Recuperação de Informações Contextuais
Implemente estratégias para recuperar o contexto mais relevante de seu armazenamento vetorial e ampliar os prompts do LLM.
Recuperação de Informações Contextuais é uma aula grátis de Vector Databases: Pinecone, Weaviate & pgvector no CoddyKit. Esta é a aula 3 de 4. Você pode ler a aula completa abaixo gratuitamente — depois pratica ao vivo no navegador com um editor de código integrado e um tutor de IA 24/7. Faz parte do caminho de aprendizado de Vector Databases: Pinecone, Weaviate & pgvector, e seu progresso é sincronizado entre a web e o app CoddyKit. O curso de Vector Databases: Pinecone, Weaviate & pgvector inclui 4 aulas no total.
Partes desta aula ainda não foram traduzidas e aparecem em inglês.
What is Context Retrieval?
In a Retrieval-Augmented Generation (RAG) system, the Large Language Model (LLM) needs relevant information to generate accurate responses.
- Contextual Information Retrieval is the process of finding and fetching this relevant data from your vector database.
- It's the bridge that connects the user's query to the knowledge stored in your specialized data.
The Retrieval Workflow
When a user asks a question, several steps happen to get the right context:
- The user's question (query) is converted into a vector embedding.
- This query embedding is sent to the vector database.
- The vector database searches for stored document embeddings that are most similar to the query embedding.
- The text chunks associated with these similar embeddings are retrieved and sent to the LLM.
Vector Similarity Basics
The core of retrieval is vector similarity. Your vector database calculates how 'close' your query vector is to all the stored document vectors.
- Closer vectors mean higher semantic similarity.
- Common similarity metrics include cosine similarity or Euclidean distance.
- The database efficiently finds the closest vectors, usually using specialized indexing techniques.
Simple Top-K Retrieval
The most straightforward retrieval strategy is Top-K Retrieval.
- You simply ask the vector database to return the
Kmost similar document chunks to your query. Kis a number you choose (e.g., 3, 5, or 10), representing how many pieces of context you want to provide to the LLM.- While simple, choosing the right
Kis crucial for balancing relevance and LLM token limits.
Python Top-K Retrieval Demo
This simple Python code simulates a vector store and demonstrates how top_k retrieval works. Try changing the top_k value!
import math
class SimpleVectorStore:
def __init__(self):
self.vectors = {}
def add_document(self, doc_id, vector, text):
self.vectors[doc_id] = {"vector": vector, "text": text}
def _cosine_similarity(self, vec1, vec2):
dot_product = sum(v1 * v2 for v1, v2 in zip(vec1, vec2))
magnitude1 = math.sqrt(sum(v**2 for v in vec1))
magnitude2 = math.sqrt(sum(v**2 for v in vec2))
if magnitude1 == 0 or magnitude2 == 0:
return 0.0
return dot_product / (magnitude1 * magnitude2)
def query(self, query_vector, top_k=3):
similarities = []
for doc_id, data in self.vectors.items():
sim = self._cosine_similarity(query_vector, data["vector"])
similarities.append((sim, doc_id, data["text"]))
similarities.sort(key=lambda x: x[0], reverse=True)
return [{"id": s[1], "text": s[2], "similarity": s[0]} for s in similarities[:top_k]]
if __name__ == "__main__":
store = SimpleVectorStore()
store.add_document("doc1", [0.1, 0.2, 0.3], "The quick brown fox jumps over the lazy dog.")
store.add_document("doc2", [0.15, 0.25, 0.35], "A fast fox leaps over a sleepy canine.")
store.add_document("doc3", [0.8, 0.7, 0.9], "Artificial intelligence is transforming industries.")
store.add_document("doc4", [0.75, 0.85, 0.95], "Machine learning algorithms are key to AI.")
query_vec = [0.12, 0.22, 0.32] # Simulating an embedding for "fast animal"
print("--- Top 2 Relevant Chunks ---")
results = store.query(query_vec, top_k=2)
for res in results:
print(f"ID: {res['id']}, Sim: {res['similarity']:.2f}, Text: {res['text']}")
print("\n--- Top 1 Relevant Chunk ---")
results_single = store.query(query_vec, top_k=1)
for res in results_single:
print(f"ID: {res['id']}, Sim: {res['similarity']:.2f}, Text: {res['text']}")Chunking for Effective Retrieval
The quality of your retrieval heavily depends on how your original documents were broken down into chunks before being embedded.
- Chunk Size: Too small, and context might be lost. Too large, and irrelevant information might be included.
- Overlap: Adding overlap between chunks helps ensure that important information isn't split across boundaries.
- Good chunking ensures each retrieved piece of context is meaningful and self-contained.
Enhancing Retrieval with Re-ranking
Sometimes, simple Top-K retrieval isn't enough. The most 'similar' vectors aren't always the most 'relevant' in context.
- Re-ranking is an optional but powerful step performed after initial retrieval.
- It takes the top K chunks from the vector database and uses a smaller, more specialized model to score their relevance more deeply.
- This secondary scoring helps filter out less useful chunks and prioritize truly pertinent information.
The Need for Re-ranking
Why do we need re-ranking?
- Vector similarity can sometimes be fooled by superficial semantic closeness.
- A re-ranker, often a smaller transformer model, can better understand the nuanced relationship between the query and the retrieved document chunks.
- It helps ensure the context provided to the LLM is not only similar but also highly relevant and useful for answering the user's specific question.
Advanced Retrieval Concepts
Beyond basic Top-K and re-ranking, advanced strategies can further improve retrieval:
- Query Expansion: Rewriting or adding terms to the user's original query to improve search results.
- Hybrid Search: Combining traditional keyword search (like full-text search) with vector similarity for a more comprehensive retrieval.
- These methods aim to make the initial retrieval even more robust before context is sent to the LLM.
Check Your Knowledge
Consider a RAG system that initially retrieves 10 document chunks using vector similarity. What are key benefits of adding a re-ranking step?
Contextual Retrieval Summary
You've learned how to retrieve relevant context for RAG systems!
- Contextual retrieval bridges user queries and stored knowledge.
- It involves converting queries to embeddings, querying the vector DB for similar vectors, and fetching associated text.
- Top-K retrieval is the basic method, but good chunking is vital.
- Re-ranking can further refine results by applying a secondary relevance filter.
- Advanced techniques like query expansion and hybrid search offer even more sophisticated retrieval.
Perguntas Frequentes
A aula “Recuperação de Informações Contextuais” é grátis?
Sim — o texto completo de “Recuperação de Informações Contextuais” é grátis para ler aqui na web. Para praticá-la interativamente (um editor de código integrado e um tutor de IA 24/7) e desbloquear o restante do curso de Vector Databases: Pinecone, Weaviate & pgvector, atualize para CoddyKit PRO. O curso de Vector Databases: Pinecone, Weaviate & pgvector inclui 4 aulas no total.
O que vou aprender em “Recuperação de Informações Contextuais”?
Implemente estratégias para recuperar o contexto mais relevante de seu armazenamento vetorial e ampliar os prompts do LLM. Você pratica Vector Databases: Pinecone, Weaviate & pgvector com código prático que executa diretamente no navegador, e um tutor de IA 24/7 responde suas dúvidas enquanto trabalha na aula.
Preciso ter experiência prévia para começar Vector Databases: Pinecone, Weaviate & pgvector?
Nenhuma experiência prévia é necessária. Vector Databases: Pinecone, Weaviate & pgvector no CoddyKit é estruturado para alunos iniciantes até avançados, então você pode começar aqui ou desde o início e aprender no seu ritmo. Esta é a aula 3 de 4.
Quanto tempo leva a aula “Recuperação de Informações Contextuais”?
A maioria das aulas CoddyKit leva cerca de 5–10 minutos. Cada uma é compacta e interativa, então você faz progresso constante e retoma exatamente de onde parou entre web e app.
Posso escrever e executar código nesta aula de Vector Databases: Pinecone, Weaviate & pgvector?
Sim. Cada aula de Vector Databases: Pinecone, Weaviate & pgvector inclui um editor de código integrado, então você escreve e executa código real direto no navegador e recebe feedback de IA instantaneamente — nenhuma configuração local necessária.
Todas as aulas deste curso
- Visão Geral da Arquitetura de um Sistema RAG
- Integração com Estruturas de LLM
- Recuperação de Informações Contextuais
- Estratégias de divisão em partes para RAG