Morning crew.
Lots of code work this weekend trying to beat it into a GA worthy release.
In fact a LOT of codework. RIP my Claude budget. is Pro really cheaper? Maybe - we may never know…
After all that - I’ve currently got 2 active ‘issue’ cases on the board and both do not appear to be from tools, rather configurations. RC2 Code has been stabile on multiple systems for most of the weekend.
That means… I’m locking RC2 feature complete, looks like ive cleared the RC2 gates so I’m going to tag it on main as RC2 and pull together a release. (Release Candidate as a release, go figure…) and then I’m on to working on GA.
That means you saw a flurry of activity Saturday as I was mucking about in the core tools - Labels getting some code normalization (thanks Tes) FileCab learned some really cool (necessary) tricks as well as tightening some edge cases we found. And the indexer pipe got a fresh coat of paint and completed my planned GA feature set AND a UAT pass. (That immediately got bugged and fixed because I forgot the main case and only tested the edges - that’s fixed too…)
So lets look at what this brings to the table.
- Bunch of quality-of-life stuff.
- added more check/restart endpoints… An agent if given access has a tool to HA check/ ha restart, reload automations, scripts, all yaml, etc thencheck the logs. Basically call down updates or trigger a patch, validate nothing broke that will hose your install and restart or reload the affected components then checkk the logs and report back… Very useful if you hook a test agent to your live HA install… Just sayin…
- Tools to manage and attempt to repair bogus cabinets (see highlander up there…) and stacks (a set of cabinets will now start showing as a stack in the docs… Libraries, stacks? - ok I guess I’m old now.)
- Final GA pass through the index pipeline and:
Made Inspect now deliver the attached device data (all of it - parent children and sibs) for an ent if available and called (your LLM can shape the output and the smallest digest style is default) If I did it right your agent can read that stuff in the proxmox integration and see what your stack looks like just by tagging the ents. (it’s pretty cool they do the parent child info stuff right in the device tree - Im using it as the model)
Well that feeds the indexer - so I had to walk that contract all the way back to the top of Index and add all those options. Meanwhile I moved the drawer explosion down into Inspect where it should have been all along setting up for security trimming and redaction because it happens in exactly ONE place.
and…
- Fixed the drawer explosion I broke in the last index push - oops.
- Actually made Regex work in ZQ1
- Fixed the break in ZQ1 where if your ai used it even a little wrong Poof - nothing.
- moved some things around structurally to make the computation return less expensive. (cause I’m about to have to start computing a secmask)
- added the first plumbing for id tool stubs
This sets up some really cool stuff for post GA if you’re following along in the code. For now, it’s faster leaner meaner more info rich searches.
In short, the Index got a LOT Lighter and more effective.
I also tuned up filecab, and the summary pipeline to remove redundancy and be less wordy. Most of the tools have a default that leads to a summary or digest or context only mode - and THATs what is / will be used in prompt gen.
If you see files versioned 4.2.1 or greater it’s got the GA treatment now. I’ll be continuing to work to a GA release on branches that split off and merge back into main. (Yes, I’m trying to swerve the repo into a normal devops cadence/maintenance style.)
All in all good weekend - but for now I have a paying job to do. 
Pull code, read docs, submit bugs, have fun!