: str, history_with_input: list[list[str | None]], request: Request, *args, ) -> tuple[list[list[str | None]], list[list[str | None]]]: history = history_with_input[:-1] inputs, _, _ = special_args( self.fn, inputs=[message, history, *args], request=request ) if self.is_async: response = await self.fn(*inputs) else: response = await anyio.to_thread.run_sync( self.fn, *inputs, limiter=self.limiter ) # history.append([message, response]) return history, history with gr.Blocks() as demo: def flatten_chain(list_of_lists): return list(chain.from_iterable(list_of_lists)) class thread_with_trace(threading.Thread): # https://www.geeksforgeeks.org/python-different-ways-to-kill-a-thread/ # https://stackoverflow.com/questions/6893968/how-to-get-the-return-value-from-a-thread def __init__(self, *args, **keywords): threading.Thread.__init__(self, *args, **keywords) self.killed = False self._return = None def start(self): self.__run_backup = self.run self.run = self.__run threading.Thread.start(self) def __run(self): sys.settrace(self.globaltrace) self.__run_backup() self.run = self.__run_backup def run(self): if self._target is not None: self._return = self._target(*self._args, **self._kwargs) def globaltrace(self, frame, event, arg): if event == "call": return self.localtrace else: return None def localtrace(self, frame, event, arg): if self.killed: if event == "line": raise SystemExit() return self.localtrace def kill(self): self.killed = True def join(self, timeout=0): threading.Thread.join(self, timeout) return self._return def update_agent_history(recipient, messages, sender, config): if config is None: config = recipient if messages is None: messages = recipient._oai_messages[sender] message = messages[-1] msg = message.get("content", "") # config.append(msg) if msg is not None else None # config can be agent_history return False, None # required to ensure the agent communication flow continues def _is_termination_msg(message): """Check if a message is a termination message. Terminate when no code block is detected. Currently only detect python code blocks. """ if isinstance(message, dict): message = message.get("content") if message is None: return False cb = extract_code(message) contain_code = False for c in cb: # todo: support more languages if c[0] == "python": contain_code = True break return not contain_code def initialize_agents(config_list): assistant = AssistantAgent( name="assistant", max_consecutive_auto_reply=10, llm_config={ # "seed": 42, "timeout": TIMEOUT, "config_list": config_list, }, ) userproxy = UserProxyAgent( name="userproxy", human_input_mode="NEVER", is_termination_msg=_is_termination_msg, max_consecutive_auto_reply=10, # code_execution_config=False, code_execution_config={ "work_dir": "coding", "use_docker": False, # set to True or image name like "python:3" to use docker }, ) # assistant.register_reply([Agent, None], update_agent_history) # userproxy.register_reply([Agent, None], update_agent_history) return assistant, userproxy def chat_to_oai_message(chat_history): """Convert chat history to OpenAI message format.""" messages = [] if LOG_LEVEL == "DEBUG": print(f"chat_to_oai_message: {chat_history}") for msg in chat_history: messages.append( { "content": msg[0].split()[0] if msg[0].startswith("exitcode") else msg[0], "role": "user", } ) messages.append({"content": msg[1], "role": "assistant"}) return messages def oai_message_to_chat(oai_messages, sender): """Convert OpenAI message format to chat history.""" chat_history = [] messages = oai_messages[sender] if LOG_LEVEL == "DEBUG": print(f"oai_message_to_chat: {messages}") for i in range(0, len(messages), 2): chat_history.append( [ messages[i]["content"], messages[i + 1]["content"] if i + 1 < len(messages) else "", ] ) return chat_history def agent_history_to_chat(agent_history): """Convert agent history to chat history.""" chat_history = [] for i in range(0, len(agent_history), 2): chat_history.append( [ agent_history[i], agent_history[i + 1] if i + 1 < len(agent_history) else None, ] ) return chat_history def initiate_chat(config_list, user_message, chat_history): if LOG_LEVEL == "DEBUG": print(f"chat_history_init: {chat_history}") # agent_history = flatten_chain(chat_history) if len(config_list[0].get("api_key", "")) < 2: chat_history.append( [ user_message, "Hi, nice to meet you! Please enter your API keys in below text boxs.", ] ) return chat_history else: llm_config = { # "seed": 42, "timeout": TIMEOUT, "config_list": config_list, } assistant.llm_config.update(llm_config) assistant.client = OpenAIWrapper(**assistant.llm_config) if user_message.strip().lower().startswith("show file:"): filename = user_message.strip().lower().replace("show file:", "").strip() filepath = os.path.join("coding", filename) if os.path.exists(filepath): chat_history.append([user_message, (filepath,)]) else: chat_history.append([user_message, f"File {filename} not found."]) return chat_history assistant.reset() oai_messages = chat_to_oai_message(chat_history) assistant._oai_system_message_origin = assistant._oai_system_message.copy() assistant._oai_system_message += oai_messages try: userproxy.initiate_chat(assistant, message=user_message) messages = userproxy.chat_messages chat_history += oai_message_to_chat(messages, assistant) # agent_history = flatten_chain(chat_history) except Exception as e: # agent_history += [user_message, str(e)] # chat_history[:] = agent_history_to_chat(agent_history) chat_history.append([user_message, str(e)]) assistant._oai_system_message = assistant._oai_system_message_origin.copy() if LOG_LEVEL == "DEBUG": print(f"chat_history: {chat_history}") # print(f"agent_history: {agent_history}") return chat_history def chatbot_reply_thread(input_text, chat_history, config_list): """Chat with the agent through terminal.""" thread = thread_with_trace( target=initiate_chat, args=(config_list, input_text, chat_history) ) thread.start() try: messages = thread.join(timeout=TIMEOUT) if thread.is_alive(): thread.kill() thread.join() messages = [ input_text, "Timeout Error: Please check your API keys and try again later.", ] except Exception as e: messages = [ [ input_text, str(e) if len(str(e)) > 0 else "Invalid Request to OpenAI, please check your API keys.", ] ] return messages def chatbot_reply_plain(input_text, chat_history, config_list): """Chat with the agent through terminal.""" try: messages = initiate_chat(config_list, input_text, chat_history) except Exception as e: messages = [ [ input_text, str(e) if len(str(e)) > 0 else "Invalid Request to OpenAI, please check your API keys.", ] ] return messages def chatbot_reply(input_text, chat_history, config_list): """Chat with the agent through terminal.""" return chatbot_reply_thread(input_text, chat_history, config_list) def get_description_text(): return """ # Microsoft AutoGen: Multi-Round Human Interaction Chatbot Demo This demo shows how to build a chatbot which can handle multi-round conversations with human interactions. #### [AutoGen](https://github.com/microsoft/autogen) [Discord](https://discord.gg/pAbnFJrkgZ) [Paper](https://arxiv.org/abs/2308.08155) [SourceCode](https://github.com/thinkall/autogen-demos) """ def update_config(): config_list = autogen.config_list_from_models( model_list=[os.environ.get("MODEL", "gpt-35-turbo")], ) if not config_list: config_list = [ { "api_key": "", "base_url": "", "api_type": "azure", "api_version": "2023-07-01-preview", "model": "gpt-35-turbo", } ] return config_list def set_params(model, oai_key, aoai_key, aoai_base): os.environ["MODEL"] = model os.environ["OPENAI_API_KEY"] = oai_key os.environ["AZURE_OPENAI_API_KEY"] = aoai_key os.environ["AZURE_OPENAI_API_BASE"] = aoai_base def respond(message, chat_history, model, oai_key, aoai_key, aoai_base): set_params(model, oai_key, aoai_key, aoai_base) config_list = update_config() chat_history[:] = chatbot_reply(message, chat_history, config_list) if LOG_LEVEL == "DEBUG": print(f"return chat_history: {chat_history}") return "" config_list, assistant, userproxy = ( [ { "api_key": "", "base_url": "", "api_type": "azure", "api_version": "2023-07-01-preview", "model": "gpt-35-turbo", } ], None, None, ) assistant, userproxy = initialize_agents(config_list) description = gr.Markdown(get_description_text()) with gr.Row() as params: txt_model = gr.Dropdown( label="Model", choices=[ "gpt-4", "gpt-35-turbo", "gpt-3.5-turbo", ], allow_custom_value=True, value="gpt-35-turbo", container=True, ) txt_oai_key = gr.Textbox( label="OpenAI API Key", placeholder="Enter OpenAI API Key", max_lines=1, show_label=True, container=True, type="password", ) txt_aoai_key = gr.Textbox( label="Azure OpenAI API Key", placeholder="Enter Azure OpenAI API Key", max_lines=1, show_label=True, container=True, type="password", ) txt_aoai_base_url = gr.Textbox( label="Azure OpenAI API Base", placeholder="Enter Azure OpenAI Base Url", max_lines=1, show_label=True, container=True, type="password", ) chatbot = gr.Chatbot( [], elem_id="chatbot", bubble_full_width=False, avatar_images=( "human.png", (os.path.join(os.path.dirname(__file__), "autogen.png")), ), render=False, height=600, ) txt_input = gr.Textbox( scale=4, show_label=False, placeholder="Enter text and press enter", container=False, render=False, autofocus=True, ) chatiface = myChatInterface( respond, chatbot=chatbot, textbox=txt_input, additional_inputs=[ txt_model, txt_oai_key, txt_aoai_key, txt_aoai_base_url, ], examples=[ ["write a python function to count the sum of two numbers?"], ["what if the production of two numbers?"], [ "Plot a chart of the last year's stock prices of Microsoft, Google and Apple and save to stock_price.png." ], ["show file: stock_price.png"], ], ) if __name__ == "__main__": demo.launch(share=True, server_name="0.0.0.0")