Hi all, Just an idea I had a while ago, when upda...
# infrastructure
k
Hi all, Just an idea I had a while ago, when updating some of the rather central nf-core components, e.g aligners. Do we have something like a knowledge base / graph for nf-core components? Could also be expanded to pipelines... E.g. to answer a few of these questions: - a given module_X inputs depends on module_Y outputs in tests. - changing inputs/outputs in module_A will lead to changes in module_A, subworkflow_B and test_C. - Over time a given module / subworkflows is used less and less by other nf-core components / pipelines so it could be licensed for depreciation. - If an agent designs some subworkflows / pipelines, it could be aware of nf-core component dependencies, reading those components from the official repo in every session could be quite inefficient. I'm an obsidian heavy-user and since LLMs utilizing md files as token efficient solutions, I was thinking of creating a knowledge base for nf-core components, with a particular focus on interdependencies. Regarding the design, I think it would be good to brainstorm a bit first, a) whether it's worth to develop such a DB and b) what might be some critical design choices. Cheers, Kevin
p
I think that this idea has been proposed before. Maybe even implemented? 🤔
I'm AFK right now but worth having a search on slack
k
Thanks Phil! I had a look but didn't find any discussion about explicitly creating a knowledge graph / db for modules and subworkflows. But could also be me, not knowing what to search for 😓
k
Perfect! Thank you Rike
p
That was the one! Thanks @Rike Hanssen !
k
I'll ping Joao, I have a few ideas to drive this a bit further. Thank you two for the feedback 🙂
m
Note for me. https://raw.githubusercontent.com/nf-core/website/main/public/components.json is the graph of nf-core components (from João)