Interrogation et génération de réponses
Développez la logique nécessaire pour traiter les requêtes utilisateur, récupérer le contexte pertinent et synthétiser des réponses à l’aide du LLM.
Interrogation et génération de réponses est une leçon LangChain / RAG / Vector DBs gratuite sur CoddyKit. Ceci est la leçon 2 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.
Querying RAG: The Answer Flow
After integrating RAG components, the next step is to use them to answer user questions. This lesson covers the full process from a user's query to a generated answer.
We'll focus on the 'query-time' logic: how your system takes a question, finds relevant context, and synthesizes a coherent response using an LLM.
Understanding the User Query
A RAG system starts with a user's question, just like a search engine. This raw input is the trigger for the entire process.
- It defines what information needs to be retrieved.
- It guides the LLM on what kind of answer to generate.
No special formatting is typically needed at this initial stage; it's just plain text.
Setting Up Your Retriever
To find relevant documents, you need a retriever. This component knows how to query your vector store. You typically obtain it from your VectorStore instance.
The as_retriever() method creates this component, and you can configure parameters like k (number of top documents to fetch).
from langchain_community.vectorstores import InMemoryVectorStore
from langchain_core.documents import Document
from langchain_core.embeddings import Embeddings
from typing import List
# A simple mock for embeddings
class MockEmbeddings(Embeddings):
def embed_documents(self, texts: List[str]) -> List[List[float]]:
return [[i * 0.1] * 10 for i in range(len(texts))]
def embed_query(self, text: str) -> List[float]:
return [0.5] * 10
def main():
# Create a dummy vector store with some content
embeddings = MockEmbeddings()
docs = [
Document(page_content="The capital of France is Paris.", metadata={"source": "wiki"}),
Document(page_content="Eiffel Tower is in Paris, France.", metadata={"source": "travel"})
]
vectorstore = InMemoryVectorStore.from_documents(docs, embeddings)
# Create a retriever from the vector store
retriever = vectorstore.as_retriever(search_kwargs={"k": 1})
print("Retriever created successfully!")
if __name__ == "__main__":
main()Fetching Contextual Documents
Once you have a retriever, you can invoke it with the user's query. It will perform a similarity search in your vector store and return the most relevant Document objects.
These documents form the context that will be passed to the LLM.
from langchain_community.vectorstores import InMemoryVectorStore
from langchain_core.documents import Document
from langchain_core.embeddings import Embeddings
from typing import List
class MockEmbeddings(Embeddings):
def embed_documents(self, texts: List[str]) -> List[List[float]]:
return [[i * 0.1] * 10 for i in range(len(texts))]
def embed_query(self, text: str) -> List[float]:
return [0.5] * 10
def main():
embeddings = MockEmbeddings()
docs = [
Document(page_content="The capital of France is Paris.", metadata={"source": "wiki"}),
Document(page_content="Eiffel Tower is in Paris, France.", metadata={"source": "travel"}),
Document(page_content="London is the capital of the UK.", metadata={"source": "wiki"})
]
vectorstore = InMemoryVectorStore.from_documents(docs, embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 2})
user_query = "What is the capital of France?"
retrieved_docs = retriever.invoke(user_query)
print(f"Retrieved {len(retrieved_docs)} documents:")
for doc in retrieved_docs:
print(f"- {doc.page_content[:50]}...")
if __name__ == "__main__":
main()Preparing Context for the LLM
LLMs usually prefer a single string of text as context. The retrieved Document objects need to be combined into a coherent format.
A common approach is to concatenate their page_content fields, perhaps with separators, and include source metadata if desired.
- Ensures all context fits within the LLM's token window.
- Presents a clean input for the LLM to reason over.
Crafting the RAG Prompt
The prompt template is crucial. It instructs the LLM on how to use the provided context to answer the user's question. It typically includes placeholders for both the context and the question.
A well-designed prompt guides the LLM to be factual and avoid hallucination.
from langchain_core.prompts import ChatPromptTemplate
def main():
# Define a RAG-specific prompt template
rag_prompt = ChatPromptTemplate.from_messages([
("system", "You are an AI assistant for Q&A. Use the context to answer. If you don't know, say that you don't know."),
("human", "Context: {context}\nQuestion: {question}")
])
print("RAG Prompt Template created!")
# Example of how it formats:
# print(rag_prompt.format(context="some info", question="a query"))
if __name__ == "__main__":
main()Connecting the Generation Engine
The final step in generating an answer is to pass the prepared context and the user's question to a Large Language Model. LangChain allows you to easily plug in various LLM providers.
For this example, we'll use a mock LLM to demonstrate the integration without needing an API key.
from langchain_core.language_models import BaseChatModel
from langchain_core.messages import BaseMessage, AIMessage
from typing import List, Any
# A simple mock LLM
class MockChatLLM(BaseChatModel):
def invoke(self, input: Any, config: Any = None) -> BaseMessage:
# Simulate LLM response based on input
if "Paris" in str(input):
return AIMessage(content="Paris is the capital of France.")
elif "London" in str(input):
return AIMessage(content="London is the capital of the UK.")
else:
return AIMessage(content="I don't have enough info to answer.")
async def ainvoke(self, input: Any, config: Any = None) -> BaseMessage:
return self.invoke(input, config) # Simple async pass-through
@property
def _llm_type(self) -> str:
return "mock-chat-llm"
def main():
llm = MockChatLLM()
print("Mock LLM initialized!")
# Example invocation (not part of the RAG chain yet)
response = llm.invoke("Tell me about Paris.")
print(f"LLM Response: {response.content}")
if __name__ == "__main__":
main()Assembling the End-to-End RAG Chain
Now, we combine the retriever, prompt template, and LLM using LangChain Expression Language (LCEL) to create a powerful, flexible RAG chain. This chain handles the entire flow.
We'll use RunnablePassthrough to manage inputs and StrOutputParser to extract the final text answer.
from langchain_core.runnables import RunnablePassthrough, RunnableLambda
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.documents import Document
from langchain_community.vectorstores import InMemoryVectorStore
from langchain_core.embeddings import Embeddings
from langchain_core.language_models import BaseChatModel
from langchain_core.messages import BaseMessage, AIMessage
from typing import List, Any
# Mock Embeddings
class MockEmbeddings(Embeddings):
def embed_documents(self, texts: List[str]) -> List[List[float]]:
return [[i * 0.1] * 10 for i in range(len(texts))]
def embed_query(self, text: str) -> List[float]:
return [0.5] * 10
# Mock LLM
class MockChatLLM(BaseChatModel):
def invoke(self, input: Any, config: Any = None) -> BaseMessage:
input_str = str(input)
if "Paris" in input_str and "capital of France" in input_str:
return AIMessage(content="Based on context, Paris is the capital of France.")
elif "Eiffel Tower" in input_str and "Paris" in input_str:
return AIMessage(content="The Eiffel Tower is in Paris, France.")
else:
return AIMessage(content="I don't have enough info in the context.")
async def ainvoke(self, input: Any, config: Any = None) -> BaseMessage:
return self.invoke(input, config)
@property
def _llm_type(self) -> str:
return "mock-chat-llm"
def main():
# 1. Setup Retriever
embeddings = MockEmbeddings()
docs = [
Document(page_content="The capital of France is Paris.", metadata={"source": "wiki"}),
Document(page_content="Eiffel Tower is in Paris, France.", metadata={"source": "travel"})
]
vectorstore = InMemoryVectorStore.from_documents(docs, embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 1})
# 2. Setup Prompt
rag_prompt = ChatPromptTemplate.from_messages([
("system", "You are an AI assistant. Use the following context to answer: {context}. If you don't know, say 'I don't know.'"),
("human", "Question: {question}")
])
# 3. Setup LLM
llm = MockChatLLM()
# 4. Define how to format retrieved documents
def format_docs(docs: List[Document]) -> str:
return "\n\n".join(doc.page_content for doc in docs)
# 5. Build the RAG chain
rag_chain = (
{"context": retriever | RunnableLambda(format_docs),
"question": RunnablePassthrough()}
| rag_prompt
| llm
| StrOutputParser()
)
print("RAG chain assembled!")
if __name__ == "__main__":
main()Querying Your RAG Application
With the RAG chain fully constructed, you can now invoke it with a user's question. The chain will internally handle retrieval, context formatting, prompting, and LLM generation, returning a direct answer.
This is the final step in getting a response from your RAG system.
from langchain_core.runnables import RunnablePassthrough, RunnableLambda
from langchain_core.output_parsers import StrOutputParser
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.documents import Document
from langchain_community.vectorstores import InMemoryVectorStore
from langchain_core.embeddings import Embeddings
from langchain_core.language_models import BaseChatModel
from langchain_core.messages import BaseMessage, AIMessage
from typing import List, Any
# Mock Embeddings
class MockEmbeddings(Embeddings):
def embed_documents(self, texts: List[str]) -> List[List[float]]:
return [[i * 0.1] * 10 for i in range(len(texts))]
def embed_query(self, text: str) -> List[float]:
return [0.5] * 10
# Mock LLM
class MockChatLLM(BaseChatModel):
def invoke(self, input: Any, config: Any = None) -> BaseMessage:
input_str = str(input)
if "Paris" in input_str and "capital of France" in input_str:
return AIMessage(content="Based on context, Paris is the capital of France.")
elif "Eiffel Tower" in input_str and "Paris" in input_str:
return AIMessage(content="The Eiffel Tower is in Paris, France.")
else:
return AIMessage(content="I don't have enough info in the context.")
async def ainvoke(self, input: Any, config: Any = None) -> BaseMessage:
return self.invoke(input, config)
@property
def _llm_type(self) -> str:
return "mock-chat-llm"
def main():
# Setup Retriever
embeddings = MockEmbeddings()
docs = [
Document(page_content="The capital of France is Paris.", metadata={"source": "wiki"}),
Document(page_content="Eiffel Tower is in Paris, France.", metadata={"source": "travel"})
]
vectorstore = InMemoryVectorStore.from_documents(docs, embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 1})
# Setup Prompt
rag_prompt = ChatPromptTemplate.from_messages([
("system", "You are an AI assistant. Use the following context to answer: {context}. If you don't know, say 'I don't know.'"),
("human", "Question: {question}")
])
# Setup LLM
llm = MockChatLLM()
def format_docs(docs: List[Document]) -> str:
return "\n\n".join(doc.page_content for doc in docs)
# Build the RAG chain
rag_chain = (
{"context": retriever | RunnableLambda(format_docs),
"question": RunnablePassthrough()}
| rag_prompt
| llm
| StrOutputParser()
)
# Invoke the RAG chain with a query
query = "Where is the Eiffel Tower?"
result = rag_chain.invoke(query)
print(f"Query: {query}")
print(f"Answer: {result}")
query_no_context = "What is the capital of Japan?"
result_no_context = rag_chain.invoke(query_no_context)
print(f"\nQuery: {query_no_context}")
print(f"Answer: {result_no_context}")
if __name__ == "__main__":
main()Test Your RAG Flow Knowledge
Consider a LangChain RAG application designed to answer questions from a knowledge base.
Recap: Querying RAG
In this lesson, you learned how to bring all the RAG components together to process user queries and generate answers:
- We transformed a user's question into a query for the retriever.
- We instantiated a retriever from a vector store to fetch relevant documents.
- We crafted a prompt template to guide the LLM.
- We integrated an LLM to synthesize the final answer.
- Finally, we assembled and invoked an end-to-end RAG chain using LangChain Expression Language (LCEL).
You can now build a functional RAG system that delivers grounded, factual answers!
Apprends LangChain / RAG / Vector DBs avec un tuteur IA — gratuit
Écris et exécute du vrai code dans ton navigateur, obtiens de l'aide instantanée d'un tuteur IA disponible 24h/24, et reprends là où tu t'es arrêté sur le web ou dans l'app.
- Cours
- 12
- Leçons
- 48
Questions Fréquemment Posées
La leçon « Interrogation et génération de réponses » est-elle gratuite ?
Oui — le texte complet de « Interrogation et génération de réponses » 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 « Interrogation et génération de réponses » ?
Développez la logique nécessaire pour traiter les requêtes utilisateur, récupérer le contexte pertinent et synthétiser des réponses à l’aide du LLM. 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 2 sur 4.
Combien de temps prend la leçon « Interrogation et génération de réponses » ?
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
- Intégration de tous les composants RAG
- Interrogation et génération de réponses
- Évaluation des performances d’un système RAG
- Créer un jeu de tests de référence pour RAG