Skip to content

Getting development work done with Vibe Coding

This group of sub-pages covers how to hand development tasks to an Agent inside your Vibe Coding tool, letting it call rabetbase capabilities to work with datasets, SQL, BFF, system page development, code generation, and menu sync. You never assemble commands by hand or memorize flags — the focus is on stating your business goal, constraints, and acceptance criteria clearly.

How to use it: open Codex, Claude Code, Trae, or Cursor, and describe your requirement to the Agent using the prompts from these pages. The Agent checks the project, picks the right capability, previews risks, executes the change, and reports back.

How to read this group

TaskSub-pageWhat you tell the AgentWhat you should see when it's done
Understand your data and APIsAPIs and datasets: pull config, see your data clearlyWhich project to sync, which dataset to inspect, which fields or operations to confirm.Local API config is aligned; fields, operations, and table relations are explained.
Build reusable queriesCustom SQL: from writing to executionWhat the business needs to query, the input parameters, which fields to return, and whether to save it as platform SQL.The SQL is validated, saved or updated, and can be test-run with sample parameters.
Wrap server-side logicBFF scripts: local and remote in syncWhat kind of server-side logic to add or change, its inputs and outputs, reuse rules, and release scope.Local scripts and platform state are clear, changes are previewed, and sync happens after your confirmation.
Build system pagesSystem page development: business pages with the SDKWhat problem the page solves, which data it uses, which interaction states and acceptance paths it needs.The page is wired into the Lovrabet SDK, with clear data reads/writes, state handling, and local verification steps.
Generate call code and entry pointsCode generation and menusWhich call code to generate, where to put it, and how page menus or asset URLs should align.TypeScript call snippets, menu structure, or live asset references are in place and verifiable.

How to write a good prompt

A good prompt usually covers four things: the business goal, the objects in scope, constraints, and acceptance criteria.

text
请帮我开发一个客户列表页。
背景是:销售需要按客户等级、负责人和最近跟进时间筛选客户,并快速看到客户最近一次跟进情况。
对象是:客户、联系人、跟进记录、客户活跃度统计 SQL、客户跟进摘要 BFF 和客户列表页菜单。
约束是:不要使用 mock 数据;写入前先确认可写字段;菜单和资源更新前先预览影响范围。
完成后请告诉我:改了哪些文件,页面路由是什么,本地如何验证,还有哪些风险需要确认。

CLI capabilities the AI may call (chosen per task): rabetbase api, rabetbase dataset, rabetbase sql, rabetbase bff, rabetbase codegen, rabetbase run, rabetbase menu.

Don't use it like this

  • Don't treat these pages as a flag reference to copy command by command.
  • Don't just say "handle this for me" — state the business objects, the expected result, and how you'll accept it.
  • Don't put sensitive information, keys, or personal paths in a prompt — describe them with placeholders.
  • For writes, deletes, overwrites, and releases, require the Agent to preview the impact before executing.
  1. Start with the business goal and let the Agent map out datasets, fields, operations, and relations.
  2. For complex queries, have the Agent design and validate platform-hosted SQL.
  3. For server-side logic, have the Agent add or modify BFF and explain local vs. platform differences.
  4. For system pages, have the Agent wire in real data via the SDK and complete page states and verification paths.
  5. For frontend integration, have the Agent generate call snippets and place them in the designated files or pages.
  6. Before release, have the Agent check menus, asset references, and verification paths.

Sub-pages

<sub-page-list space-id="7535363963853209604" wiki-token="S71hw5c2HiwyoHkLgyecKrI8n3M"><sub-page doc-id="HOIEdnXNDoNEbsxlIcAcsYhNnUd" file-type="docx" title="API 与数据集:拉配置、看清数据"/><sub-page doc-id="WgMjd6rzooJPiaxgRIYcKGk1nYg" file-type="docx" title="系统页面开发:用 SDK 完成业务页面"/><sub-page doc-id="U69vdbaq9oYz2SxC1sXcwTDynQd" file-type="docx" title="BFF 脚本:本地与远端同步"/><sub-page doc-id="LiETd90f9oWk8Vx3UQUcPe7nn7b" file-type="docx" title="自定义 SQL:从写到执行"/><sub-page doc-id="DAHxdf4C8oqKwAxBU7PcfjpDnQb" file-type="docx" title="代码生成与菜单"/></sub-page-list>

基于飞书知识库同步生成,内容以飞书源文档为准