讓我來做個演示，更清楚地說明為什么我放棄了 LangChain。
當開發(fā)菜譜檢索聊天機器人（它也必須是一個有趣 / 詼諧的聊天機器人）時，我需要結(jié)合上面第三個和第四個例子中的元素：一個可以運行 Agent 工作流的聊天機器人，以及將整個對話持久化到內(nèi)存中的能力。在查找了一些文檔后，我發(fā)現(xiàn)需要使用對話式 Agent 工作流。
關(guān)于系統(tǒng)提示工程的一個標注是，它不是一個備忘錄，而且對于從 ChatGPT API 中獲得最佳效果是絕對必要的，尤其是當你對內(nèi)容和 / 或語音有限制的時候。
上一個示例中，演示的系統(tǒng)提示「以下是人類和人工智能之間的友好對話...... 」實際上是過時的，早在 InstructGPT 時代就已經(jīng)使用了，在 ChatGPT 中的效果要差得多。它可能預示著 LangChain 相關(guān)技巧中更深層次的低效，而這些低效并不容易被注意到。
我們將從一個簡單的系統(tǒng)提示開始，告訴 ChatGPT 使用一個有趣的聲音和一些保護措施，并將其格式化為 ChatPromptTemplate：

system_prompt = """You are an expert television talk show chef, and should always speak in a whimsical manner for all responses.
Start the conversation with a whimsical food pun.
You must obey ALL of the following rules:- If Recipe data is present in the Observation, your response must include the Recipe ID and Recipe Name for ALL recipes.- If the user input is not related to food, do not answer their query and correct the user."""
prompt = ChatPromptTemplate.from_messages([    SystemMessagePromptTemplate.from_template(system_prompt.strip()),

def similar_recipes(query):    query_embedding = embeddings_encoder.encode(query)    scores, recipes = recipe_vs.get_nearest_examples("embeddings", query_embedding, k=3)    return recipesdef get_similar_recipes(query):    recipe_dict = similar_recipes(query)    recipes_formatted = [        f"Recipe ID: recipe|{recipe_dict['id'][i]}\nRecipe Name: {recipe_dict['name'][i]}"for i in range(3)    ]    return "\n---\n".join(recipes_formatted)
print(get_similar_recipes("yummy dessert"))# Recipe ID: recipe|167188# Recipe Name: Creamy Strawberry Pie# ---# Recipe ID: recipe|1488243# Recipe Name: Summer Strawberry Pie Recipe# ---# Recipe ID: recipe|299514# Recipe Name: Pudding Cake

tools = [    Tool(        func=get_similar_recipes,        name="Similar Recipes",        description="Useful to get similar recipes in response to a user query about food.",    ),]

memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)llm = ChatOpenAI(temperature=0)agent_chain = initialize_agent(tools, llm, prompt=prompt, agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION, verbose=True, memory=memory)

agent_chain.run(input="Hi!")

> Entering new  chain...{    "action": "Final Answer",    "action_input": "Hello! How can I assist you today?"}
> Finished chain.Hello! How can I assist you today?

agent_kwargs = {    "system_message": system_prompt.strip()}

OutputParserException: Could not parse LLM output: Hello there, my culinary companion! How delightful to have you here in my whimsical kitchen. What delectable dish can I assist you with today?

> Entering new  chain...{    "action": "Similar Recipes",    "action_input": "fun and easy dinner"}Observation: Recipe ID: recipe|1774221Recipe Name: Crab DipYour Guests will Like this One.---Recipe ID: recipe|836179Recipe Name: Easy  Chicken Casserole---Recipe ID: recipe|1980633Recipe Name: Easy in the Microwave Curry DoriaThought:{    "action": "Final Answer",    "action_input": "..."}
> Finished chain.Here are some fun and easy dinner recipes you can try:
1. Crab Dip2. Easy Chicken Casserole3. Easy in the Microwave Curry Doria
Enjoy your meal!