Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
'use strict'
const assert = require('node:assert/strict')
const fs = require('node:fs')
const path = require('node:path')
const { describe, it } = require('node:test')
const { classifyBytes, FILE_CLASSES, MEDIA_TYPE_BY_CLASS } = require('./fileClassifier.cjs')
const EXPECTED_VECTOR_COUNT = 38
const { vectors } = JSON.parse(
fs.readFileSync(path.join(__dirname, 'fixtures', 'classifier-vectors.v1.json'), 'utf8')
)
function bytesOf(vector) {
return new Uint8Array(Buffer.from(vector.input.bytesBase64, 'base64'))
}
describe('classifier vectors', () => {
it('loads every vector', () => {
assert.equal(vectors.length, EXPECTED_VECTOR_COUNT)
})
for (const vector of vectors) {
it(vector.name, () => {
const actual = classifyBytes({
bytes: bytesOf(vector),
totalByteLength: vector.input.totalByteLength,
declaredMediaType: vector.input.declaredMediaType,
filename: vector.input.filename,
})
assert.deepEqual(actual, vector.expected)
})
}
it('covers every file class and every detection', () => {
const classes = new Set(vectors.map(vector => vector.expected.class))
assert.deepEqual([...classes].sort(), [...FILE_CLASSES].sort())
const detections = new Set(vectors.map(vector => vector.expected.detection))
assert.deepEqual([...detections].sort(), ['declared', 'magic', 'text_utf8'])
})
})
describe('classifyBytes', () => {
it('maps every class to a detected media type', () => {
for (const fileClass of FILE_CLASSES)
assert.equal(typeof MEDIA_TYPE_BY_CLASS[fileClass], 'string')
})
it('rejects input that is not a Uint8Array', () => {
assert.throws(() => classifyBytes({ bytes: 'abc', totalByteLength: 3 }), TypeError)
})
it('rejects a total length smaller than the bytes provided', () => {
assert.throws(
() => classifyBytes({ bytes: new Uint8Array([1, 2, 3]), totalByteLength: 2 }),
TypeError
)
})
it('does not decode a binary file as text', () => {
const result = classifyBytes({
bytes: new Uint8Array([0x25, 0x50, 0x44, 0x46, 0x2d, 0x00, 0xff]),
totalByteLength: 7,
filename: 'x.txt',
})
assert.equal(result.class, 'pdf')
assert.equal(result.textReadable, false)
assert.equal(result.reader, 'none')
})
it('ignores media type parameters and case', () => {
const result = classifyBytes({
bytes: new TextEncoder().encode('# T\n'),
totalByteLength: 4,
declaredMediaType: 'Text/Markdown; charset=UTF-8',
filename: null,
})
assert.equal(result.class, 'markdown')
assert.equal(result.mismatch, false)
})
})
describe('isomorphic module', () => {
// The renderer loads all three modules (index.cjs re-exports the other two).
for (const file of ['fileClassifier.cjs', 'fileReference.cjs', 'index.cjs']) {
it(`${file} uses no Node-only API, so the renderer can load it`, () => {
const source = fs.readFileSync(path.join(__dirname, file), 'utf8')
assert.match(source, /module\.exports/, `the ${file} source was not read`)
assert.doesNotMatch(source, /\bBuffer\b/)
assert.doesNotMatch(source, /require\(['"]node:/)
assert.doesNotMatch(source, /require\(['"](?:crypto|fs|path|zlib)['"]\)/)
})
}
})