The effective-model read surfaces (Networking, Dossier/Anatomy, Storage,
and Compose Doctor) avoid environment, label, and command values, but
docker compose config resolves any ${VAR} interpolation before the model
is parsed, leaving no provenance. A secret interpolated into a structural
field (a bind path, network name, published port, or extra_hosts entry)
is therefore returned resolved. That value is already readable at the
same stack:read scope through the stack's files, so this documents the
caveat rather than changing behavior.
- Reword the "secret-safe / never shows a secret value" claims on the
Networking, Dossier, Storage, and Doctor docs pages, and add a
canonical note to the Environment and Secrets Guardrails page steering
secrets to environment:/env_file: injection.
- Make the matching code comments honest in effectiveAnatomy,
composeNetworkInspector, effectiveModel (extra_hosts), and the
effective-anatomy route.
* feat: flag documentation drift in the stack dossier
Warn in the Dossier tab when a port written into a stack's access_urls is not published by the stack's compose, so operator documentation stays aligned with what Sencho can observe.
The check is deterministic, read-only, and frontend-only: it compares ports parsed from access_urls against the published ports the Anatomy panel already shows, never interprets prose, and stays quiet for port-less URLs, scheme-default ports (:80/:443), and ports published through a variable, to avoid false positives. Community tier, no gating.
* test: pin doc-drift handling of bare hosts and mixed variable ports
Make two deterministic-drift behaviors intentional and regression-proof after review: a scheme-less single-label host (plex:32400) is not checked, since it cannot be told apart from a plain note (add a scheme to opt in), and a stack mixing a variable-published port with fixed ports suppresses the whole check. Adds tests for both, a clarifying code comment, and a docs note with the http:// workaround. No behavior change.
* feat(fleet): export a whole-fleet Markdown dossier
Add an admin-only "Export Dossier" action to the Fleet view that walks
every node and stack, pairs each stack's generated Compose anatomy with
its operator notes, and downloads a folder-structured homelab-dossier.zip
(index, per-node and per-stack pages, plus fleet-wide port, volume,
network, env, access-URL, and VLAN/firewall maps).
Reuses the existing stack dossier and anatomy Markdown generators by
extracting the shared Compose parsers into a frontend lib module.
Unreachable nodes are recorded with a reason and never block the export;
only env variable names and counts are ever emitted, never values.
* fix(fleet): unique stack slugs and reproducible dossier archive
Disambiguate stack names on one node that slugify to the same value (e.g.
`Web` and `web` on a case-sensitive host) with a per-node slug map shared by
the node-page links and the file emission, so neither overwrites the other.
Pin a fixed entry timestamp on the zip so the archive bytes are a pure
function of the file map rather than the wall clock.
* feat(stacks): add Stack Dossier tab with operator notes and Markdown export
Add a Dossier tab beside Anatomy and Activity on the stack detail panel. It
shows a read-only summary auto-derived from the stack's Compose anatomy
(services, ports, volumes, network, restart policy, env file, source) plus an
editable form for the context Sencho cannot infer: purpose, owner, access URLs,
static IP, VLAN, and firewall, reverse-proxy, backup, upgrade, recovery, and
custom notes.
Notes persist per stack and per node in a new stack_dossiers table, reached
transparently through the remote-node proxy so a remote stack's dossier
round-trips to the node that owns it. Reading a dossier needs stack read
permission; saving needs stack edit. The tab exports a single Markdown document
combining the generated facts and the operator notes, with copy-to-clipboard and
download actions; env values are never exported, only variable names and counts.
Available on all tiers. The standalone anatomy copy-as-Markdown shortcut is
removed since the Dossier export supersedes it.
* fix(stacks): gate dossier reads/writes on stack existence; clear dossier on node delete
A dossier endpoint validated the stack name but not that the stack exists, so an
editor could PUT a dossier for a name with no stack, leaving an orphan row that a
later stack of the same name would inherit. Require the stack to exist (existing
requireStackExists guard) on the dossier GET and PUT, returning 404 otherwise.
Also clear a node's stack_dossiers rows when the node is deleted, alongside the
other node-scoped cleanup, so removing a node leaves no orphan dossiers.
Docs: scope the "no secret exported" statement to the generated facts (which only
ever carry variable names and counts) and clarify that operator notes are
exported exactly as written.
Anso