lab_bench

A lab bench: a place to run and preserve measurement experiments about how Soliplex installations actually behave.

Nothing here ships. This repository deliberately releases no software, and it is deliberately separate from any repository that does. Its purpose is reproducibility: an experiment run six months ago should still be findable, re-runnable, and readable.

Companion repository

soliplex/lab_harness holds the reusable machinery -- the soliplex-lab-harness package. It is ordinary released software with an ordinary branch model.

This repository holds only the experiment-specific parts: the jigs built on that harness, and the results. A jig depends on a pinned harness version; it never vendors one.

The branch model

branch	holds	merges?
main	praxis documents only	PRs to main change praxis, nothing else
set/<name>	an experiment set and its measurement jig(s)	never merged to main
jig/<set>/<topic>	in-progress jig work	merges into its set/ branch
docs/<set>/<topic>	a set's documentation, including its SET.md	merges into its set/ branch
exp/<set>/<slug>	one experiment: local jig bits, data, results	never merged anywhere

set/ and exp/ branches are the archive. They are never merged and never deleted. See PRAXIS.md.

Finding things

There is no index file. The issue tracker is the index:

• every experiment has an issue, created from the experiment template, that names the branches it uses
• every set has one too, created from the set-proposal template: a set is proposed and accepted rather than branched by hand, and accepting it is what creates set/<name>, its scaffold, and its label
• issues carry a label mirroring their branch kind -- exp:<set>, jig:<set>, docs:<set>, set:<set>, praxis -- so a set's listing separates experiments run from work on the apparatus, and both from what the set is for
• a status: label says where a proposal stands -- proposed, accepted, declined -- which is a different question from what kind of work it is

So the list of experiments for a set is a label query, and it is always current because nobody has to remember to update it.

Start here

• PRAXIS.md -- how and when to create a set, a jig, and an experiment; what to record; what may not be committed here
• AGENTS.md -- rules for AI coding agents, which differ by branch and are not what an agent would otherwise assume