Voulons-nous ajuster les liens « suggérés » et « connexes » en bas des publications ?

[Falco]

@schneeland

Among the various AI features that Discourse offers, related topics is actually one of the two (the other is AI-based SPAM classification) I would be willing to turn on, if we ever get to the point, where we get affordable vGPUs or vNPUs and self-hosting the LLM becomes an option.

Running https://hf.co/Qwen/Qwen3-Embedding-0.6B with GitHub - huggingface/text-embeddings-inference: A blazing fast inference solution for text embeddings models should be good doable in a server with 2-4GB RAM without a GPU.