diff --git a/db/library.js b/db/library.js
new file mode 100644
index 0000000..d8e4b0b
--- /dev/null
+++ b/db/library.js
@@ -0,0 +1,224 @@
+// ETL for open library: https://openlibrary.org/developers/dumps
+
+const { pipeline } = require('node:stream/promises')
+const { createGunzip } = require('node:zlib')
+
+const cds = require('@sap/cds');
+
+(async () => {
+  process.env.CDS_TEST_FAKE = 'true'
+  await cds.test(cds.utils.path.resolve(__dirname, '..'))
+  await cds.deploy(cds.options.from)
+})()
+  .then(() => Promise.all([
+    store('https://openlibrary.org/data/ol_dump_authors_latest.txt.gz', 'Authors'),
+    store('https://openlibrary.org/data/ol_dump_works_latest.txt.gz', 'Books'),
+  ]))
+  .then(() => { // Create foreign key index
+    const { Books } = cds.entities
+    return cds.run(`CREATE INDEX ${Books}_author ON ${Books}(author_ID)`)
+  })
+  .then(() => cds.shutdown())
+
+async function store(src, target, maxsize = 1000) {
+  let transform
+  switch (target) {
+    case 'Authors': transform = Authors
+      break;
+    case 'Books': transform = Books
+      break;
+    default: cds.error`unknown target: ${target}`
+  }
+  return pipeline(await open(src), createGunzip(), extract, transform, throughput.bind(target), insert.bind(null, target, maxsize))
+}
+
+async function* Authors(data) {
+  let errors = 0
+  let success = 0
+  for await (const row of data) {
+    try {
+      yield {
+        ID: ID(row.key),
+        name: row.name.slice(0, 255),
+        createdAt: timestamp(row.created),
+        modifiedAt: timestamp(row.last_modified),
+      }
+      success++
+    } catch { errors++ }
+  }
+  console.log(`Transformed ${success} Authors (ignored ${errors})`)
+}
+
+async function* Books(data) {
+  let errors = 0
+  let success = 0
+  for await (const row of data) {
+    try {
+      yield {
+        ID: ID(row.key),
+        title: row.title.slice(0, 255),
+        author_ID: ID(row.authors[0].author.key),
+        createdAt: timestamp(row.created),
+        modifiedAt: timestamp(row.last_modified),
+      }
+      success++
+    } catch { errors++ }
+  }
+  console.log(`Transformed ${success} Books (ignored ${errors})`)
+}
+
+async function insert(target, maxsize, data) {
+  if (typeof maxsize !== 'number') {
+    data = maxsize
+    maxsize = 1000
+  }
+  target = cds.model.entities('sap.capire.bookshop')[target]
+  let buffer = []
+  for await (const row of data) {
+    if (buffer.length >= maxsize) {
+      await UPSERT(buffer).into(target).then(() => { }, err => console.log('error:', err.message))
+      buffer = []
+    }
+    buffer.push(row)
+  }
+  if (buffer.length) await UPSERT(buffer).into(target)
+}
+
+async function write(target, maxsize, data) {
+  if (typeof maxsize !== 'number') {
+    data = maxsize
+    maxsize = 1024 * 1024 * 100
+  }
+  const filename = cds.utils.path.resolve(__dirname, 'data', `sap.capire.bookshop-${target}`)
+  let file, index = 0, size = 0, sep = ''
+  for await (const row of data) {
+    if (!file || size > maxsize) {
+      if (file) file.end(']')
+      file = cds.utils.fs.createWriteStream(`${filename}-${index++}.json`)
+      file.write('[')
+      size = 1
+      sep = ''
+    }
+
+    const chunk = JSON.stringify(row)
+    size += sep.length + chunk.length
+
+    if (sep) { file.write(sep) } else { sep = ',' }
+    file.write(chunk)
+  }
+  file.end(']')
+}
+
+async function* extract(data) {
+  for await (const line of split(data)) {
+    try { yield JSON.parse(line.split('\t').at(-1)) } catch { }
+  }
+}
+
+async function* split(stream) {
+  let leftover = ''
+  for await (const chunk of stream) {
+    const split = `${leftover}${chunk}`.split('\n')
+    for (const line of split) if (line) yield line
+  }
+  if (leftover) yield leftover
+}
+
+const { stdout } = process
+async function throughput_print() {
+  const now = Date.now()
+
+  stdout.cursorTo(0, 0)
+
+  const nameSize = Object.keys(throughput.trackedStream).reduce((l, c) => Math.max(l, c.length), 0)
+  const width = 35 + nameSize
+  const wrap = '\n'.padStart(width, '=')
+  stdout.write(wrap)
+  for (const name in throughput.trackedStream) {
+    const stats = throughput.trackedStream[name]
+    if (!stats.start) {
+      stdout.write(`${name.padStart(nameSize, ' ')}: waiting...\n`)
+      continue
+    }
+    const format = stats.type === 'objects' ? objects : bytes
+    // stdout.write(`${name.padStart(nameSize, ' ')}: ${format(stats.size / ((now - stats.start) / 1000)).padStart(10, ' ')}/sec total: ${format(stats.size).padStart(10, ' ')} (${time(now - stats.start)})\n`)
+    stdout.write(`${name.padStart(nameSize, ' ')}: ${stats.size.toLocaleString().padStart(23, ' ')} (${time(now - stats.start)})\n`)
+  }
+  stdout.write(wrap)
+}
+
+function bytes(bytes) {
+  let scale = 0
+  while (bytes > 1024) {
+    bytes /= 1024
+    scale++
+  }
+  const scales = ['  B', 'KiB', 'MiB', 'GiB', 'TiB']
+  return `${(bytes >>> 0).toLocaleString()} ${scales[scale]}`
+}
+
+function objects(count) {
+  let scale = 0
+  while (count > 1000) {
+    count /= 1000
+    scale++
+  }
+  const scales = ['K  ', 'K  ', 'M  ', 'B  ', 'T  ', 'Q  ', 'P  ']
+  return `${(count >>> 0).toLocaleString()} ${scales[scale]}`
+}
+
+function time(milliseconds) {
+  const scales = ['mil', 'sec', 'min', 'Hur', 'Day', 'Wek', 'Mon', 'Yer']
+  const scaleSize = [1000, 60, 60, 24, 7, 29, 256]
+  let scale = 0
+  while (milliseconds > scaleSize[scale]) {
+    milliseconds /= scaleSize[scale]
+    scale++
+  }
+  return `${(milliseconds >>> 0).toLocaleString()} ${scales[scale]}`
+}
+
+async function* throughput(name, stream) {
+  if (typeof name !== 'string') {
+    stream = name
+    name = this
+  }
+
+  const stats = (throughput.trackedStream ??= {})[name] = { start: undefined, size: 0, type: undefined }
+  for await (const chunk of stream) {
+    throughput.interval ??= setInterval(throughput_print, stdout.isTTY ? 100 : 10_000)
+    stats.start ??= Date.now()
+    stats.type ??= typeof (chunk?.bytelength ?? chunk?.length) === 'number' ? 'bytes' : 'objects'
+    stats.size += chunk?.bytelength ?? chunk?.length ?? 1
+    yield chunk
+  }
+  delete throughput.trackedStream[name]
+  if (!Object.keys(throughput.trackedStream).length) throughput.interval = clearInterval(throughput.interval)
+}
+
+async function open(src) {
+  const filename = cds.utils.path.resolve(__dirname, 'data', cds.utils.path.basename(src))
+
+  if (!cds.utils.fs.existsSync(filename)) {
+    const download = await fetch(src)
+    if (download.status > 399) cds.error`Failed to download ${filename} (${download.statusText || download.status})`
+    await pipeline(download.body, cds.utils.fs.createWriteStream(filename))
+  }
+
+  return cds.utils.fs.createReadStream(filename)
+}
+
+function timestamp(val) {
+  if (!val) return null
+  const value = val.value || val
+  return value[value.length - 1] === 'Z' ? value : `${value}Z`
+}
+
+function ID(key) {
+  return Number.parseInt(/\d+/.exec(key)[0])
+}
+
+module.exports = {
+  open, extract, write, throughput,
+  Authors, Books,
+}