diff --git a/.gitignore b/.gitignore
index 0e5a8a8..77485ad 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,5 +1,6 @@
 coverage
 node_modules
 package-lock.json
-dist
 *.tgz
+.DS_Store
+/types
diff --git a/README.md b/README.md
index cd50f63..07d1134 100644
--- a/README.md
+++ b/README.md
@@ -1,17 +1,21 @@
 # Hyparquet Writer
 
+[![workflow status](https://github.com/hyparam/hyparquet-writer/actions/workflows/ci.yml/badge.svg)](https://github.com/hyparam/hyparquet-writer/actions)
 [![mit license](https://img.shields.io/badge/License-MIT-orange.svg)](https://opensource.org/licenses/MIT)
 ![coverage](https://img.shields.io/badge/Coverage-96-darkred)
+[![dependencies](https://img.shields.io/badge/Dependencies-0-blueviolet)](https://www.npmjs.com/package/hyparquet?activeTab=dependencies)
 
 ## Usage
 
-```javascript
-import { writeParquet } from 'hyparquet-writer'
+Call `parquetWrite` with a list of columns, each column is an object with a `name` and `data` field. The `data` field should be an array of same-type values.
 
-const arrayBuffer = writeParquet({
-  name: ['Alice', 'Bob', 'Charlie'],
-  age: [25, 30, 35],
-})
+```javascript
+import { parquetWrite } from 'hyparquet-writer'
+
+const arrayBuffer = parquetWrite([
+  { name: 'name', data: ['Alice', 'Bob', 'Charlie'] },
+  { name: 'age', data: [25, 30, 35] },
+])
 ```
 
 ## References
diff --git a/package.json b/package.json
index 06d63b8..0ddf1ff 100644
--- a/package.json
+++ b/package.json
@@ -6,7 +6,8 @@
   "homepage": "https://hyperparam.app",
   "keywords": [
     "hyparquet",
-    "parquet"
+    "parquet",
+    "thrift"
   ],
   "license": "MIT",
   "repository": {
@@ -15,14 +16,27 @@
   },
   "main": "src/index.js",
   "files": [
-    "src"
+    "src",
+    "types"
   ],
   "type": "module",
   "types": "src/index.d.ts",
+  "exports": {
+    ".": {
+      "import": "./src/index.js",
+      "types": "./types/index.d.ts"
+    },
+    "./src/*.js": {
+      "import": "./src/*.js",
+      "types": "./types/*.d.ts"
+    }
+  },
   "scripts": {
+    "build:types": "tsc -p ./tsconfig.build.json",
     "coverage": "vitest run --coverage",
     "lint": "eslint",
     "lint:fix": "eslint --fix",
+    "prepare": "npm run build:types",
     "test": "vitest run"
   },
   "devDependencies": {
diff --git a/src/convert.js b/src/convert.js
index 74e4787..6973990 100644
--- a/src/convert.js
+++ b/src/convert.js
@@ -13,9 +13,14 @@ export function unconvert(schemaElement, values) {
     return values.map(v => v.getTime())
   }
   if (ctype === 'JSON') {
-    const encoder = new TextEncoder()
     if (!Array.isArray(values)) throw new Error('JSON must be an array')
+    const encoder = new TextEncoder()
     return values.map(v => encoder.encode(JSON.stringify(v)))
   }
+  if (ctype === 'UTF8') {
+    if (!Array.isArray(values)) throw new Error('strings must be an array')
+    const encoder = new TextEncoder()
+    return values.map(v => encoder.encode(v))
+  }
   return values
 }
diff --git a/src/plain.js b/src/plain.js
index c8be649..a2f42ef 100644
--- a/src/plain.js
+++ b/src/plain.js
@@ -84,10 +84,9 @@ function writePlainDouble(writer, values) {
  * @param {DecodedArray} values
  */
 function writePlainByteArray(writer, values) {
-  const encoder = new TextEncoder()
   for (const value of values) {
-    const bytes = typeof value === 'string' ? encoder.encode(value) : value
-    writer.appendUint32(bytes.length)
-    writer.appendBytes(bytes)
+    if (!(value instanceof Uint8Array)) throw new Error('BYTE_ARRAY must be Uint8Array')
+    writer.appendUint32(value.length)
+    writer.appendBytes(value)
   }
 }
diff --git a/src/schema.js b/src/schema.js
index bc68035..62e5859 100644
--- a/src/schema.js
+++ b/src/schema.js
@@ -29,7 +29,13 @@ export function getSchemaElementForValues(name, values) {
       else if (typeof value === 'bigint') valueType = 'INT64'
       else if (Number.isInteger(value)) valueType = 'INT32'
       else if (typeof value === 'number') valueType = 'DOUBLE'
-      else if (typeof value === 'string') valueType = 'BYTE_ARRAY'
+      else if (value instanceof Uint8Array) valueType = 'BYTE_ARRAY'
+      else if (typeof value === 'string') {
+        valueType = 'BYTE_ARRAY'
+        // make sure they are all strings
+        if (type && !converted_type) throw new Error('mixed types not supported')
+        converted_type = 'UTF8'
+      }
       else if (value instanceof Date) {
         valueType = 'INT64'
         // make sure they are all dates
diff --git a/src/thrift.js b/src/thrift.js
index 0124965..ab626e9 100644
--- a/src/thrift.js
+++ b/src/thrift.js
@@ -21,7 +21,7 @@ const CompactType = {
  *
  * Expects keys named like "field_1", "field_2", etc. in ascending order.
  *
- * @import {Writer} from './types.js'
+ * @import {Writer} from '../src/types.js'
  * @param {Writer} writer
  * @param {Record<string, any>} data
  */
diff --git a/src/types.d.ts b/src/types.d.ts
index b3dabb5..f10648b 100644
--- a/src/types.d.ts
+++ b/src/types.d.ts
@@ -1,3 +1,9 @@
+import { DecodedArray, ParquetType } from "hyparquet"
+
+export interface ColumnData {
+  name: string
+  data: DecodedArray
+}
 
 export interface Writer {
   buffer: ArrayBuffer
diff --git a/src/write.js b/src/write.js
index 93e5587..f4a146e 100644
--- a/src/write.js
+++ b/src/write.js
@@ -7,17 +7,17 @@ import { getSchemaElementForValues } from './schema.js'
  * Write data as parquet to an ArrayBuffer
  *
  * @import {ColumnChunk, DecodedArray, FileMetaData, SchemaElement, SchemaTree} from 'hyparquet'
- * @param {Record<string, DecodedArray>} columnData
+ * @import {ColumnData} from '../src/types.js'
+ * @param {ColumnData[]} columnData
  * @returns {ArrayBuffer}
  */
 export function parquetWrite(columnData) {
   const writer = new Writer()
 
   // Check if all columns have the same length
-  const columnNames = Object.keys(columnData)
-  const num_rows = columnNames.length ? BigInt(columnData[columnNames[0]].length) : 0n
-  for (const name of columnNames) {
-    if (BigInt(columnData[name].length) !== num_rows) {
+  const num_rows = columnData.length ? BigInt(columnData[0].data.length) : 0n
+  for (const { data } of columnData) {
+    if (BigInt(data.length) !== num_rows) {
       throw new Error('parquetWrite: all columns must have the same length')
     }
   }
@@ -29,7 +29,7 @@ export function parquetWrite(columnData) {
   /** @type {SchemaElement[]} */
   const schema = [{
     name: 'root',
-    num_children: columnNames.length,
+    num_children: columnData.length,
   }]
 
   // row group columns
@@ -37,9 +37,9 @@ export function parquetWrite(columnData) {
   const columns = []
 
   // Write columns
-  for (const name of columnNames) {
-    const values = columnData[name]
-    const schemaElement = getSchemaElementForValues(name, values)
+  for (const { name, data } of columnData) {
+    // auto-detect type
+    const schemaElement = getSchemaElementForValues(name, data)
     if (!schemaElement.type) throw new Error(`column ${name} cannot determine type`)
     const file_offset = BigInt(writer.offset)
     /** @type {SchemaElement[]} */
@@ -47,7 +47,7 @@ export function parquetWrite(columnData) {
       schema[0],
       schemaElement,
     ]
-    const meta_data = writeColumn(writer, schemaPath, values)
+    const meta_data = writeColumn(writer, schemaPath, data)
 
     // save metadata
     schema.push(schemaElement)
diff --git a/src/writer.js b/src/writer.js
index e0e1749..b83d593 100644
--- a/src/writer.js
+++ b/src/writer.js
@@ -2,8 +2,7 @@
 /**
  * Self-expanding buffer view
  *
- * @import {Writer} from './types.js'
- * @returns {Writer}
+ * @returns {import('../src/types.js').Writer}
  */
 export function Writer() {
   this.buffer = new ArrayBuffer(1024)
diff --git a/test/metadata.test.js b/test/metadata.test.js
index 582c1d5..fa224e3 100644
--- a/test/metadata.test.js
+++ b/test/metadata.test.js
@@ -16,7 +16,7 @@ export const exampleMetadata = {
     { name: 'int', type: 'INT32', repetition_type: 'REQUIRED' },
     { name: 'bigint', type: 'INT64', repetition_type: 'REQUIRED' },
     { name: 'double', type: 'DOUBLE', repetition_type: 'REQUIRED' },
-    { name: 'string', type: 'BYTE_ARRAY', repetition_type: 'REQUIRED' },
+    { name: 'string', type: 'BYTE_ARRAY', repetition_type: 'REQUIRED', converted_type: 'UTF8' },
     { name: 'nullable', type: 'BOOLEAN', repetition_type: 'OPTIONAL' },
   ],
   num_rows: 4n,
@@ -110,7 +110,7 @@ export const exampleMetadata = {
     total_byte_size: 236n,
     num_rows: 4n,
   }],
-  metadata_length: 336,
+  metadata_length: 338,
 }
 
 describe('writeMetadata', () => {
diff --git a/test/plain.test.js b/test/plain.test.js
index 0e9b0c7..78113f7 100644
--- a/test/plain.test.js
+++ b/test/plain.test.js
@@ -62,7 +62,10 @@ describe('writePlain', () => {
   it('writes BYTE_ARRAY', () => {
     const writer = new Writer()
     const strings = ['a', 'b', 'c', 'd']
-    writePlain(writer, strings, 'BYTE_ARRAY')
+    // strings must be pre-converted to Uint8Array
+    const encoder = new TextEncoder()
+    const bytes = strings.map(s => encoder.encode(s))
+    writePlain(writer, bytes, 'BYTE_ARRAY')
 
     let offset = 0
     for (const s of strings) {
diff --git a/test/write.test.js b/test/write.test.js
index 1afd009..c973b61 100644
--- a/test/write.test.js
+++ b/test/write.test.js
@@ -6,32 +6,33 @@ import { exampleMetadata } from './metadata.test.js'
 /**
  * Utility to encode a parquet file and then read it back into a JS object.
  *
- * @param {Record<string, any[]>} columnData
+ * @import {ColumnData} from '../src/types.js'
+ * @param {ColumnData[]} columnData
  * @returns {Promise<Record<string, any>>}
  */
 async function roundTripDeserialize(columnData) {
   const file = parquetWrite(columnData)
-  return await parquetReadObjects({ file })
+  return await parquetReadObjects({ file, utf8: false })
 }
 
-const data = {
-  bool: [true, false, true, false], // BOOLEAN
-  int: [0, 127, 0x7fff, 0x7fffffff], // INT32
-  bigint: [0n, 127n, 0x7fffn, 0x7fffffffffffffffn], // INT64
-  double: [0, 0.0001, 123.456, 1e100], // DOUBLE
-  string: ['a', 'b', 'c', 'd'], // BYTE_ARRAY
-  nullable: [true, false, null, null], // BOOLEAN nullable
-}
+const basicData = [
+  { name: 'bool', data: [true, false, true, false] },
+  { name: 'int', data: [0, 127, 0x7fff, 0x7fffffff] },
+  { name: 'bigint', data: [0n, 127n, 0x7fffn, 0x7fffffffffffffffn] },
+  { name: 'double', data: [0, 0.0001, 123.456, 1e100] },
+  { name: 'string', data: ['a', 'b', 'c', 'd'] },
+  { name: 'nullable', data: [true, false, null, null] },
+]
 
 describe('parquetWrite', () => {
   it('writes expected metadata', () => {
-    const file = parquetWrite(data)
+    const file = parquetWrite(basicData)
     const metadata = parquetMetadata(file)
     expect(metadata).toEqual(exampleMetadata)
   })
 
   it('serializes basic types', async () => {
-    const result = await roundTripDeserialize(data)
+    const result = await roundTripDeserialize(basicData)
     expect(result).toEqual([
       { bool: true, int: 0, bigint: 0n, double: 0, string: 'a', nullable: true },
       { bool: false, int: 127, bigint: 127n, double: 0.0001, string: 'b', nullable: false },
@@ -46,16 +47,17 @@ describe('parquetWrite', () => {
     bool[100] = false
     bool[500] = true
     bool[9999] = false
-    const buffer = parquetWrite({ bool })
+    const buffer = parquetWrite([{ name: 'bool', data: bool }])
     expect(buffer.byteLength).toBe(1399)
     const metadata = parquetMetadata(buffer)
     expect(metadata.metadata_length).toBe(89)
   })
 
   it('serializes list types', async () => {
-    const result = await roundTripDeserialize({
-      list: [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12]],
-    })
+    const result = await roundTripDeserialize([{
+      name: 'list',
+      data: [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12]],
+    }])
     expect(result).toEqual([
       { list: [1, 2, 3] },
       { list: [4, 5, 6] },
@@ -65,9 +67,10 @@ describe('parquetWrite', () => {
   })
 
   it('serializes object types', async () => {
-    const result = await roundTripDeserialize({
-      obj: [{ a: 1, b: 2 }, { a: 3, b: 4 }, { a: 5, b: 6 }, { a: 7, b: 8 }],
-    })
+    const result = await roundTripDeserialize([{
+      name: 'obj',
+      data: [{ a: 1, b: 2 }, { a: 3, b: 4 }, { a: 5, b: 6 }, { a: 7, b: 8 }],
+    }])
     expect(result).toEqual([
       { obj: { a: 1, b: 2 } },
       { obj: { a: 3, b: 4 } },
@@ -77,9 +80,10 @@ describe('parquetWrite', () => {
   })
 
   it('serializes date types', async () => {
-    const result = await roundTripDeserialize({
-      date: [new Date(0), new Date(100000), new Date(200000), new Date(300000)],
-    })
+    const result = await roundTripDeserialize([{
+      name: 'date',
+      data: [new Date(0), new Date(100000), new Date(200000), new Date(300000)],
+    }])
     expect(result).toEqual([
       { date: new Date(0) },
       { date: new Date(100000) },
@@ -88,8 +92,21 @@ describe('parquetWrite', () => {
     ])
   })
 
+  it('serializes byte array types', async () => {
+    const result = await roundTripDeserialize([{
+      name: 'bytes',
+      data: [Uint8Array.of(1, 2, 3), Uint8Array.of(4, 5, 6), Uint8Array.of(7, 8, 9), Uint8Array.of(10, 11, 12)],
+    }])
+    expect(result).toEqual([
+      { bytes: Uint8Array.of(1, 2, 3) },
+      { bytes: Uint8Array.of(4, 5, 6) },
+      { bytes: Uint8Array.of(7, 8, 9) },
+      { bytes: Uint8Array.of(10, 11, 12) },
+    ])
+  })
+
   it('throws for mixed types', () => {
-    expect(() => parquetWrite({ mixed: [1, 2, 3, 'boom'] }))
-      .toThrow('parquet cannot write mixed types: INT32 and BYTE_ARRAY')
+    expect(() => parquetWrite([{ name: 'mixed', data: [1, 2, 3, 'boom'] }]))
+      .toThrow('mixed types not supported')
   })
 })
diff --git a/tsconfig.build.json b/tsconfig.build.json
new file mode 100644
index 0000000..51c713e
--- /dev/null
+++ b/tsconfig.build.json
@@ -0,0 +1,11 @@
+{
+  "extends": "./tsconfig.json",
+  "compilerOptions": {
+    "noEmit": false,
+    "declaration": true,
+    "emitDeclarationOnly": true,
+    "outDir": "types",
+    "declarationMap": true
+  },
+  "include": ["src"]
+}