Hallo everyone! Apologies if this isn't quite the right channel, but I just published a new plugin, nf-runinsights, for cross-run benchmarking of Nextflow pipelines. registry.nextflow.io/plugins/nf-runinsights github.com/QuietFlare/nf-runinsights Nextflow's built-in reports only describe a single run. This plugin remembers your runs and compares them. Every run writes its per-process metrics into a local history store (a shared directory or S3 works too, no server needed) and each new run gets checked against the older ones. It flags regressions and tries to name the likely cause, like queue wait vs actual execution, a container change, or plain input growth. It also spots over-provisioned resources and pipeline-wide slowdowns. It only observes, so it can't change or fail a run. There is also a small dashboard with no dependencies and an MCP server, so you can ask your AI assistant things like "why was last night's run slow?" Trying it is one config line

plugins { id 'nf-runinsights@0.1.0' }

So far I've only validated it on local runs of nf-core/sarek. If you run the same pipeline repeatedly on a cluster or in the cloud, I'd love to hear whether the findings hold up at that scale and what's missing. Thank you in Advance!

I like this idea

Thank you @Maxime Garcia Happy to hear on how we could make it useful for pipeline maintainers, like catching performance changes between the runs!

Hi, does this plugin need internet access?

Hi,The core plugin works fully offline, so no internet access is needed. It only reads pipeline's own trace data and writes JSON files to a local directory, there are no outbound calls at runtime. The internet is needed to download the plugin from the registry, or for fully offline setups you can use the plugin zip. The only feature that needs internet is the optional AI narration, which is off by default, and also pointing the history store at s3:// obviously needs network too. The bundled dashboard and MCP server are localhost only.

sounds interesting, thank you 🙂