Almacenes vectoriales para recuperación
Aprenda a utilizar bases de datos vectoriales para almacenar y recuperar eficazmente fragmentos de documentos relevantes según su similitud semántica para RAG (Retrieval Augmented Generation).
Almacenes vectoriales para recuperación es una lección gratuita de AI Agents with LangChain & Autonomous Workflows en CoddyKit. Esta es la lección 3 de 4. Puedes leer la lección completa abajo gratuitamente — luego la practicas en el navegador con un editor de código integrado y un tutor de IA 24/7. Forma parte de la ruta de aprendizaje de AI Agents with LangChain & Autonomous Workflows, y tu progreso se sincroniza en la web y la app de CoddyKit. El curso de AI Agents with LangChain & Autonomous Workflows incluye 4 lecciones en total.
Partes de esta lección aún no han sido traducidas y se muestran en inglés.
Intro to Vector Stores
Welcome to the final lesson on Data Loading & Retrieval! Today, we'll dive into Vector Stores, a crucial component for building intelligent AI agents.
Think of vector stores as specialized databases designed to store and efficiently search through numerical representations of information, called embeddings.
Why Vector Stores for RAG?
Vector stores are the backbone of Retrieval Augmented Generation (RAG). RAG allows Large Language Models (LLMs) to access external, up-to-date information, overcoming their inherent limitations like:
- Knowledge cutoffs: LLMs only know what they were trained on.
- Hallucinations: Making up facts when uncertain.
Vector stores provide the relevant context for the LLM to generate accurate responses.
How Vector Stores Work
When you have text documents (which you learned to load and split in previous lessons), they are first converted into numerical embeddings.
These embeddings are then stored in a vector store. When a query comes in, it's also converted into an embedding. The vector store then finds document embeddings that are 'closest' (most similar) to the query embedding.
LangChain's VectorStore Abstraction
LangChain provides a powerful abstraction for interacting with various vector stores. This means you can swap out different vector database providers (like Chroma, Pinecone, FAISS) with minimal code changes.
Key methods include from_documents() to create a store from documents, add_documents() to add more, and similarity_search() to find relevant content.
Local Store: ChromaDB
For our examples, we'll use ChromaDB. It's an open-source, lightweight vector database that can run locally, making it perfect for development and testing without needing cloud services.
First, make sure you have the necessary packages installed:
pip install chromadbpip install langchain-communitypip install sentence-transformers
Initializing Chroma & Embeddings
Let's set up ChromaDB with a local embedding model. The embedding model converts text into vectors.
This example uses SentenceTransformerEmbeddings, which runs directly on your machine.
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import SentenceTransformerEmbeddings
import os
def main():
# Define a path for ChromaDB to store data locally
# This creates a 'chroma_db' folder if it doesn't exist
persist_directory = "./chroma_db"
# Initialize a local embedding function
# 'all-MiniLM-L6-v2' is a small, efficient model
embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
# Initialize ChromaDB. It will load if exists, or create new.
vectordb = Chroma(
persist_directory=persist_directory,
embedding_function=embeddings
)
print("ChromaDB initialized successfully!")
print(f"Database will persist at: {os.path.abspath(persist_directory)}")
if __name__ == "__main__":
main()Adding Documents to Chroma
Once initialized, we can add Document objects (which contain page_content and optional metadata) to our vector store. LangChain handles the embedding process automatically.
We'll add a few sample documents to our ChromaDB instance.
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import SentenceTransformerEmbeddings
from langchain_core.documents import Document
import os
def main():
persist_directory = "./chroma_db"
embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
# Load the existing ChromaDB or create a new one
vectordb = Chroma(
persist_directory=persist_directory,
embedding_function=embeddings
)
# Example documents to add
documents = [
Document(page_content="The quick brown fox jumps over the lazy dog."),
Document(page_content="Artificial intelligence is transforming industries."),
Document(page_content="Machine learning is a subset of AI."),
Document(page_content="Dogs are known for their loyalty and companionship.")
]
print(f"Adding {len(documents)} documents to ChromaDB...")
# add_documents handles embedding and storing
vectordb.add_documents(documents)
print("Documents added.")
if __name__ == "__main__":
main()Performing Similarity Search
Now that documents are stored, we can query the vector store to find content semantically similar to our query. The similarity_search() method returns a list of relevant Document objects.
The k parameter specifies how many top similar documents to retrieve.
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import SentenceTransformerEmbeddings
import os
def main():
persist_directory = "./chroma_db"
embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
# Load the existing ChromaDB
vectordb = Chroma(
persist_directory=persist_directory,
embedding_function=embeddings
)
query = "What is AI?"
print(f"Searching for documents similar to: '{query}'")
# Perform similarity search, retrieve top 2 results
docs = vectordb.similarity_search(query, k=2)
print("\nFound relevant documents:")
for i, doc in enumerate(docs):
print(f"--- Document {i+1} ---")
print(doc.page_content)
if __name__ == "__main__":
main()VectorStore as a Retriever
In LangChain, a Retriever is an interface that returns Documents given an unstructured query. A vector store is one of the most common ways to create a retriever.
By converting your vector store into a retriever, you can easily plug it into more complex LangChain chains and agents, especially for RAG applications.
Activating the Retriever
Here's how to convert your ChromaDB instance into a retriever and use it to fetch documents based on a query. This is the final step before integrating it into a full RAG chain.
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import SentenceTransformerEmbeddings
import os
def main():
persist_directory = "./chroma_db"
embeddings = SentenceTransformerEmbeddings(model_name="all-MiniLM-L6-v2")
# Load the existing ChromaDB
vectordb = Chroma(
persist_directory=persist_directory,
embedding_function=embeddings
)
# Convert the vector store into a retriever
# search_kwargs allows passing arguments like 'k' to the underlying search
retriever = vectordb.as_retriever(search_kwargs={"k": 2})
query = "Tell me about AI."
print(f"Using retriever to find documents for: '{query}'")
# Use the retriever to get relevant documents
relevant_docs = retriever.get_relevant_documents(query)
print("\nRelevant documents retrieved by the retriever:")
for i, doc in enumerate(relevant_docs):
print(f"--- Retrieved Document {i+1} ---")
print(doc.page_content)
if __name__ == "__main__":
main()Quick Check: Vector Stores
Which of the following best describes the primary purpose of a vector store in the context of Retrieval Augmented Generation (RAG)?
Recap: Vector Stores for RAG
Great job! In this lesson, you learned about:
- The role of vector stores in enhancing LLMs through RAG.
- How vector stores store embeddings for semantic search.
- Setting up and interacting with a local ChromaDB.
- Adding documents and performing similarity searches.
- Converting a vector store into a LangChain Retriever.
You now have a solid foundation for implementing data retrieval in your AI agents!
Preguntas frecuentes
¿La lección «Almacenes vectoriales para recuperación» es gratis?
Sí — el texto completo de «Almacenes vectoriales para recuperación» es gratis para leer aquí en la web. Para practicarla de forma interactiva (editor de código integrado y tutor de IA 24/7) y desbloquear el resto del curso de AI Agents with LangChain & Autonomous Workflows, actualiza a CoddyKit PRO. El curso de AI Agents with LangChain & Autonomous Workflows incluye 4 lecciones en total.
¿Qué aprenderé en «Almacenes vectoriales para recuperación»?
Aprenda a utilizar bases de datos vectoriales para almacenar y recuperar eficazmente fragmentos de documentos relevantes según su similitud semántica para RAG (Retrieval Augmented Generation). Practicas AI Agents with LangChain & Autonomous Workflows con código real que ejecutas directamente en el navegador, y un tutor de IA 24/7 responde tus preguntas mientras trabajas en la lección.
¿Necesito experiencia previa para empezar AI Agents with LangChain & Autonomous Workflows?
No se requiere experiencia previa. AI Agents with LangChain & Autonomous Workflows en CoddyKit está estructurado para principiantes hasta estudiantes avanzados, así que puedes empezar aquí o desde el inicio y avanzar a tu ritmo. Esta es la lección 3 de 4.
¿Cuánto tiempo toma la lección «Almacenes vectoriales para recuperación»?
La mayoría de las lecciones de CoddyKit toman alrededor de 5–10 minutos. Cada una es compacta e interactiva, así que avanzas constantemente y retomas exactamente por donde dejaste en la web y la app.
¿Puedo escribir y ejecutar código en esta lección de AI Agents with LangChain & Autonomous Workflows?
Sí. Cada lección de AI Agents with LangChain & Autonomous Workflows incluye un editor de código integrado, así que escribes y ejecutas código real directamente en tu navegador y obtienes retroalimentación instantánea de IA — sin configuración local necesaria.
Todas las lecciones de este curso
- Explicación de los cargadores de documentos
- Divisores de texto y embeddings
- Almacenes vectoriales para recuperación
- Retrievers y compresión contextual