You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Stop committing generated/*. Once the regeneration triggers + gateway are wired, generated artifacts (e.g. generated/k9iser/*.k9) should be .gitignored and produced on-trigger, not hand-committed (the rot that started this — k9iser#8 / idaptik#77). Blocked on sub-issues 1 + 2.
Sub-issue 1 (iseriser scaffold roll): PARTLY DONE as of this session — see iseriser#12 (DRAFT PR, Refs standards#89). The scaffold now emits *-regen.yml and the unified adapter for NEW repos. The 28 existing -iser repos are NOT yet migrated (correctly: estate-wide mutation is blocked until sub-issue 2 is done).
Sub-issue 2 (gateway tier-2 production-wiring): OPEN, ~8–12 weeks, external-scale. Until done, regen triggers are fire-and-forget loopback. Gitignoring generated/* BEFORE this is wired would break consumers (no regeneration path exists externally). The blocker is real.
Conclusion:#93 remains correctly gated. Do NOT gitignore/delete committed generated/* now.
Inventory of hand-committed generated/* across the estate (read-only sweep)
tlaiser: tlaiser-mcp cartridge (not yet a boj-server cartridge)
wokelangiser: wokelangiser-mcp cartridge (not yet a boj-server cartridge)
alloyiser: alloyiser-mcp cartridge (not yet a boj-server cartridge)
abi/*.h / README.adoc: hand-generated headers/docs (not from any -mcp cartridge)
Prepared .gitignore stanza (apply THE MOMENT sub-issue 2 unblocks)
# generated/* — produced by boj-server -mcp cartridges on push via *-regen.yml.# DO NOT commit these. If missing locally, trigger regeneration:# curl -X POST http://boj-server.local:7700/cartridges/k9iser-mcp/invoke \# -d '{"repo": "hyperpolymath/<repo>", "tool": "k9_generate"}'# Standards#93 / #89 sub-issue 1+2 must be wired before this line is safe.generated/k9iser/
generated/tlaiser/
generated/wokelangiser/
generated/alloyiser/
generated/abi/
Prepared regenerate-on-trigger contract
Each repo's *-regen.yml (already scaffolded into new repos via iseriser#12) is the regeneration trigger. The workflow:
Fires on push to main/master and workflow_dispatch
Checks for the -iser.toml manifest presence
POSTs to boj-server/cartridges/<name>-mcp/invoke with {"tool": "<name>_generate"}
The cartridge regenerates generated/* and commits back (or pushes a PR) via the BoJ REST API
Unblock condition: http-capability-gateway tier-2 production-wired (ADR-0004, standards#89 sub-issue 2, ~8–12 wk). When that ships:
Apply the .gitignore stanza above to all 10 repos (one PR per repo or one batch PR)
Remove committed generated/* files (they will be regenerated on next push)
Verify all regen triggers fire successfully against the real gateway endpoint
The two blockers cited in the May 19 comment remain OPEN:
standards#90 (Roll unified-gated-adapter + SSE + regen-trigger into the iseriser scaffold) — still OPEN
standards#91 (http-capability-gateway tier-2 production-wiring (ADR-0004)) — still OPEN
Until #91 ships, regen triggers are fire-and-forget loopback. The 66 hand-committed generated/* files across 10 repos (inventory in May 19 comment) cannot be gitignored without breaking consumers. Issue remains correctly gated.
Next step (re-verified 2026-08-27): generated/ is tracked on origin/main in ~15 repos (conflow 8, echidna 7, panll 6, januskey 6, laniakea 4, kea 4, boj-server 3, plus singles). Only reposystem does it right — .gitignore carries generated/* with the regen provenance comment, leaving generated/README.adoc tracked.
→ Replicate reposystem's gitignore+regen pattern across the ~14 remaining repos. Small per repo, gated on #90's regen trigger. Both provenance links in the body are dead: hyperpolymath/k9iser 404s and idaptik is now metadatastician/IDApTIK.
Measured today (2026-09-19) across the actual -iser fleet — the outcome this issue asks for is largely already true, and its own blockers have moved.
I searched the estate for -iser repositories and found 27 (affinescriptiser, anvomidaviser, atsiser, betlangiser, bqniser, chapeliser, dafniser, eclexiaiser, ephapaxiser, futharkiser, halideiser, idrisiser, julianiser, lustreiser, mylangiser, nimiser, oblibeniser, otpiser, ponyiser, tlaiser, typedqliser, wokelangiser, and others) — consistent with #89's "28 -iser generators".
Committed files under generated/: the issue wants these to stop being hand-committed.
count
repos with nogenerated/ at all
26 of 27
repos still committing generated/*
1 — proven-servers (2 files, .gitignore does not cover them)
So the rot this issue was filed about no longer exists at scale. The residual is one repo and two files.
Blockers re-checked: this issue says "Blocked on sub-issues 1 + 2" — standards#89 / #90. Those are still open, and the regeneration-trigger half of the ask (produce on trigger) is theirs, not this issue's.
Suggested disposition: the "stop committing generated/*" half is effectively delivered; either close this and let #89/#90 own the trigger wiring, or narrow it to proven-servers + whatever iseriser still emits. Both of the repos its preamble names (k9iser#8, idaptik#77) no longer exist in the org, so the original framing cannot be re-verified as written.
Activity
hyperpolymath commentedon May 19, 2026
Blocker verification + prepared execution contract (session 2026-05-19)
Blocker confirmed real
Sub-issue 1 (iseriser scaffold roll): PARTLY DONE as of this session — see iseriser#12 (DRAFT PR,
Refs standards#89). The scaffold now emits*-regen.ymland the unified adapter for NEW repos. The 28 existing -iser repos are NOT yet migrated (correctly: estate-wide mutation is blocked until sub-issue 2 is done).Sub-issue 2 (gateway tier-2 production-wiring): OPEN, ~8–12 weeks, external-scale. Until done, regen triggers are fire-and-forget loopback. Gitignoring
generated/*BEFORE this is wired would break consumers (no regeneration path exists externally). The blocker is real.Conclusion: #93 remains correctly gated. Do NOT gitignore/delete committed
generated/*now.Inventory of hand-committed
generated/*across the estate (read-only sweep)66 tracked files across 10 repos:
boj-servergenerated/abi/boj_catalogue.h,generated/alloyiser/boj_server.als,generated/alloyiser/run-analysis.shidaptikgenerated/k9iser/*.k9(3 files),generated/wokelangiser/compliance-report.txt,generated/wokelangiser/consent/(3 files)reposystemgenerated/README.adocstapelngenerated/README.adoc,generated/alloyiser/stapeln.als,generated/alloyiser/run-analysis.shburblegenerated/k9iser/*.k9(4),generated/tlaiser/*.{cfg,tla,PlusCal.tla,sh}(12),generated/wokelangiser/(4)gossamergenerated/README.adoc,generated/tlaiser/*.{cfg,tla,PlusCal.tla,sh}(12),generated/wokelangiser/(4)typed-wasmgenerated/abi/README.adocpanllgenerated/k9iser/*.k9(6 files)echidnagenerated/abi/*.h(5 C headers),generated/alloyiser/echidna.als,generated/alloyiser/run-analysis.shrb56-workgenerated/README.adocGenerator breakdown:
k9iser: k9iser-mcp cartridge (blocker = sub-issue 2 gateway)tlaiser: tlaiser-mcp cartridge (not yet a boj-server cartridge)wokelangiser: wokelangiser-mcp cartridge (not yet a boj-server cartridge)alloyiser: alloyiser-mcp cartridge (not yet a boj-server cartridge)abi/*.h/README.adoc: hand-generated headers/docs (not from any -mcp cartridge)Prepared
.gitignorestanza (apply THE MOMENT sub-issue 2 unblocks)Prepared regenerate-on-trigger contract
Each repo's
*-regen.yml(already scaffolded into new repos via iseriser#12) is the regeneration trigger. The workflow:pushtomain/masterandworkflow_dispatch-iser.tomlmanifest presenceboj-server/cartridges/<name>-mcp/invokewith{"tool": "<name>_generate"}generated/*and commits back (or pushes a PR) via the BoJ REST APIUnblock condition: http-capability-gateway tier-2 production-wired (ADR-0004, standards#89 sub-issue 2, ~8–12 wk). When that ships:
.gitignorestanza above to all 10 repos (one PR per repo or one batch PR)generated/*files (they will be regenerated on next push)🤖 Generated with Claude Code
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#45)
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#127)
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#38)
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#67)
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#90)
ci(gitignore): ignore generated/* artefacts (Refs standards#93) (#69)
ci(gitignore): ignore generated/* artefacts (Refs standards#93)
hyperpolymath commentedon May 27, 2026
Residue verified 2026-05-27 — leaving OPEN.
The two blockers cited in the May 19 comment remain OPEN:
Roll unified-gated-adapter + SSE + regen-trigger into the iseriser scaffold) — still OPENhttp-capability-gateway tier-2 production-wiring (ADR-0004)) — still OPENUntil #91 ships, regen triggers are fire-and-forget loopback. The 66 hand-committed
generated/*files across 10 repos (inventory in May 19 comment) cannot be gitignored without breaking consumers. Issue remains correctly gated.hyperpolymath commentedon Aug 27, 2026
Next step (re-verified 2026-08-27):
generated/is tracked onorigin/mainin ~15 repos (conflow8,echidna7,panll6,januskey6,laniakea4,kea4,boj-server3, plus singles). Onlyreposystemdoes it right —.gitignorecarriesgenerated/*with the regen provenance comment, leavinggenerated/README.adoctracked.→ Replicate reposystem's gitignore+regen pattern across the ~14 remaining repos. Small per repo, gated on #90's regen trigger. Both provenance links in the body are dead:
hyperpolymath/k9iser404s andidaptikis nowmetadatastician/IDApTIK.fix(templates): STATE template must not tell mints to default to PMPL
hyperpolymath commentedon Sep 19, 2026
Measured today (2026-09-19) across the actual
-iserfleet — the outcome this issue asks for is largely already true, and its own blockers have moved.I searched the estate for
-iserrepositories and found 27 (affinescriptiser,anvomidaviser,atsiser,betlangiser,bqniser,chapeliser,dafniser,eclexiaiser,ephapaxiser,futharkiser,halideiser,idrisiser,julianiser,lustreiser,mylangiser,nimiser,oblibeniser,otpiser,ponyiser,tlaiser,typedqliser,wokelangiser, and others) — consistent with #89's "28-isergenerators".Committed files under
generated/: the issue wants these to stop being hand-committed.generated/at allgenerated/*proven-servers(2 files,.gitignoredoes not cover them)So the rot this issue was filed about no longer exists at scale. The residual is one repo and two files.
Blockers re-checked: this issue says "Blocked on sub-issues 1 + 2" —
standards#89 / #90. Those are still open, and the regeneration-trigger half of the ask (produce on trigger) is theirs, not this issue's.Suggested disposition: the "stop committing
generated/*" half is effectively delivered; either close this and let #89/#90 own the trigger wiring, or narrow it toproven-servers+ whateveriseriserstill emits. Both of the repos its preamble names (k9iser#8,idaptik#77) no longer exist in the org, so the original framing cannot be re-verified as written.