Description:This plugin adds cross-run benchmarking to Nextflow. It records per-process metrics for every run into a durable history store and compares each new run against previous runs of the same pipeline, reporting what changed and why.
Key features:
Trace observer records task duration, peak memory, queue wait, CPU efficiency, I/O volume, retries, and container id per process
Deterministic findings: regressions with likely causes (queue wait vs execution, container change, input growth), pipeline-wide slowdowns, over-provisioned memory, retry storms
One JSON file per run, safe for concurrent runs on shared filesystems and compatible with s3:// via Nextflow's filesystem layer, so teams share a history with no server
Includes a zero-dependency local dashboard and a read-only MCP server for querying run history from AI assistants; optional AI report narration, off by default
This benefits the Nextflow community by filling a gap: built-in reports describe one run in isolation, while this plugin gives individuals, core facilities, and HPC users regression detection and resource right-sizing evidence across runs. Validated against nf-core/sarek.
License: Apache 2.0. QuietFlare is my personal GitHub organization. I am its sole owner and the plugin's author.