BFF scripts: local and remote in sync

This page explains how to have an Agent develop and sync BFF scripts on the platform. You don't need to memorize sync commands — describe the server-side logic you want to add, change, reuse, or publish, and the Agent checks local files, platform resources, and risks.

Tasks to hand to the Agent

• Check whether local BFF scripts match the platform.
• Add an endpoint-style script for the frontend or pages to call.
• Extract duplicated logic into a shared function.
• Add hook logic to dataset operations.
• Push local changes to the platform, or pull platform scripts back to local.

Prerequisites

• The target project is open, and you've confirmed it's the app you want to modify.
• You know this logic's inputs, outputs, and callers.
• When live impact is possible, ask the Agent to preview the diff and risks first.

Common prompts

Check local vs. platform consistency

请帮我检查当前项目的 BFF 脚本和平台是否一致。
请列出：本地已修改但未同步的脚本、平台有但本地没有的脚本、可能冲突的脚本。
先不要执行覆盖或删除，只给我差异和建议处理方式。

The AI may call: rabetbase bff status, rabetbase bff list, rabetbase bff detail.

Add an endpoint-style script

请帮我新增一个 BFF 接口，名称是「客户跟进摘要」。
它用于「客户详情页展示最近 5 条跟进记录、下一次跟进时间和未完成待办数量」。
输入参数包括「customerId」，返回字段包括「最近跟进记录列表、下一次跟进时间、未完成待办数量、最后跟进人」。
请先根据相关数据集确认字段和查询方式，再生成脚本，并告诉我如何验证。

The AI may call: rabetbase dataset detail, rabetbase dataset operations, rabetbase bff create, rabetbase bff push.

Extract shared logic

我发现「客户详情页、销售工作台、客户列表页」里都在计算客户跟进状态。
请帮我抽成一个公共 BFF 函数，并调整调用方复用它。
修改前先列出会影响哪些脚本，修改后给出验证清单。

The AI may call: rabetbase bff list, rabetbase bff detail, rabetbase bff status, rabetbase bff push.

Add a data operation hook

请为「客户」数据集的「更新客户状态」操作增加一个钩子。
业务规则是：当客户状态变为「已成交」时，自动写入一条成交跟进记录，并更新客户最近跟进时间。
请先确认这个操作的入参和返回结构，再生成脚本。
执行前预览影响范围，避免影响无关操作。

The AI may call: rabetbase dataset operations, rabetbase bff create, rabetbase bff push.

Sync to the platform

本地的「客户跟进摘要」脚本已经改完。
请先检查它和平台当前版本的差异，说明是否会覆盖远端内容。
如果风险可接受，再把它同步到平台，并告诉我同步结果和验证方式。

The AI may call: rabetbase bff status, rabetbase bff detail, rabetbase bff push.

What to add to your prompt

• Script goal: query, write, validate, extract shared logic, or hook into an operation.
• Inputs and outputs: parameter names, required fields, returned fields, error-handling requirements.
• Dependencies: related datasets, SQL, frontend pages, or callers.
• Sync strategy: check only, pull to local, push to platform, or preview-then-execute.

Completion checks

• The Agent has listed which local files were added or modified.
• Platform and local script states match, or the differences are explicitly kept.
• Script inputs and outputs match the business requirement.
• There's a runnable way to verify — sample parameters, a page call, or a runtime test run.

Next step: if the script depends on complex queries, start with "Custom SQL: from writing to execution". To generate frontend call snippets, continue with "Code generation and menus".