Stanislav

From my experience wvegas is best for wireless(3g/4g etc..) connections using MPTCP, also MPTCP dev's don't recommend using different congestions controls on client/server, they always say that they must be...

Have you each tested connections rtt ? (ping) and also mptcp capable tests (on mptcp status page)? Difference in ping can cause big impact on results.

That can cause very huge impact from my experience, on even 2 LTE modems, if 1 gets 20mb/s and second 5-10 but with 3 times more ping, result would be...

I was doing tests for example currently with 4 LTE modems, installed in vehicle and in city if all 4 have good signal/RTT i'm pushing it to limit (i'm using...

> > I was doing tests for example currently with 4 LTE modems, installed in vehicle and in city if all 4 have good signal/RTT i'm pushing it to limit...

Same problem here :( - latest rpi os bullseye, and no vout_drm. Also during configure i get warning: > WARNING: Option --enable-hwaccel=h264_v4l2request did not match anything

I also noticed that in 'Openai chat node' on first step output is: > { > "response": { > "generations": [ > [ > { > "text": "", > "generationInfo":...

[My_workflow.json](https://github.com/user-attachments/files/18916681/My_workflow.json) Missed your message 😔, i'm using v1.79.3, workflow attached. Sorry for 'spam', tried to use VLLM instead of sglang, got some other qdrant error, but flow is kind of...

SGlang has VLLM core, they both have openai compatible API, currently it looks like there is problem with "stream": true option that n8n sends, because if output from LLM is...

I have done more tests, looks like its not problem with streaming, for some reason n8n is not able to parse 'content' array.