Practical Project: Personal Chef Agent
Web Search + Memory + System Prompt Comprehensive Application

1Project goal: Integrate all capabilities of Module 1

Personal Chef Agent is a comprehensive practical project of Module 1, which integrates the core capabilities of all previous courses:

2Complete code: a combination of 5 key components

from dotenv import load_dotenv
load_dotenv()

import os
from langchain.chat_models import init_chat_model
from langchain.agents import create_agent
from langchain.messages import HumanMessage
from langchain.tools import tool
from langgraph.checkpoint.memory import InMemorySaver
from tavily import TavilyClient

# ────────────────────────────────────
# 组件 1：模型工厂函数
# ────────────────────────────────────
def deepseek_model(**kwargs):
    return init_chat_model(
        model="deepseek-chat",
        model_provider="openai",
        api_key=os.environ["DEEPSEEK_API_KEY"],
        base_url="https://api.deepseek.com",
        max_tokens=1000,
        **kwargs
    )

# ────────────────────────────────────
# 组件 2：Web 搜索工具
# ────────────────────────────────────
tavily_client = TavilyClient()

@tool
def web_search(query: str) -> dict:
    """Search the web for information"""
    return tavily_client.search(query)

# ────────────────────────────────────
# 组件 3：System Prompt（厨师角色定义）
# ────────────────────────────────────
system_prompt = """

You are a personal chef. The user will give you a list of ingredients they have left over in their house.

Using the web search tool, search the web for recipes that can be made with the ingredients they have.

Return recipe suggestions and eventually the recipe instructions to the user, if requested.

"""

# ────────────────────────────────────
# 组件 4 + 5：Agent = 模型 + 工具 + System Prompt + 记忆
# ────────────────────────────────────
agent = create_agent(
    model=deepseek_model(),
    tools=[web_search],
    system_prompt=system_prompt,
    checkpointer=InMemorySaver()
)

3System Prompt: Define the behavior of the Chef Agent

System Prompt is the "soul" of this Agent - it defines the Agent's role, capability boundaries and response methods:

system_prompt = """

You are a personal chef. The user will give you a list of ingredients they have left over in their house.

Using the web search tool, search the web for recipes that can be made with the ingredients they have.

Return recipe suggestions and eventually the recipe instructions to the user, if requested.

"""

This prompt does three things:

1. Define roles: "You are a personal chef" - the model knows who it is and its behavioral style will be different
2. Define tool usage strategies："Using the web search tool, search the web..."——Explicit instructionsModels use search tools instead of making up recipes based on LLM knowledge
3. Define response format："Return recipe suggestions and eventually the recipe instructions..."——Give suggestions first, and then give detailed steps after the user requests them.

4web_search tool: Search recipes in real time

Here'sweb_searchThe tool is slightly different than the previous one - using@toolSimplified writing method (no name specified, function name automatically used), and description comes from docstring:

@tool                    # 简化写法：name 自动为 "web_search"
def web_search(query: str) -> dict:
    """Search the web for information"""   # docstring 作为 description
    return tavily_client.search(query)

5Multi-turn dialogue demonstration: remember the ingredients and continue the dialogue

config = {"configurable": {"thread_id": "1"}}

# 轮次 1：告诉 Agent 有什么食材
response = agent.invoke(
    {"messages": [HumanMessage(content="I have some leftover chicken and rice. What can I make?")]},
    config
)
print(response['messages'][-1].content)

Agent calledweb_search, searches for "chicken and rice recipes", gives 3 suggestions based on the search results, and invites the user to choose a recipe with detailed steps.

Round 2: Ask how to do it (use memory)

# 轮次 2：用户要鸡肉炒饭的做法（Agent 记住了上一轮的食材）
response2 = agent.invoke(
    {"messages": [HumanMessage(content="Can you give me the detailed recipe for chicken fried rice?")]},
    config
)
print(response2['messages'][-1].content)

6In-depth analysis of Agent execution process

When the user says "I have some leftover chicken and rice. What can I make?", the complete execution process inside the Agent:

• 1

  receive user input

  HumanMessage adds to the message list. checkpointer loads the history of the current thread_id (the first round is empty).

• 2

  LLM inference (model node)

  The model sees system prompt (chef role) + historical messages + user messages. It reasoned: "The user has chicken and rice, I should use the web_search tool to find the recipe." Output a tool_call.

• 3

  Tool execution (tools node)

  LangGraph executionweb_search("chicken and rice recipes"), obtain Tavily search results, encapsulate them into ToolMessage and add them to the message history.

• 4

  LLM re-inference (model node)

  The model sees the search results and generates recipe suggestions based on real data. Output the final AIMessage (text answer).

• 5

  Status saving (checkpointer)

  InMemorySaver saves the complete message history of this execution (including HumanMessage + AIMessage + ToolMessage). It will be loaded automatically the next time it is called with the same thread_id.

7Module 1 Knowledge Panorama Review and Next Step

At this point, all seven Lessons of Module 1 are completed. The following is a complete knowledge panorama: