Unverified50% confidenceSolutionExact time
The retrieval step in RAG (embedding plus vector database querying) is typically a CPU-intensive or lightweight GPU task and should be deployed separately from LLM generation.
1
Sources
50%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedRAG（检索增强生成）技术架构将文档切片、向量化、存入向量数据库，查询时先检索相关片段再交给模型生成回答76% similarUnverifiedRAG (Retrieval-Augmented Generation) and vector database technologies can partially alleviate the problem of LLMs lacking persistent memory, but introduce additional engineering complexity.74% similarUnverified向量数据库是 RAG（检索增强生成）技术的核心基础设施，底层核心算法是近似最近邻搜索（ANN）73% similarVerifiedRAG (Retrieval-Augmented Generation) typically comprises three core components: a vector database, a retriever (which recalls relevant document chunks based on cosine similarity), and a generator.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/912011API
curl https://kongchang.com/api/v1/knowledge/claims/912011MCP
get_claim(id=912011)