LangChain / RAG / Vector DBs · Leçon

Agents LangChain et concepts d’outils

Découvrez comment les agents LangChain peuvent raisonner et utiliser des outils pour accomplir des tâches complexes allant au-delà de la simple réponse aux questions.

Agents LangChain et concepts d’outils est une leçon LangChain / RAG / Vector DBs gratuite sur CoddyKit. Ceci est la leçon 1 sur 4. Tu peux lire la leçon complète ci-dessous gratuitement — puis la pratiquer en direct dans le navigateur avec un éditeur de code intégré et un tuteur IA 24/7. Elle fait partie du parcours d'apprentissage LangChain / RAG / Vector DBs, et ta progression se synchronise sur le web et l'application CoddyKit. Le cours LangChain / RAG / Vector DBs comprend 4 leçons au total.

Certaines parties de cette leçon n'ont pas encore été traduites et s'affichent en anglais.

What are LangChain Agents?

Welcome to Agents! So far, you've learned to build chains that perform a fixed sequence of steps. But what if you need more flexibility?

LangChain Agents are systems that allow an LLM to dynamically decide which actions to take, observe the results, and then decide the next action. Think of them as giving the LLM a 'brain' to reason and 'hands' to interact with the world.

The Agent's Reasoning Loop

Agents operate on a continuous Observe-Think-Act loop:

Observe: The agent receives an input (your query) and the results of its last action.
Think: The LLM reasons about the current situation and decides what to do next.
Act: The agent performs an action, often by using a tool.

This loop continues until the agent determines it has enough information to answer your question or complete its task.

Introducing Tools

An LLM alone can only access the knowledge it was trained on. To perform actions in the real world or access up-to-date information, it needs Tools.

Tools are functions or APIs that an agent can call. They extend the LLM's capabilities, allowing it to:

Search the internet for current events.
Perform calculations.
Query databases.
Interact with other applications.

Anatomy of a Tool

Each tool needs a few key pieces of information for the agent to use it effectively:

Name: A unique identifier (e.g., 'Google Search').
Description: A clear explanation of what the tool does and when it should be used. This is crucial for the LLM's reasoning.
Input Schema: What kind of input the tool expects (e.g., a search query string, two numbers).

The LLM reads the descriptions to decide which tool is appropriate for a given step.

Common Built-in Tools

LangChain provides many ready-to-use tools. Here are a couple of popular examples:

SerpAPIWrapper: Allows the agent to perform Google searches. Useful for current information or specific data points.
LLMMathChain: Enables the agent to perform mathematical calculations. This is more reliable than asking the LLM to do complex math directly.

These tools are like plugins that give your agent superpowers!

Setting Up a Simple Agent

To create an agent, you typically need:

An LLM (e.g., ChatOpenAI).
A list of Tools the agent can use.
An Agent Type or a prompt that defines how the agent should reason (e.g., ReAct framework).

Let's prepare these components to build an agent that can do math!

Agent in Action: Calculator

Here's a Python example of an agent using a custom 'Calculator' tool. The LLM decides when to use the tool based on the question.

Note: For a real application, replace eval() with a secure math parser.

import os
from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_react_agent, Tool
from langchain import hub

# Ensure OPENAI_API_KEY is set in your environment variables

def get_calculator_tool():
    """A simple calculator tool."""
    def calculate(expression: str) -> str:
        try:
            # DANGER: In real apps, use safer math parsers!
            return str(eval(expression))
        except Exception as e:
            return f"Error: {e}"

    return Tool(
        name="Calculator",
        func=calculate,
        description="Useful for when you need to answer questions about math. Input should be a mathematical expression."
    )

def main():
    llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
    tools = [get_calculator_tool()]
    prompt = hub.pull("hwchase17/react") # ReAct agent prompt

    agent = create_react_agent(llm, tools, prompt)
    agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

    print("--- Agent Running ---")
    result = agent_executor.invoke({"input": "What is 123 + 456?"})
    print("\n--- Agent Result ---")
    print(result["output"])

if __name__ == "__main__":
    main()

Tracing Agent's Thought Process

When you run an agent with verbose=True, LangChain shows you the agent's internal monologue:

Thought: The LLM's reasoning about the current situation.
Action: The tool it decides to use and its input.
Observation: The result returned by the tool.

This trace helps you understand how the agent arrived at its answer and debug its behavior.

Different Agent Types

LangChain supports various agent types, each with a specific reasoning strategy:

zero-shot-react-description: A general-purpose agent that uses the ReAct framework. It relies heavily on tool descriptions.
OpenAIFunctionsAgent: Leverages OpenAI's native function calling capabilities, often leading to more robust and concise tool usage.
Others exist for specific use cases or models.

Choosing the right agent type depends on your LLM and task complexity.

Agent Limitations & Considerations

While powerful, agents have limitations:

Cost & Latency: Multiple LLM calls and tool invocations can increase cost and response time.
Reliability: Agents can still 'hallucinate' or misuse tools if descriptions aren't precise or the LLM struggles with complex reasoning.
Security: Tools that interact with external systems (like eval() or APIs) need careful handling to prevent vulnerabilities.

Design your tools and prompts carefully!

Agent Concepts Check

Which of the following statements about LangChain Agents and Tools are TRUE?

Recap: The Power of Agents

You've learned that LangChain Agents empower LLMs to move beyond simple question-answering by enabling them to:

Reason dynamically through an 'Observe-Think-Act' loop.
Utilize Tools to interact with external data and services.
Perform complex tasks that require multiple steps and external interactions.

This combination makes agents incredibly versatile for building intelligent applications!

Questions Fréquemment Posées

La leçon « Agents LangChain et concepts d’outils » est-elle gratuite ?

Oui — le texte complet de « Agents LangChain et concepts d’outils » est gratuit à lire ici sur le web. Pour la pratiquer de manière interactive (un éditeur de code intégré et un tuteur IA 24/7) et déverrouiller le reste du cours LangChain / RAG / Vector DBs, passe à CoddyKit PRO. Le cours LangChain / RAG / Vector DBs comprend 4 leçons au total.

Qu'est-ce que j'apprendrai dans « Agents LangChain et concepts d’outils » ?

Découvrez comment les agents LangChain peuvent raisonner et utiliser des outils pour accomplir des tâches complexes allant au-delà de la simple réponse aux questions. Tu pratiques LangChain / RAG / Vector DBs avec du code pratique que tu exécutes directement dans le navigateur, et un tuteur IA 24/7 répond à tes questions au fur et à mesure que tu avances dans la leçon.

Dois-je avoir de l'expérience pour commencer LangChain / RAG / Vector DBs ?

Aucune expérience préalable n'est requise. LangChain / RAG / Vector DBs sur CoddyKit est structuré pour les débutants jusqu'aux apprenants avancés, donc tu peux commencer ici ou depuis le début et avancer à ton rythme. Ceci est la leçon 1 sur 4.

Combien de temps prend la leçon « Agents LangChain et concepts d’outils » ?

La plupart des leçons CoddyKit prennent environ 5–10 minutes. Chacune est courte et interactive, tu progresses régulièrement et tu repiques exactement où tu t'es arrêté sur le web et l'app.

Peux-tu écrire et exécuter du code dans cette leçon LangChain / RAG / Vector DBs ?

Oui. Chaque leçon LangChain / RAG / Vector DBs inclut un éditeur de code intégré, tu écris et exécutes du vrai code directement dans ton navigateur et tu reçois des retours IA instantanés — aucune configuration locale requise.

Toutes les leçons de ce cours

Agents LangChain et concepts d’outils
Créer des flux de travail RAG multi-agents
Intégrer des API externes comme outils
Mémoire et état dans RAG agentique

← Retour à LangChain / RAG / Vector DBs