📄 中文摘要
在构建 ADK（Agent Development Kit）应用程序时，遇到了常见的架构选择问题。最初使用 Chroma 作为 RAG（检索增强生成）后端，在本地开发时运行良好，但在迁移到云端时遇到复杂性。需要一个生产就绪且具有弹性的解决方案，避免在 Mac Mini 上管理有状态资产或支付额外的托管向量数据库费用。最终选择了 Postgres，通过 pgvector 将关系数据库转变为强大的向量存储，提供了理想的解决方案。
简单且廉价的 RAG - genai-toolbox 和 pgvector
出处: Simple and cheap RAG - genai-toolbox and pgvector
发布: 2026年3月18日
📄 English Summary
Simple and cheap RAG - genai-toolbox and pgvector
While building the ADK (Agent Development Kit) application, a common architectural decision point was encountered. Initially, Chroma was used as the RAG (Retrieval-Augmented Generation) backend, which worked well for local development but became complicated when moving to the cloud. A production-ready and resilient solution was needed that did not involve managing stateful assets on a Mac Mini or paying for a separate managed vector database. The solution was found in Postgres, which, by utilizing pgvector, can transform a relational database into a powerful vector store, providing an ideal solution.