// 把 kb 快照灌回一个本地空库。只写本地,不碰生产。 // // 这个文件要放在 streamline 仓库【根目录】下再跑(跟 package.json 同一层), // 因为下面 import schema 用的是相对路径。 // // 用法: // 1) 先把本地空库的表建出来(本地库,不是生产,直接 push 就行): // DATABASE_URL=postgres://127.0.0.1:5432/streamline_dev bun x drizzle-kit push // 2) 解压快照: gunzip -k streamline-kb-snapshot.json.gz // 3) 灌数据: // LOCAL_DATABASE_URL=postgres://127.0.0.1:5432/streamline_dev \ // bun kb_snapshot_restore.ts ./streamline-kb-snapshot.json // // 全程只对 LOCAL_DATABASE_URL 写;这个变量跟生产的 DATABASE_URL 无关,自己起的本地库随便填。 import { drizzle } from "drizzle-orm/bun-sql"; import { SQL } from "bun"; import { readFileSync } from "node:fs"; import * as schema from "./control-server/src/schema"; const local = process.env.LOCAL_DATABASE_URL; if (!local) { console.error("设置 LOCAL_DATABASE_URL 指向你本地那个空库(不是生产 DATABASE_URL)"); process.exit(1); } const file = process.argv[2]; if (!file) { console.error("用法: bun kb_snapshot_restore.ts "); process.exit(1); } const client = new SQL(local); const db = drizzle({ client, schema, casing: "snake_case" }); const data = JSON.parse(readFileSync(file, "utf8")) as Record[]>; // kb_claims 必须先于 kb_evidence(外键) —— 其余表互相之间没有外键约束,顺序不敏感。 const ORDER = [ "kb_domains", "kb_claims", "kb_evidence", "kb_distill_runs", "kb_proposals", "kb_injections", "run_conditions", ] as const; for (const name of ORDER) { const rows = data[name] ?? []; const table = (schema as Record)[name]; if (!rows.length) { console.log(`${name}: 0 行,跳过`); continue; } const batch = 400; for (let i = 0; i < rows.length; i += batch) { await db.insert(table as never).values(rows.slice(i, i + batch) as never); } console.log(`${name}: 写入 ${rows.length} 行`); } console.log("done"); process.exit(0);