Pglite Client DB Feedback | 客户端 Pglite DB 功能反馈

[arvinxx]

Looking to the future, LobeChat introduces a brand new Client DB engine in v1.37.0: PGlite. It is the wasm version of our server-side database Postgres, capable of running in the browser with a relatively small footprint while providing database capabilities that are fully consistent with Postgres, which means:

• In a pure client, we will have complete knowledge base dialogue capabilities (pglite supports core vector retrieval functions);
• Due to the complete consistency of the schema between the two ends of the db, capabilities that were previously only available in the server-side db will subsequently be fully available on the client side as well. (branch conversations is available in 1.38.0); The implementation of the database on both the client and server sides will only need to be done once, significantly reducing future development costs, and most features will be able to be updated synchronously to both the client and server versions in the future; With the help of a CRDT collaboration engine like electric-sql, we will be able to truly integrate a unified client/server in the future, achieving a local-first, on-demand synchronization ultimate form.
• In the future, we may even achieve multi-person online collaborative AI conversations through group chats!

However, any forward-looking innovation comes at a cost. While we are very optimistic about the potential of PGlite, it is still a very early project at this point in time, and there are currently no large-scale user projects that are truly using it in production scenarios. As a pioneering explorer of open-source AI projects, LobeChat hopes to explore the possibilities of using PGlite together with the community and to discover new best practices for applications in the AI era!

How to use

This change has no impact on users using the server-side DB.

To avoid affecting the existing client DB users, we have added a new environment variable to enable pglite. You can enable the use of pglite by adding NEXT_PUBLIC_CLIENT_DB=pglite during the application build.

[!IMPORTANT] Note: Currently, automatic data migration has not been implemented. Please manually back up the existing data before adding variables.

We will enable this engine by default in the major version of LobeChat V2 and remove all existing implementations of dexie.

---

面向未来，LobeChat 在 v1.37.0 中提供了一种全新的 Client DB 引擎： PGlite。它是我们服务端数据库 Postgres 的 wasm 版本，可以以相对较小的体积在浏览器中运行，并获得与 Postgres 完全一致的数据库能力，这意味着：

• 在纯客户端中，我们将可以拥有完整的知识库对话能力（pglite 支持核心的向量检索功能）；
• 由于两端 db 的schema 做到了完全一致，之前只有在服务端 db 中才可用的能力，后续将在客户端中也完全可用。（1.38.0 已支持分支对话）；
• 客户端和服务端的 db 实现将只需做一次，大大降低了未来的研发成本，未来大部分特性将能同步更新到客户端与服务端版本中；
• 借助 electric-sql 这样的 CRDT 协同引擎，我们将可以在后续真正融合统一客户端/服务端，实现本地优先，按需同步的终极形态。
• 甚至，在未来我们可以通过群聊中实现多人在线协同的 ai 对话！

但是，任何面向未来的创新都是有代价的，虽然我们非常看好 PGlite 的潜力，但 PGlite 在当前的时间点下仍然是一个非常早期的项目，目前也没有一个大规模用户的项目真正在生产场景下使用它。而 LobeChat 作为开源 AI 项目的先锋探索者，我们希望与社区一起探索使用 PGlite 的可能性，一起来探索 AI 时代新的应用最佳实践！

如何使用

本次变更对于使用服务端 DB 的用户来说没有任何影响。

为了不影响原有客户端 DB 用户的使用，我们新增了一个环境变量用于启用 pglite，你可以通过在应用构建时添加 NEXT_PUBLIC_CLIENT_DB=pglite 来开启 pglite 的使用。

[!IMPORTANT] 注意事项：目前没有实现数据的自动迁移，请在添加变量前手动备份原有的数据。

我们将会在 LobeChat V2 大版本中默认开启该引擎，并移除原有的 dexie 的所有实现。