[JS SDK] uploadFile uses chunked transfer encoding, causing 501 NotImplemented on S3 presigned PUT URLs

[g-veritas]

Bug Report

Summary

When building a template using the JS SDK in a self-hosted deployment, uploading files via .copy() fails with 501 NotImplemented from S3. The root
cause is that uploadFile sends the tar stream with Transfer-Encoding: chunked, which S3 presigned PUT URLs do not support.

Environment

• e2b JS SDK: v2.18.0 (latest)
• Node.js: v22
• Storage: Self-hosted with S3-compatible storage (presigned PUT URL returned directly to SDK)

Error

501 NotImplemented
A header you provided implies functionality that is not implemented

Root Cause

In dist/index.js, uploadFile passes a Node.js Readable stream directly as the fetch body:

const res = await fetch(url, {
  method: "PUT",
  body: uploadStream,  // Node.js Readable, no Content-Length
  duplex: "half"
});

Since Content-Length is unknown, Node.js undici falls back to Transfer-Encoding: chunked. S3 presigned PUT URLs reject chunked requests with 501
NotImplemented — this is an S3 protocol-level restriction that cannot be resolved via IAM or bucket policy.

Comparison with Python SDK

The Python SDK (v2.19.0) already handles this correctly — it buffers the tar archive into io.BytesIO first, then uploads bytes via httpx, which sets
Content-Length automatically:

tar_buffer = tar_file_stream(...)
response = client.put(url, content=tar_buffer.getvalue())  # bytes, not stream

┌───────────────────┬────────────────────┬───────────────────┐
│                   │       JS SDK       │    Python SDK     │
├───────────────────┼────────────────────┼───────────────────┤
│ Body type         │ ReadableStream     │ bytes             │
├───────────────────┼────────────────────┼───────────────────┤
│ Content-Length    │ Not set            │ Set automatically │
├───────────────────┼────────────────────┼───────────────────┤
│ Transfer-Encoding │ chunked            │ fixed-length      │
├───────────────────┼────────────────────┼───────────────────┤
│ S3 presigned PUT  │ 501 NotImplemented │ Works             │
└───────────────────┴────────────────────┴───────────────────┘

Suggested Fix

Buffer the stream into memory before uploading, consistent with how the Python SDK works:

// Before
const res = await fetch(url, {
  method: "PUT",
  body: uploadStream,
  duplex: "half"
});

// After
const _buf = await new Promise((resolve, reject) => {
  const chunks = [];
  uploadStream.on("data", (chunk) => chunks.push(Buffer.from(chunk)));
  uploadStream.on("end", () => resolve(Buffer.concat(chunks)));
  uploadStream.on("error", reject);
});
const res = await fetch(url, {
  method: "PUT",
  body: _buf,
  headers: { "Content-Length": String(_buf.length) }
});

This is a 5-line change and aligns the JS SDK behavior with the Python SDK. Happy to submit a PR if helpful.

---

[linear]

ENG-3584