Header 'Transfer-Encoding' in Google Cloud Run

[xgeek95]

Hello, 

Has anyone tested or worked with the response header : 'Transfer-Encoding' to obtain it as a response header in a service deployed in Cloud Run ?

I can't find it in my API response headers, if I create any other custom header I can get it.

Backend : NestJS

Best regards

[robertcarlos]

Based on this Google Cloud Blog: 

Cloud Run can now run gRPC services with server-streaming RPCs and send partial responses in a single request—in addition to existing support for unary (non-streaming) RPCs.

You could test this sample gRPC server application provided from the blog that you can test to stream responses from the Cloud Run-based serverless application.

[AlexTheOne]

Hi guys,

I have the same issue, please give me some clue to debug what is wrong.

Backend: Nest.js (10.4.7)

Cloud: Cloud Run behind Google Frontend

Code part:

@Post('/')
async chat(@Body() chatRequestBody: TheType, @Res() response: express.Response) {
  // ... some secret code, pretty much data processing

  // Manually handle the response as a streaming text endpoint
  response.setHeader('Content-Type', 'text/plain');
  response.setHeader('Transfer-Encoding', 'chunked');
  // Force Google Frontend to enable chunked response
  response.setHeader('x-goog-response-info', 'disable-response-buffering');
  response.setHeader('Cache-Control', 'no-store, no-cache, must-revalidate, proxy-revalidate');
  response.removeHeader('Content-Length');
  response.flushHeaders();

  try {
    for await (const chunk of chatResponse.partialObjectStream) {
      logger.debug(`Received chunk: ${JSON.stringify(chunk)}`);
      response.write(JSON.stringify(chunk));
    }
  } catch (error: any) {
    response.status(400).json({ type: 'error', error: error.message });
    return;
  }

  response.end();
}

What I see from logs from Cloud Run is that each chunk is processed correctly, so I see something like:

Received chunk: {"chat":["Okay, I've reviewed the"]}
Received chunk: {"chat":["Okay, I've reviewed the message here is what I think"]}

But client gets response without `Transfer-Encoding: chunked`

Basically this:

vary	Origin
access-control-allow-credentials	true
x-request-id	***
content-type	text/plain
x-goog-response-info	disable-response-buffering
cache-control	no-store, no-cache, must-revalidate, proxy-revalidate
X-Cloud-Trace-Context	***
Date	Tue, 14 Jan 2025 19:06:01 GMT
Server	Google Frontend
Content-Length	1980
Alt-Svc	h3=":443"; ma=2592000,h3-29=":443"; ma=2592000

---

@robertcarlos wrote:

 Google Cloud Blog: 

---

So, according to the post (from October 7, 2020), it should work right, why it does not?

Thanks.

[AlexTheOne]

I wrote then server-side events implementation, but it still returns one big chunk.

So according to this article I am using WebSockets right now, but what about this streaming approach? It should be working fine, it is just normal valid web approach.