Trois questions indépendantes, tranchées lors du branchement de Kadans sur
MinIO (2026-07-26) : qui déclare les buckets d'une app, qui détient les
identifiants capables de les créer, et comment l'app lit les siens.
La décision de fond est du fondateur : CHACUN SON PÉRIMÈTRE. Une application
déclare ses buckets depuis son propre dépôt ; `tools` fournit le serveur, un
module de standardisation et un compte de provisionnement — pas la liste. Une
première version faisait tout porter par l'infra partagée : à ce rythme, chaque
bucket de chaque app devenait une PR sur le dépôt commun.
L'ADR consigne aussi les trois identités et leurs portées (root / provisionneur
/ compte de service), pourquoi la lecture des identifiants est une propriété
inconditionnelle de la plateforme plutôt qu'une déclaration par app, et pourquoi
les octets ne transitent pas par l'API — avec les conséquences que ça impose
(endpoint public, CORS aux origines exactes, pas de basic-auth sur l'ingress S3).
Les alternatives écartées sont listées avec leur motif, dont deux que j'avais
moi-même proposées et qui étaient plus faibles.
Deux limites assumées y figurent : le provisionneur est un secret PARTAGÉ entre
rôles CI (sa compromission permet de créer des buckets, pas de lire des objets),
et les noms d'actions d'administration MinIO n'ont pas été éprouvés contre le
serveur au moment d'écrire.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01CoafGWmRVESaWX819USUUA
L'app url-shortener était en SyncError permanent : son PVC live porte un
spec.volumeName épinglé (rebind du volume Longhorn après le drill coupure de
courant) absent du chart ; chaque sync tentait donc de le vider, refus API
(spec immuable après création), échec en boucle malgré automated+selfHeal.
- apps.yaml : passthrough générique ignoreDifferences + syncOptions par app.
- values.yaml : url-shortener ignore /spec/volumeName du PVC, avec
RespectIgnoreDifferences=true pour que l'apply réinjecte la valeur live au
lieu de la vider (le cas d'usage documenté d'ArgoCD pour les champs
immuables).
Rendu helm vérifié : seule l'Application url-shortener change.
Co-Authored-By: Claude Fable 5 <[email protected]>
Le chapitre « service compagnon » montrait `vaultConnectionRef: default` dans
l'exemple VaultAuth — c'est faux hors du namespace `tools` et ça a réellement bloqué
le déploiement de kadans-api (pods en CreateContainerConfigError, VaultDynamicSecret
sur « VaultConnection default not found »).
VSO résout vaultConnectionRef dans le namespace DU CR ; la VaultConnection `default`
ne vit que dans `tools`. Les apps hors `tools` (erp, webapp) OMETTENT le champ et
laissent VSO retomber sur sa defaultVaultConnection. On retire donc la ligne de
l'exemple + on ajoute un encart WARNING dédié au piège.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_013ws8L74dVZmp97Wu36fm8j
Le runbook new-web-app couvre l'app autonome (dépôt/base/Vault/namespace propres,
tout nommé <app>). Il manquait le cas du SERVICE COMPAGNON : un second service qui
partage le namespace — et parfois le stack Vault/DB — d'une app existante (API cœur
à côté de son front, façade d'analyse). Deux précédents vivants non documentés :
kadans-jobs (namespace seul) et kadans-api (namespace + base + Vault).
- Nouvelle page 09-service-compagnon.md : compagnon vs app autonome ; les deux
formes (sans état / partage Vault+DB) ; le PIÈGE du VaultAuth manquant quand
l'app primaire ne consomme pas Vault (front statique) → le compagnon pose son
propre VaultAuth mais avec le rôle+SA du PRIMAIRE ; carte, précédents, delta de
checklist.
- 07-argocd-register.md : ajoute la ligne `namespace:` aux options (elle existait
dans values.yaml — kadans-jobs — mais n'était pas documentée) ; corrige le
callout qui affirmait le namespace « non configurable ».
- conventions.md : note l'exception compagnon à la règle « tout est <app> ».
- README.md : entrée 09 dans l'index + Last Updated.
Vérifié : VaultAuth erp nommé `auth` ; connexion via pgbouncer.tools ; liens
internes tous résolus.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_013ws8L74dVZmp97Wu36fm8j
Ajoute l'API cœur au registre gitea_applications. ArgoCD crée une Application
`kadans-api` (source arcodange/kadans-api, path chart, targetRevision HEAD), sync
automatique prune+selfHeal, image-updater par digest sur :latest — même moule que
les autres apps.
Namespace `kadans` (comme kadans-jobs) : kadans-api partage le stack Vault/DB déjà
en place pour l'app front (VaultAuth `kadans`, rôle Postgres dynamique
postgres/creds/kadans, ServiceAccount `kadans`, policy KV `kadans`). Aucun nouvel
iac/DB/Vault à provisionner.
À merger APRÈS le fix chart kadans-api (VaultAuth + hôte DB pgbouncer.tools) pour
que la première synchro ArgoCD parte d'un chart correct.
helm template rend l'Application kadans-api → repoURL arcodange/kadans-api,
namespace kadans, CreateNamespace, digest.
Co-Authored-By: Claude Opus 4.8 <[email protected]>
Claude-Session: https://claude.ai/code/session_013ws8L74dVZmp97Wu36fm8j
The never-yet-applied k3s_dns.yml placed 'import /etc/coredns/custom/*.server'
INSIDE the .:53 server block. *.server files hold full server blocks
(arcodange.lab:53 {…}), which only parse at Corefile root — inside a block
CoreDNS dies at startup with "Unknown directive 'arcodange.lab:53'"
(CrashLoopBackOff, cluster DNS fully down; lived it on 2026-07-24 while
restoring the expired *.arcodange.lab certificate).
Also restores the stock 'loadbalance' plugin dropped by the playbook.
Context: cluster CoreDNS forwarded to the node's resolv.conf, which lists the
ISP box's IPv6 RDNSS next to the Pi-holes — NXDOMAIN roulette for *.lab names.
That's what left step-issuer unable to reach ssl-ca.arcodange.lab:8443 and let
the 24h wildcard cert expire this morning. The (fixed) playbook pins .lab
resolution to the Pi-holes via the coredns-custom ConfigMap; applied live on
2026-07-24, wildcard renewed, strict TLS verified on gitea/argocd/grafana.
Co-Authored-By: Claude Fable 5 <[email protected]>
Incident 2026-07-23: an uncapped nuxt generate (3.5G RSS) on pi1 starved the
k3s control-plane and traefik (load >150, no swap, no OOM-kill) — every
*.arcodange.lab endpoint went dark, Gitea included, while Gitea itself was
healthy on pi2. Job containers are spawned via the host docker socket, so
cgroup caps on the job container are the only guardrail.
Applied live on pi1+pi3 via 03_cicd.yml on 2026-07-24 (both runners
re-registered; pi3's runner was down and is back in service).
Co-Authored-By: Claude Fable 5 <[email protected]>
kadans-jobs is the tier-2 (homelab, 24/7) piece of the Kadans topology: a job
queue plus the store of published analysis results, decoupling the product from
the volatile Mac worker.
org: arcodange — the repo does not live under the default arcodange-org.
Digest-based image-updater annotations follow the fleet pattern; the
cluster-wide ImageUpdater CR (namePattern "*", useAnnotations) picks them up,
so there is no per-app CR to maintain. The image is already in the registry.
No postgres/iac/terraform.tfvars entry, on purpose: the façade runs a memory
store in v0, so it needs neither a database nor Vault. That is the runbook's
degraded mode — the DB, the Vault JWT role and the app's own iac/ will land
together with the Postgres store.
Chart realigned on the runbook conventions first, in kadans-jobs#1.
Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_013ws8L74dVZmp97Wu36fm8j
Co-authored-by: Gabriel Radureau <[email protected]>
Co-committed-by: Gabriel Radureau <[email protected]>
- qa-strategy › Independent verification: with Mistral (vibe -p,
mistral-medium-3.5) and Ornith 35B admitted to verifier duty by verdict
parity (erp#63 evidence, blind-judged), the independent verifier SHOULD be
a different model family than the builder; journal records which family
verified what.
- STATUS: #63✅ (PR erp#69, harness home erp:fleet/harness/), #56✅
(PR erp#68, authored by the Mistral builder bench), #39 built on local
branch (push+PR = operator step), PR-log rows, P3 flipped to in-progress.
Paired with erp#69 (Closes erp#63).
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
Ajouté par bda53f29 pour « monter les credentials dans le repo-server »,
mais ces values sont passées au chart argocd-image-updater (HelmChart
kube-system) qui n'a pas de clé repoServer : no-op intégral. Le vrai
repo-server ArgoCD est déployé par l'addon k3s et n'a pas besoin de ce
montage — les credentials repo passent par les secrets étiquetés
argocd.argoproj.io/secret-type.
Co-Authored-By: Claude Fable 5 <[email protected]>
Ces champs (commit 7aa789d1) n'existent pas dans le schéma Application
d'ArgoCD : l'API server les élague à l'apply, d'où un diff permanent →
factory OutOfSync en boucle (296 tentatives selfHeal) sur kadans,
telegram-gateway et dance-lessons-coach. L'authentification aux repos
privés passe par un secret repo-creds (label
argocd.argoproj.io/secret-type: repo-creds, url préfixe
https://gitea.arcodange.lab/arcodange) — corrigé côté cluster sur le
secret gitea-credentials existant.
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator direction 2026-07-15: the orchestration layer itself (builder
sessions, cold verifiers) must run on Mistral or hermes+Ornith/MLX too.
The protocol already carries everything in files+issues; new model-fleet
section defines the evidence-gated ladder — verifier roles migrate
first (cross-family refutation is stronger verification), scoped
builders benched on unchanged acceptance gates, Claude default until
the bench says otherwise. D2 row records the direction; spike = erp#63.
Co-Authored-By: Claude Fable 5 <[email protected]>
The resume-protocol fresh-reader test (context-free subagent) passed
on substance (picked erp#38, correct first command, skipped the
human-gated erp#46) and surfaced two doc gaps: milestone due dates
were only on the forge (rule says order by due date), and nothing
arbitrated one-session-one-lane vs orchestrated fan-out. Both fixed;
#54 map entry now mentions the ADC register.
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator ask 2026-07-12: an ADR-equivalent for accounting so method
choices are consistent AND justifiable. Accounting scatters this across
permanence des méthodes (PCG 121-5), the annexe, the organisation doc
(PCG 911-3) and audit position memos; the ADC unifies them as one
lightweight versioned record: MADR-lite + base légale/effective-dates/
annexe-impact fields, immutable once Accepted (supersede = the
permanence principle made structural), fiscal.yaml rules cite their
ADC (écriture → règle → ADC → base légale in four hops), annexe
generated from the register, acceptance human-only. Seeds adc-001..007
from decisions already made this exercice; two new obligation-table
rows (121-5, 911-3); expert-comptable agenda updated.
Co-Authored-By: Claude Fable 5 <[email protected]>
- STATUS resume protocol: milestones ordered by due date, skip
human-gated tops, named entry points (erp#38 / #51 / write-skill
quartet); every issue now carries an Execution footer.
- Backlog map: +erp#59 (T14 split from #48), +erp#60 (T11 loop split
from #54), retitles, post-replay markers.
- D9 meeting capture parked (nice-to-have; calls are iPhone-first).
- prd_check.py preserved from the session scratchpad into scripts/
(the closure protocol references the pattern — now it's runnable).
Co-Authored-By: Claude Fable 5 <[email protected]>
Diarization and Google Calendar sync are both on Hyprnote's free plan,
which satisfies the two operator criteria at once; Meetily (MIT,
diarization in the community core) stays as OSS fallback with sb.py
ICS-matching to compensate its missing calendar sync. Gate: quality
judged on a real bilingual call before the lane is trusted (erp#49).
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator has no Granola account (proprietary, paid, cloud
transcription — misfit with the vault doctrine). The delivery-agents
backlog line now specifies the local transcription lane: Whisper-class
model on the M4 + Ornith summary, as an sb.py job.
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator insight 2026-07-11: the second brain's client dossiers
(01-Projects/Clients/<client>/) are where the back-office fleet meets
the sales chain and the future delivery agents working on client
projects. Knowledge-layer bullet: agent-owned deposit files (never
edits of human notes), each family feeds and reads the dossier
(billing state + contract facts in; dunning tone, contract scope,
new-business sightings out — the KM 4% settlement clause makes
delivery observations a billing input); vault routing doctrine
(local-first) for confidential client content. Backlog updated with
the 2026-07-11 CRM audit facts (propal/contrat/projet modules empty,
KM retainer terms only in WIP JSON — deferred schedule = T06's source
of truth) and the third family (delivery agents; Granola→vault ingest
as an sb.py job; REX → Mode B evidence → CMS).
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator addition to the closure protocol: the QA gate is held by an
independent verifier subagent — context-free, prompted to REFUTE, repo
+ issues + journals as its only inputs; verdict posted on the milestone,
unresolved refutation blocks. New qa-strategy section extends no-self-
grading to POC exit gates and autonomy promotions (verdict attached to
the artifact it gates), mirroring at process level what the pipelines
do at data level (dual extraction, seeded-wrong fixtures).
Co-Authored-By: Claude Fable 5 <[email protected]>
Operator requirement: the resume protocol's twin. A milestone closes
only through an 8-step sweep that OPENS with a QA gate (acceptance
criteria re-verified with evidence, test suites green — golden-set
bars, injection fixtures, linter suites incl. seeded-wrong provenance
FAILs, idempotency replay; catalog status flips backed by proving-
protocol evidence; can't pass QA -> shed scope, don't close) and ENDS
with doc-surface QA (link/anchor check, deprecation grep of the claims
the milestone retired, fresh-reader smoke test by a context-free
subagent). In between: STATUS flip, roadmap re-baseline (diamonds never
move), no-tombstone truth-pass of leaves, AGENTS.md/SKILL.md/guidebook
sweep, outward loop (ADRs, memories, second-brain REX). Reader's half:
trust order live system > code > STATUS > leaves > memories; stale
Last Updated = suspect. Summarized as a house rule in factory
AGENTS.md, canonical checklist in the PRD STATUS.
Co-Authored-By: Claude Fable 5 <[email protected]>