I’ve needed to brush up on my webhosting skills for an unrelated work project, and have access to some older GPU-enabled computers, so I’ve started a little project where I host a couple of new-ish bioinformatics tools that anyone can use. I’ll be putting out a few of these in the coming weeks but the first one I wanted to talk about was Metappuccino (paper here, though it looks like oxford university press is down right now??) - a light LLM that uses free-text info in SRA records to try and fill in missing metadata tags, as well as assigning a confidence score to those tags. You can give it a try at metappuccino.biobench.ai if you have some SRA data you’d like use.

I’m hoping to get a bit of load on the server so I can test if my load handling is working as intended, but if you have any feedback on the tool itself I’m happy to hear it (just know that I’m not affiliated with the authors at all - I just have servers, have been frustrated by SRA metadata issues in the past, and needed to host something, so I decided on this).

I’m also looking for some ideas/inspiration for other tools to stand up. The only real requirements are that it shouldn’t depend on end-users uploading big proprietary data - I’m just a rando on the internet so I don’t want to ask people to trust me with private things. I’m not too constrained by hardware, so if there’s something out there you wanted to check out but didn’t have the resources to run, I’m happy to take a look.