Files
sencho/docs/features/stack-file-explorer.mdx
Anso b1decbb32a docs: v1 docs refresh (batch 7) (#1627)
* docs(introduction): refresh screenshots and correct stale nav/tab coverage

Replace all 5 screenshots with fresh 1920x1080 production captures.
Document the shipped Take down stack action, the Compose Labels stack
tab, and the Docker Labels Fleet tab, none of which were mentioned.
Correct the Console nav item to note it is a limited-availability
surface rather than a plain role/tier-gated view.

* docs(quickstart): refresh screenshots and correct preflight, dashboard, and menu drift

Replace all three quickstart screenshots with fresh captures and correct
several claims that drifted from the current UI:

- Document the environment preflight's 7th check (Sencho compose
  location), previously missing entirely from both the text and the
  screenshot alt text.
- Add the Stack health table's Source and Port columns, and note that
  columns are sortable and default to load order.
- Note the masthead's running-container count, live CPU/memory readout,
  and alert count.
- Fix the profile menu list: replace the nonexistent "Feedback" entry
  with "Open New Issue", drop "Appearance" (it lives in Settings, not
  the profile menu), and add the conditional "Billing" entry.
- Note that the sidebar groups stacks by Docker Compose label once
  labels are assigned, with pinned stacks first.

* docs(configuration): document missing env vars and fix JWT secret wording

Add SENCHO_UPLOAD_DIR, TRIVY_CACHE_DIR, SENCHO_MESH_RECONCILE_INTERVAL_MS,
SENCHO_MESH_PROXY_TUNNEL_IDLE_MS, SENCHO_COMPOSE_COMMAND_TIMEOUT_MS, and
SSO_OIDC_CUSTOM_ENABLED, all real environment variables that were missing
from the reference tables. Clarify that the JWT signing secret has no
environment variable at all, generated and stored in the database only,
rather than implying an unused JWT_SECRET var exists.

* docs(stack-management): refresh against current live app and source

Rewrites the Stack Management page against the production node and
frontend source: adds the Compose Labels anatomy tab (new since the
last refresh), the Mute action on the stack header and sidebar
context menu, corrects the update-available banner wording and the
sidebar context menu's lifecycle ordering, notes the Doctor tab's
severity dot and the scan-status banner, and replaces all 14
screenshots with fresh production captures.

* docs(editor): refresh against current live app and source

Replace all 8 screenshots with fresh 1920x1080 production captures (mobile
shot at phone viewport). Document the self-stack protection dialog that
blocks deploy/delete actions on the stack running the current Sencho
instance, the multi-container summary strip and Compact/Detailed density
toggle, the mutually exclusive Expand containers / Expand logs controls,
the Files tab full-screen toggle, and the post-deploy scan-status banner
on the Anatomy panel, none of which were previously covered.

* docs(editor): recapture mobile compose screenshot without redaction

The previous mobile screenshot used the plex stack, whose compose volume
mount includes a real host path that had to be blacked out and overlaid
with placeholder text, leaving a visible seam. Recapture against the
dozzle stack instead, whose only volume is the Docker socket, so the
screenshot needs no editing.

* docs(stack-file-explorer): refresh page against current app and code

Renames "Files & Volumes tab" references to the current "Files" tab
label, documents the full-screen toggle and word-wrap control, expands
the non-browsable-volume reasons to match the current containment
logic (single-file binds, the full protected host-path list, and the
Docker-unreachable named-volume case), documents the 100-item bulk
selection cap and the 5000-file/1 GiB archive limits, corrects the
non-existent DISK_FULL error code, notes that override compose files
are unprotected, and scopes the atomic-write claim to fs-backed roots.
Replaces all 9 screenshots with fresh production captures taken in the
Files tab's full-screen mode, so no other stack UI (health metrics,
logs) appears in the background.

* docs(resources): refresh against current live app and source

* docs(dashboard): refresh against current live app and source

* docs(app-store): refresh against current live app and source

Replaces all 6 screenshots with fresh production captures and corrects
the Environment Variables and About-panel metadata claims to match what
the bundled LinuxServer.io registry actually returns today.

* fix(docs): recapture app-store advanced-tab screenshot without scroll cut-off

The prior capture was taken mid-scroll to reveal the Security checkbox,
leaving the template logo and About text cropped awkwardly at the top.

* docs(global-search): refresh against current live app and source

* docs(deep-links): refresh against current live app and source

- Add App Store, Logs, Update, Console, and Audit to the URL table (previously undocumented)
- Document hub-only URL redirect behavior and cross-link to Multi-Node Fleet
- Note Fleet tab URL segments do not always match their on-screen label (Status/configuration, Map/dependencies)
- Document the no-env-files edge case (Env tab absent, falls back to compose)
- Clarify Settings section-list state is phone-only; desktop always normalizes to /settings/appearance
- Note which top-level views share identical URLs between desktop and phone
- Remove greenfield-violating temporal phrasing ("now has", "as today")
- Replace em dashes in touched bullet list

* docs(appearance): refresh against current live app and source

* docs(stack-activity): refresh against current live app and source

Add the missing "Stack taken down" event category, cross-link the
Suppressed badge to notification mute rules, list the new Compose
Labels tab in the Anatomy panel strip, and replace both screenshots
with current production captures.

* docs(dossier): fix generated-facts wording, add rollback readiness section

Corrects three Dossier tab Generated Facts rows against current frontend
behavior: the Ports count is not filtered to host-published mappings, the
Network row's "bridge" label is fixed text rather than a derived driver,
and missing env var names appear only in the Markdown export, not the
live tab. Adds a Rollback Readiness subsection under Connected Features
(previously only an orphaned Limitations bullet), links to Compose
Networking and Storage Portability, and clarifies that a missing export
section can mean either an older node build or a failed on-demand fetch,
not only a version gap. Recaptures all three screenshots against
production.

* docs(stack-drift): refresh against current live app and source

Fixes the stack detail tab bar description (Compose Labels tab was
missing, Files/Edit compose were wrongly listed as tabs), documents
that re-check only requires read access to the stack (so viewer and
auditor roles can trigger a ledger write), and cross-links the network
findings to the Networking tab's runtime drift section and the Fleet
Overview's Drift filter. Replaces all four screenshots with fresh
1920x1080 production captures.

* docs(compose-doctor): refresh against current live app and source

Corrects the rule count from 31 to 32 and documents the previously
undocumented self-managed-stack guard rule and enforcement, the
per-browser dismiss control for the summary card and Doctor tab dot,
the acknowledged summary state, the full Health-Gated Updates verdict
mapping, and the temporary exposure intent option. Replaces all four
screenshots with production captures.

* docs(compose-networking): refresh against current live app and source

Adds cross-links to the new node-wide Networking operator page, documents
the "view node networking" link and the unset/inherit exposure-intent pill
labels, corrects the stale Resources Hub network-tab reference, and
replaces all screenshots against the production node.

* fix(docs): unbreak stack-activity page render

The screenshot alt text used backslash-escaped quotes, which is
invalid MDX/HTML attribute syntax. Mintlify failed to parse the file
and silently dropped it from routing, so the page 404'd despite being
listed in docs.json navigation.

* docs(environment-guardrails): refresh against current live app and source

Fix the ENV FILES section description: it lists env_file: entries and
project files with an existence problem, not every declared env file.
A cleanly-resolving project file is already named in the Project
Environment File panel above it. Replace all three screenshots with
current production captures.

* docs(docker-label-audit): refresh against current live app and source

Add production screenshots (page had none), a Capability gating
section, a Limitations section, and a Troubleshooting accordion.
Tighten the secret-redaction heuristic description and cross-link
Stack Labels, Compose Doctor, Environment Guardrails, and Node
Compatibility.

* docs(compose-storage): refresh against current live app and source

Captured fresh production screenshots (both prior images were stale)
and fixed a leftover pre-rename card title pointing at the Files tab.

* docs(stack-labels): refresh against current live app and source

Removed the stale trailing-dot sidebar claim (that rendering was
removed from the sidebar in a prior UI pass), documented the new
label-scoped notification muting available from the sidebar, stack
context menu, and Settings panel, corrected the Fleet Actions card
copy to match the redesigned cards, added the previously undocumented
bulk-assign size cap and the node-scoped label bulk-action API, and
replaced all 8 screenshots with fresh production captures.

* docs(sidebar): refresh against current live app and source

Removes the stale per-row label dot claim (removed from StackRow), corrects the update indicator and Updates chip color from orange to fuchsia, and documents Mute/Take down in the context menu, the label group mute kebab, Ctrl+A/Esc bulk-mode shortcuts, pin-eviction toast, and the offline-node skip behavior in cross-node search. Replaces all 8 screenshots with fresh production captures.

* docs(deploy-progress): refresh against current live app and source

Corrected the Scan entry point (node-wide Scan this node from Security
Overview, not a per-stack config scan), added the missing Take down
verb and entry point, broadened recovery actions to cover restart and
rollback failures, fixed the modal status text and raw-output color
claims to match the live UI, and replaced all screenshots with fresh
captures from a throwaway demo stack on the production node.

* docs(atomic-deployments): refresh against current live app and source

* docs(health-gated-updates): refresh against current live app and source

* docs(deploy-enforcement): refresh against current live app and source

Correct the tier note to every-tier (verified against the current
policy gate, which no longer has a paid-only blocking switch), add the
Security Overview deploy-enforcement summary card, tighten the honor-
suppressions default wording, cross-link the separate pre-deploy scan
advisory dialog to avoid confusion with the block dialog, and replace
all three screenshots against the current production UI.

* docs(blueprint-model): refresh against current live app and source

Replaced all 11 screenshots with fresh production captures. Corrected
the volume-destroying-drift Enforce-downgrade claim (appeared three
times): that code path has no call sites in the runtime reconciler, so
any compose edit on a stateful or state-unknown blueprint always
re-enters Awaiting confirmation regardless of drift mode. Corrected
the Create workflow (creation does not trigger an immediate
reconciliation tick) and the Delete workflow (now requires typing the
blueprint name to confirm).

* docs(scheduled-operations): refresh against current live app and source

* docs(blueprint-model): drop tier framing, treat as Community-native

Blueprints carry no tier gate, so "available on every tier" implied a
comparison that doesn't exist. Removed the tier framing from the intro
note, the Prerequisites table, and the Security section; the role
requirement (admin to write, every role to read) already says
everything that matters.

* docs(auto-update-policies): refresh against current live app and source

Documents that every update trigger on this page, Apply now and
scheduled Auto-update tasks alike, runs through the same atomic backup,
build-aware rebuild, and post-update health gate as a manual update
from the stack editor, and is subject to the same deploy enforcement
policy gate. Corrects the readiness-computation description to scope it
to registry-image services and cross-links to Health-Gated Updates,
Atomic Deployments, Deploy Enforcement, and Stack Management. Replaces
the readiness board screenshot with a fresh production capture and
removes three orphaned screenshots left over from the page's prior
CRUD-policy layout.

* docs(auto-heal-policies): refresh against current live app and source

Correct the admin-only prerequisite: viewing policies and history is
open to every signed-in role, only creating/toggling/deleting requires
admin. Rewrite the matching troubleshooting entry to match, note the
overlap behavior between an All-services and a named-service policy on
the same container, and replace all three screenshots with a real
production policy.

* docs(webhooks): refresh against current live app and source

Replaced all three screenshots with fresh production captures (single
webhook create/reveal flow, two-card configured list), verified every
claim against WebhookService/StackOpLockService/registry.ts, and added
the previously undocumented per-stack lock skip behavior for
non-Git-source-sync actions with a matching troubleshooting entry.

* docs(global-observability): refresh against current live app and source

* docs(audit-log): refresh against current live app and source

- Fix nav terminology: Audit is a top navigation tab, not a sidebar tab.
- Correct retention claim: Admiral shows full retained history (default
  90 days), not a fixed 14-day window; the 14-day clamp only applies to
  the Community API, which has no navigation entry point at all.
- Fix settings path: Settings - Operations - Data Retention (previously
  pointed at the now-split Developer Diagnostics section).
- Restore current Recovery Vault naming (was stale Sencho Cloud Backup).
- Expand the tracked-actions list: stack take-down, label management,
  MFA reset, and notification suppression rules were missing.
- Note that assigning the Auditor role itself requires Admiral.
- Replace all four screenshots with fresh production captures (Stream,
  Table, expanded row, Data Retention) reflecting the current five-row
  retention card.

* fix(audit-log): reframe data-retention screenshot to card content only

The prior crop included the Settings sub-navigation panel, which isn't
part of what the surrounding text describes. Retake tightly cropped to
match the page's other screenshots (card content only, no chrome).

* docs(multi-node): refresh against current live app and source

Rewrites the Pilot Agent enrollment flow (now a generated compose.yaml
plus docker compose up -d, not a single docker run command) and the
Settings scope table (current registry: Stacks, Container Alerts,
Docker & Storage, Data Retention, Mute Rules, Recovery, and the
Admiral Account / Recovery Vault renames). Adds a Sencho Mesh
cross-link matching the live add-node form copy. Replaces all eight
screenshots with sanitized production captures.

* docs(pilot-agent): refresh against current live app and source

* docs(readme): refresh GitHub README against current live app

Replace all 9 screenshots (stale top nav showing a removed Console
item and missing the new Networking page); document Take down,
Drift Detection, Environment and Secrets Guardrails, Storage
Portability, Docker Label Audit, Remote Updates, and the node-wide
Networking dashboard; fix a broken non-root-user anchor link; correct
the notification channels list (no email channel exists yet) and the
global search scope (pages, nodes, and stacks, not containers or
services); rename "scan policy packs" to the current "scan policies"
terminology; broaden the App Store bullet to cover custom registries.

* docs(readme): reposition intro copy and refresh badge row

Lead with DevOps/platform/sysadmin framing instead of homelab-first,
drop pre-1.0 "production" language, state plainly that Sencho
provides a UI instead of promotional "does the work you do" phrasing,
and call out that multi-node was part of the architecture from the
start. Swap the Discussions badge for CodeQL, last-commit, open
issues, and live website/docs status badges, and add a blog link.

* docs: retake Dashboard and Global Search screenshots for accuracy

Configuration Status now shows Recovery Vault instead of the stale
Cloud Backup label, and the search palette capture includes the
Networking page entry added after the prior screenshot batch.

* docs: refresh Fleet View page against current app

Fleet View has drifted since its last rewrite: the Docker Labels tab,
Export Dossier action, Networking filter/badge, node label pills and
latency in topology, the Policy sync status row, and the node card
Mute submenu were all shipped but undocumented. The Check Updates and
Add Node actions also moved into the Overview toolbar (renamed Node
Update / Manage Nodes) and no longer sit in the shared action row.

Replaces all five screenshots with fresh production captures and
corrects the masthead's motion description (a calm shimmer on Healthy,
a steady glow on Degraded/Critical, not a pulsing dot). Also fixes a
Fleet Sync limitation that claimed no first-party sync-status panel
exists, now that the Status tab surfaces one.

* docs: refresh Fleet Dossier page against current app

Adds the page's first screenshot, documents the network exposure summary now included per stack, documents the storage-summary omission versus the Stack Dossier export, and adds two troubleshooting entries.

* docs(fleet-federation): deep rewrite for the rollout-approval gate

Federation's cordon and pin controls no longer take effect by
themselves: the reconciler requires a confirmed rollout preview
(Apply now -> Confirm Apply) before it mutates the fleet, and pinning
or unpinning always clears a blueprint's approval. Rewrote the mental
model, step-by-step, lifecycle table, security section, limitations,
practical workflows, troubleshooting, and cross-links to reflect that
gate. Also corrected the audit-log claims: cordon/uncordon/pin are not
filterable by a node.cordon/blueprint.pin action taxonomy, only by the
free-text search box against the method, path, and summary.

Replaced 4 screenshots and added 3 new ones (production fleet plus an
isolated instance to capture the rollout preview dialog in both a
safe and a blocked state), and fixed one stale pin-workflow sentence
in blueprint-model.mdx that the same audit surfaced.

* docs(fleet-federation): drop tier-availability framing

The feature isn't tier-gated, so calling out "every tier" reads as an
unnecessary advertisement. Removed the tier note and the Community/
Admiral prerequisite row, and reworded two role-visibility sentences
that had conflated tier with role.

* docs(fleet-actions): deep rewrite for the v2 action-card redesign

Rewrites the page against the shipped Fleet Action Card redesign: a
single cyan rail with per-card action-class chips instead of the old
per-card rose/purple/amber rails, plus the live blast-radius preview,
dry-run, and confirmed-target binding mechanics that replaced the old
static warning banners. Documents two previously-undocumented
endpoints (match-preview, prune/estimate), the preview-confirm-execute
model shared by all three cards, per-node locking, timeouts, and the
version-gated confirmed-target contract for mixed-version fleets. All
five screenshots recaptured against the live production UI.

* docs(fleet-actions): fix nested quote in screenshot alt text

* docs(fleet-sync): refresh against current app and document proxy gap

Retake all screenshots against the live fleet (control authoring view,
a genuine replica showing the read-only banner, and replicated
suppression rows) and document a previously-unwritten behavior: scan
policies, CVE suppressions, and misconfig acknowledgements are fetched
localOnly and are not proxied through the node switcher like most
other Security page tabs, so viewing a remote's Fleet Sync state
requires signing into that instance directly.

* docs(fleet-backups): refresh against current app and note new integration points

Replaces all 6 screenshots with fresh 1920x1080 production captures and
corrects several drifted details: the Cloud Snapshots panel's pagination
and per-item delete action, the exact Recovery Vault storage-mode label,
and the real per-file size-cap behavior (an oversized compose.yaml skips
the whole stack; an oversized .env is dropped but the stack still
captures). Documents two entry points that were missing from the page:
the pre-update snapshot checkbox in Health-Gated Updates and the snapshot
coverage nudge on the Storage Portability tab. Expands the single warning
banner description in the detail view to cover all four banner types,
notes the Snapshots tab and Recovery Vault are hub-only, and adds a
Where Fleet Backups fits table cross-linking the six adjacent features.

* docs(remote-updates): refresh against current app and add hardened-channel notes

Replace all screenshots with fresh 1920x1080 production captures and fix the
Node Updates trigger label (Check Updates -> Node Update). Add the Changelog
tab, correct the reconnect overlay's stale 'Update timed out'/Try Reloading
copy to the current Taking longer than expected/Reload to check text, and
document the Admiral Hardened Build channel's carve-outs (pinned-but-not-blocked,
entitlement-gated local update path). Disambiguate this page from the
unrelated top-level Update (Health-Gated Updates) nav tab.

* docs(node-compatibility): refresh against current app and expand capability list

Replace all screenshots with fresh production captures (node switcher,
capability lock card, connection test panel) and swap the lock-card example
from a now-hidden Host Console to Audit Log, which is directly reachable.
Bring the capability table from 26 to 35 entries to match the current
registry, add a Mute Rules row, and split out fine-grained fallback
capabilities (update-guard, service-scoped-update, cross-node-rbac,
stack-down-remove-volumes) with their non-lock-card fallback behavior.
Fix the compose-networking row, which incorrectly claimed to also gate the
node-level Networking overview. Note that Pilot Agent nodes never advertise
host-console regardless of version.

* docs(security): refresh against current app and document scan-node launcher

Replaces all screenshots with fresh production captures, documents the
Scan this node launcher and the third Scanner setup toggle (pre-deploy
scan advisory), tightens the Compose risks example list and the Policies
block-condition description to match the live app, and adds an On a
phone section.

* docs(vulnerability-scanning): refresh for exploit intel and risk-based policies

Documents exploit intelligence (KEV/EPSS evidence tags), the redesigned
risk-based scan policies (severity/known-exploited/fixable block
conditions replacing the single max-severity field), and the new
Findings badge state. Replaces all screenshots with fresh production
captures and adds a Secrets tab example.

* docs: refresh CVE suppressions page

Retook all screenshots against production (prior ones predated the
triage-status/OpenVEX UI and the edit capability). Documented the
suppression edit flow, which the page previously described as
delete-and-recreate only. Corrected the remote-node banner copy on the
Suppressions tab and added a screenshot of it. Cross-linked the
Misconfig acknowledgements panel that now sits on the same tab.

* docs(two-factor-authentication): refresh screenshots and document rate-limit layers

Replace all 11 non-count-variant screenshots with fresh production captures.
Document the throttled sign-in state precisely (kicker/hero/caption change,
not just the input), that a replayed TOTP counts toward the lockout counter,
and the separate per-network-address sign-in rate limit that sits in front
of the per-account MFA lockout.

* docs(api-tokens): refresh screenshots and document service-scoped deploy actions

Deploy Only now authorizes eight lifecycle POSTs (was six): the per-service
update and restore endpoints were missing from the scope table. Universal
restrictions gained a row for image channel management, which was already
rejecting API tokens in code but undocumented. Replaced all five screenshots
with fresh captures showing the current three-scope create flow and a
populated list with one token of each scope.

* docs(upgrade): refresh against current app and remove legacy-version framing

Fixes migration-list inaccuracies (registry credentials were never
plaintext, unlike node API tokens), rephrases the SSH/TLS and JSON-config
migration bullets to drop version-numbered legacy framing, adds a GHCR
mirror note matching the quickstart pattern, and cross-links the Hardened
Build entitlement-gated update path so digest-pinned installs aren't sent
down the manual docker pull steps.

* docs: refresh Backup & Restore against current backend

Broadens the encryption.key warning to cover everything CryptoService
actually encrypts (node tokens, Git source and Fleet Secrets, SSO/MFA,
Recovery Vault, fleet snapshot contents), not just registry credentials.
Adds the built-in backupData CLI as a no-host-tooling alternative to the
sqlite3 .backup command, cross-linked to Emergency command-line recovery.
Corrects node-token migration guidance: tokens are signed with a secret
stored in sencho.db and remain valid after a host move, so no
regeneration is required.

* docs: refresh Recovery guide against current Settings page and CLI

Documents the Settings · Recovery page's full System health / Environment /
Safe actions / Command-line hub instead of only its environment-preflight
section, adds the missing "Sencho compose location" preflight check, adds the
SSO/OIDC/LDAP lockout scenario now that disableSso.js exists, and corrects the
rollback description (exact UI label, backup-required and image-layer caveats).

* docs(emergency-cli): sync in-app page description, add CLI operational details

Aligns the "in-app Recovery page" summary with the freshly refreshed Recovery
guide (System health / Environment / Safe actions / Command-line recovery, SSO
providers, one-click command download). Adds constraints verified against the
CLI source: password/username minimums, the valid SSO provider identifiers with
an example, and a note that diagnostics.js always reports Docker as unreachable
(it runs without a live Docker connection, unlike the in-app page).

* docs(troubleshooting): refresh against current backend and nav

Corrects several claims that had drifted from the implementation:
network name validation (underscore is not a valid leading character),
the remote-update delayed-failure window (3 minutes, not 90 seconds),
and an unsubstantiated version-compatibility check on the Admiral 403
entry. Removes two entries describing legacy pre-v0.39 node behavior
that no longer applies to any currently shipped build. Replaces stale
"Profile > Settings > X" navigation with the current Settings hub
paths, and "wifi icon" with the current Test Connection label. Adds
Pilot Agent awareness to the remote-offline entry with a cross-link to
its dedicated troubleshooting section, and trims the docker-run
converter entry to point at the fuller, already-current version on the
Stack Management page instead of duplicating it with a broken anchor.

* docs(verifying-images): document the :dev GHCR integration tag

Verified the existing cosign/SBOM/VEX/tag-policy content against
docker-publish.yml and docker-preview.yml; all accurate. Added the
previously undocumented :dev/:dev-<sha> integration tag published on
every push to main via docker-dev.yml.

* docs(trivy-setup): sync install guide with node-scoped scanner and current Trivy upstream docs

Documents that Trivy installs independently per node, adds the TRIVY_BIN
override and non-PATH mounting option, fixes the deprecated apt-key install
flow and RHEL repo gpgkey placement, notes the Exploit intelligence toggle
for air-gapped hosts, and replaces the stale Scanner setup screenshot.

* docs(two-factor-admin): document lockout recovery and self-reset gotcha

Verified the reset flow, CLI recovery, and SSO toggle claims against
current source and replaced the stale admin-reset modal screenshot.
Added the failed-attempt lockout behavior (undocumented despite being
promised in the page description) and a warning about resetting your
own 2FA from the Users list, which signs you out immediately unlike
the self-service disable flow.

* fix(docs): make Settings and Self-hosting discoverable in the sidebar

The Reference and Operations groups used root pages that were only
reachable by clicking the ambiguous group label, and that click also
triggered an unwanted double action (expand plus navigate). List both
pages as explicit sidebar entries and disable global drilldown so group
headers only expand or collapse.

* docs(settings): sync reference page with current Settings Hub

Renamed License to Admiral Account throughout (Hardened Build channel
switch, image channel display, DURATION pill), corrected the password
policy, documented the new Appearance navigation and log-chip-color
controls plus font options, noted Mesh data plane is not on every
installation, fixed the label cap (50, not 100), and replaced every
screenshot with a fresh capture from the production node.

* fix(docs): apply the sidebar toggle-only fix to Start here and Product guide

These two groups had the same click-to-navigate-and-expand pattern as
Reference and Operations. Flattening them is safe here too: the
Product guide root page has no directory listing depending on it, and
the Start here root page's own hand-authored Next steps CardGroup
already covers the same links the auto-generated listing duplicated.

* docs(licensing): refresh licensing page for Hardened Build and sales-led Admiral pricing

Admiral pricing moved from self-serve checkout to a contact-sales model, and
a new Hardened Build image-channel switcher shipped in the Admiral Account
settings page; neither was reflected in the docs. Also documents the
lifetime-license edge case (no Manage subscription button, no Billing row)
and refreshes all four screenshots against the current UI.

* docs(security): sync reference page with current API-token and encryption scope

Verified every claim against the live implementation: the API-token universal
restrictions list was missing MFA management, Recovery Vault, and image
channel management; the encryption-at-rest field list was missing Recovery
Vault credentials; added a note on the password strength indicator's
recommended-vs-enforced distinction. Refreshed the SSO settings, API tokens,
and audit log screenshots against the production node.

* docs(settings): fix password policy and session-invalidation claims

Cross-checked against auth.ts while verifying the same claims on the
security reference page: the enforced minimum is 8 characters (the
12+/mixed-case/number text is a frontend strength hint, not a validated
rule), and a password change invalidates every other session for the
account rather than leaving them valid.

* docs(contact): sync contact channels with sales-led pricing and current support gating

Replace the retired contact@sencho.io with hello@sencho.io (the address actually
used in the website footer and the pricing page's Get in touch CTA), narrow the
licensing@sencho.io scope to existing-license activation/billing/refunds now that
new Admiral conversations route through hello@sencho.io, drop the stale LICENSE-file
and in-app upgrade-prompt claims (the AGPLv3 relicense removed both), correct the
Support settings path and add the published response-time targets, and remove the
unsubstantiated bug bounty mention.
2026-07-21 09:13:12 -04:00

313 lines
26 KiB
Plaintext

---
title: Files & Volumes
description: Browse and safely edit the config files inside a stack's mounted volumes, plus the files in the stack's own directory, from the dashboard.
---
Files & Volumes gives you direct access to the files your containers actually read at runtime: the configuration inside a stack's **mounted volumes** (bind mounts and named Docker volumes), as well as the files in the stack's own source directory. A typical use is editing the Nginx and proxy configuration inside a reverse proxy's `/config` volume without leaving the dashboard. It lives on the **Files** tab inside the stack editor, alongside the dedicated `compose.yaml` and `.env` editors.
<Frame>
<img src="/images/stack-file-explorer/overview.png" alt="Files tab open with a populated nginx configuration tree on the left and default.conf open in the editor on the right" />
</Frame>
<Note>
Browsing is limited to the roots the stack itself declares: its source directory and the volumes mounted by its services. You cannot browse other stacks or arbitrary locations on the host.
</Note>
## Opening the Files tab
1. Click any stack in the left sidebar.
2. Click **Edit compose** in the right-hand Anatomy panel header to open the editor.
3. Switch to the **Files** tab.
The **Files** shortcut button next to **Edit compose** opens the editor with the **Files** tab already selected. The shortcut only appears when your account has **stack read** permission on the stack; without it, the tab is not offered.
Files & Volumes is available on every Sencho tier. Read actions (browse, preview, download, inspect permissions) require **stack read** permission, which every signed-in role has by default. Write actions (upload, edit, create, rename, move, change permissions, delete) require **stack edit** permission on your account.
## File roots: Volumes and stack source
The left pane opens with a **Browsing** selector that lists the roots Sencho discovered for the stack, grouped into **Volumes** and **Stack source**. When the stack declares at least one browsable volume, the explorer defaults to it; otherwise it opens on the stack source.
Each volume root is labelled by its container path (for example `/config`) and the type of mount behind it:
- **Bind mounts** map a host directory into a service (for example `./config:/config`). When the directory is reachable by Sencho, the root is fully browsable and editable.
- **Named volumes** are Docker-managed volumes. Sencho browses and edits them through a short-lived, locked-down helper container, so they work even though their on-disk location is not directly visible to the dashboard.
The same source mounted by more than one service is shown as a single root, summarised as `N mounts`. A root that any service mounts read/write is editable; a root that every service mounts read-only is browse-only.
<Note>
Editing a config file does not restart the service that reads it. After you save, restart the relevant service so it picks up the new content.
</Note>
### When a volume cannot be browsed
A volume root shows a short explanation in place of its tree and is not browsable when:
- the bind mount points at a host path Sencho cannot reach (mount that path into the Sencho container to browse it),
- the mount targets a single file rather than a directory (a bind that maps one file straight into the container has nothing to open a tree on),
- the mount targets a protected host location: the filesystem root, kernel/OS state (`/etc`, `/proc`, `/sys`, `/dev`, `/run`, `/var/run`), the base-image binaries and libraries Sencho's own runtime depends on (`/usr`, `/bin`, `/sbin`, `/lib`, `/lib64`, `/boot`, `/root`), or the Docker socket,
- the mount overlaps an area Sencho manages at runtime: its own application directory, its data directory, the OS temp directory, or a configured upload/scanner directory (browse the owning stack's source instead), or
- a named volume cannot be resolved: either Sencho could not find it on the node that owns the stack, or it could not reach Docker to check (a transient daemon or proxy issue, distinct from the volume simply not existing).
### Named-volume editing details
Named-volume editing is best-effort and bound by file ownership. The helper container writes as an unprivileged user, so a file owned by `root` or a specific application user may not be writable; Sencho returns a clear permission error in that case. New files created in a named volume are owned by the helper user, while edits to an existing file keep that file's owner and mode. Named-volume saves are written in place and are not atomic, so a save interrupted mid-write can leave the file partially written; the files involved are small config files, which keeps the window tiny.
## Layout
The Files tab splits into two panes. The left pane holds the Browsing selector, the upload affordance, the **New file** and **New folder** buttons, and the directory tree for the selected root. The right pane is the action bar plus the file viewer.
A full-screen toggle sits next to **Close editor** whenever the Files tab is active. It collapses the identity, health, and log column on the left of the stack workspace so the tree and viewer use the full width, which helps when you are working with wide config files or a deep directory tree. Click it again (or switch to a different tab) to return to the normal two-column layout.
<Frame>
<img src="/images/stack-file-explorer/full-screen-mode.png" alt="Files tab in full-screen mode, with the Command Center column collapsed so the tree and viewer fill the width" />
</Frame>
## Browsing the directory tree
Folders sort before files, and entries within each group sort alphabetically.
Click a folder to expand or collapse it. Click a file to open it in the viewer on the right. Symlinks render with a chain icon and behave like files when clicked. Deleting a symlink removes only the link entry; the file it points to is untouched.
The tree is fully keyboard navigable. Tab into it and use the arrow keys to move between rows: **Up** and **Down** move row to row, **Right** expands a folder (then steps into it), **Left** collapses it (or steps out to the parent), **Home** and **End** jump to the first and last visible row, and **Enter** or **Space** opens a file or toggles a folder.
Each row is fully clickable across the pane, so right-clicking anywhere on a row (not just on its name) opens that entry's context menu. Long names are never truncated: the tree scrolls horizontally so you can read the full name.
**Display cap.** Each directory render is capped at 1000 entries. A folder with more than 1000 children shows the first 1000 alphabetically with a footer noting how many entries the directory holds in total. The tree also has a filter input at the top of the list so you can narrow a large directory to the entries you care about without dropping to a shell.
## Protected files
Protection applies to the **Stack source** root. On a volume root a file named `compose.yaml` or `.env` is an ordinary config file: it opens directly in the viewer and has no delete restriction, because it is the application's own file rather than the stack's compose definition.
On the stack source root, the amber dot in the tree marks the five canonical stack files: `compose.yaml`, `compose.yml`, `docker-compose.yaml`, `docker-compose.yml`, and `.env`.
<Frame>
<img src="/images/stack-file-explorer/protected-tree-marker.png" alt="Stack source tree with an amber dot marker next to compose.yaml" />
</Frame>
Two behaviours follow from the marker:
- **Dedicated tab redirect.** Clicking `compose.yaml`, `compose.yml`, or `.env` jumps you to the matching **compose.yaml** or **.env** tab so the save-and-deploy controls stay in front of you. `docker-compose.yaml` and `docker-compose.yml` are still flagged as protected, but they open in the regular file viewer because they are not the canonical Sencho file.
- **Delete is blocked at the stack root.** The five canonical files at the stack root cannot be deleted through the explorer: the delete is rejected. Remove the whole stack via Stack Actions instead. A same-named file nested in a subdirectory is an ordinary file and can be deleted after the type-to-confirm step.
An override file such as `compose.override.yaml` or `docker-compose.override.yml` is not one of the five canonical names, so it carries no amber dot and no delete restriction: Sencho treats it as an ordinary file even at the stack root.
## Viewing files
When you click a text file, its contents appear in the editor on the right. The Monaco editor auto-detects the language from the extension. A **word wrap** toggle in the viewer toolbar wraps long lines instead of scrolling horizontally; it defaults on and your choice is remembered across files and sessions.
| File type | Behaviour |
|-----------|-----------|
| Text file up to 2 MB | Rendered inline with syntax highlighting. |
| Text file over 2 MB | Panel with the filename, size, and a **Download** button. |
| Binary file | Same panel layout, label is `Binary file`. |
## Editing and saving
When you have stack edit permission, the editor opens in write mode. The toolbar shows the filename and a **Save** button that activates as soon as there are unsaved changes.
When you do not, the toolbar shows a `Read-only` chip and the editor refuses input.
<Frame>
<img src="/images/stack-file-explorer/viewer-edit-mode.png" alt="Viewer toolbar showing nginx.conf with an unsaved edit and the Save button now active" />
</Frame>
Click **Save** to write the file to disk. If you have unsaved edits and click a different file in the tree, a confirmation dialog asks whether to discard them; cancelling keeps you on the current file.
Saves use optimistic concurrency: the editor remembers when the file was last loaded, and a save that targets a stale version returns a "file changed elsewhere" notice with the current server-side content. Your typed buffer is preserved so you can review the new version and reapply your edits before saving again.
Writes to the stack source root and to bind-mount volumes are atomic at the filesystem level: Sencho stages the new content into a sibling temporary file, fsyncs, and promotes it via rename, so a crash or power loss never leaves a half-written target on disk. Named-volume writes go through the helper container instead and are not atomic; see [Named-volume editing details](#named-volume-editing-details) above.
<Warning>
Editing a file does not restart any containers. If your stack reads the file at runtime (for example, a config file mounted as a volume), restart the relevant service after saving so the container picks up the new content.
</Warning>
## Creating files and folders
The toolbar **New file** and **New folder** buttons at the top of the tree create an entry in the currently selected directory (the parent of the file you have open, or the stack root if nothing is open). Use the toolbar buttons to create an entry directly at the stack root, where there is no folder row to right-click.
Right-click any folder for **New File** and **New Folder** entries that scope to the right-clicked folder. These write controls appear only when your account has stack edit permission.
<Frame>
<img src="/images/stack-file-explorer/new-file-dialog.png" alt="New file modal scoped to the nginx folder, with the file name field populated and a Create button" />
</Frame>
Filenames cannot be empty, cannot contain `/` or `\`, and cannot be `.` or `..`. The Create button stays disabled until the input passes validation.
Creating a file never overwrites an existing one. If a file of that name already exists in the target directory, the dialog reports inline that the name is taken; if a folder of that name exists, Sencho rejects the creation with a message. Either way the existing entry is left untouched.
## Uploading files
The dashed **Upload file** affordance at the top of the tree opens a file picker. You can also drag and drop a file onto the dashed zone; the border lights up in the brand colour when a file is hovered over a valid drop target. Uploads are one file at a time.
| Limit | Value |
|-------|-------|
| Maximum file size | 25 MB. Uploads above the limit return a 413 error with `TOO_LARGE`. |
| Target directory | The currently selected directory, or the stack root if no file is open. |
| Same-name files | A confirmation dialog asks whether to replace the existing file. Cancelling keeps the original. |
The upload affordance is hidden for users without stack edit permission.
<Tip>
For bulk transfers or files above 25 MB, use `scp` or `rsync` from your workstation directly to the stack directory on the host.
</Tip>
## Downloading files
When a file is selected, the right pane action bar shows **Download**. Files stream straight to your browser. Files that exceed the inline preview limit also expose a Download button inside the oversized-file panel itself.
## Renaming
Right-click any file or folder and choose **Rename**. The dialog accepts a new name following the same rules as creation. Rename keeps the entry in its current folder.
Rename appears only when your account has stack edit permission.
## Moving
Move a file or folder into a different directory in one of two ways:
- **Move to…** Right-click the entry and choose **Move to…**. The dialog shows the stack's folder tree; pick a destination folder (or **Stack root**) and choose **Move**. Destinations that would collide or make no sense (the entry's current folder, a folder inside the entry itself, or a name that is reserved at the stack root) are not selectable.
- **Drag and drop.** Drag the entry onto a folder in the tree, or onto empty space to move it to the stack root.
Moving requires stack edit permission. The protected stack files (compose / docker-compose / `.env`) at the stack root stay put: the compose CLI reads them from the stack directory, so they are not offered as move sources or root destinations. Moving a file changes only where it lives on disk; it does not redeploy or restart the stack.
## Copying and duplicating
Copy a file or folder without removing the original in one of two ways:
- **Duplicate.** Right-click the entry and choose **Duplicate** to make a copy in the same folder under an auto-suffixed name (`config.yaml` becomes `config copy.yaml`, then `config copy 2.yaml`, and so on). This is the quickest way to snapshot a file before you edit it.
- **Copy to…** Right-click the entry and choose **Copy to…** to pick a destination folder, the same way Move works. The current folder stays disabled in the picker, since a same-folder copy is what Duplicate is for.
Copying requires stack edit permission and stays within the current root. A folder is copied with its full contents, and a symlink is copied as a link rather than as the file it points to. Unlike Move, a protected stack file can be copied: duplicating `compose.yaml` to `compose.yaml.bak`, or copying it into a subfolder, is allowed because the copy is an ordinary file. Only creating a reserved name (`compose.yaml`, `.env`, and the rest) directly at the stack root is blocked.
## Working with multiple files
Select more than one entry to act on them together.
- **Select.** Hover a row to reveal its checkbox, or hold **Ctrl** (**Cmd** on macOS) and click a row to add it to the selection. Hold **Shift** and click to select a contiguous range. A plain click still opens a file in the viewer and leaves the selection untouched.
- **Act.** Once anything is selected, a bar at the top of the tree shows the count and the bulk actions.
| Bulk action | What it does |
|---|---|
| Download | Streams the whole selection as a single `.tar.gz` archive, preserving folder structure. Available to anyone who can read the stack. |
| Move | Opens the destination picker (the same one as single Move) and relocates every selected item into the chosen folder. Requires stack edit. |
| Delete | Asks for confirmation, then removes every selected item; selected folders are removed with their contents. Requires stack edit. |
| Clear | Drops the current selection. |
Bulk Move and Delete skip the protected stack files (`compose.yaml`, `.env`, and the rest) at the stack root, which are reported as kept rather than acted on; they can still be included in a Download.
Each item is processed independently. If some succeed and others fail (for example, a permission error on one file), Sencho reports how many succeeded and leaves the items that failed selected so you can see which they were and retry.
**Selection and archive limits.** A bulk action accepts at most 100 selected paths per request; select fewer items and repeat the action for a larger batch. A bulk Download is also capped at 5000 files and 1 GiB of uncompressed content: a selection that would exceed either limit is refused before the archive starts building, so narrow the selection or download large folders in smaller batches.
## Permissions (chmod)
Right-click any file and choose **Permissions** to inspect or edit its Unix mode bits. The dialog shows a 3 by 3 grid of `r` / `w` / `x` toggles for Owner, Group, and Other, plus the current octal value.
<Frame>
<img src="/images/stack-file-explorer/permissions-dialog.png" alt="Permissions modal for a file showing the rwx grid for Owner, Group, and Other plus the current octal value" />
</Frame>
When your account has stack edit permission, the toggles are interactive and the footer adds **Save**. For viewer accounts the dialog opens read-only: the toggles render the current state and the footer shows only **Close**.
<Note>
Permissions are applied with `chmod`. Permission changes on symlinks are not supported; edit the target file's permissions directly.
</Note>
## Deleting
There are three delete entry points. All three require stack edit permission, and all three open the same confirmation modal.
- **Toolbar delete.** With a file open in the viewer, click **Delete** in the right-pane action bar.
- **Context-menu delete.** Right-click any file or folder in the tree and choose **Delete**.
- **Non-empty folder delete.** The first attempt is non-recursive. If the folder has contents, the modal switches to `This folder is not empty. Delete everything inside?` with a destructive **Delete all** button.
When the entry is one of the five protected names, the modal asks you to type the filename before the destructive button activates.
<Frame>
<img src="/images/stack-file-explorer/delete-protected-confirm.png" alt="Delete confirmation modal for compose.yaml requiring the user to type the filename before the destructive button activates" />
</Frame>
<Warning>
Deletes are permanent. Sencho does not keep a trash bin or undo history for file operations.
</Warning>
## Context menu reference
<Frame>
<img src="/images/stack-file-explorer/context-menu-folder.png" alt="Right-click menu on a folder showing New File, New Folder, Rename, Duplicate, Copy to, Move to, and Delete entries" />
</Frame>
<Frame>
<img src="/images/stack-file-explorer/context-menu-file.png" alt="Right-click menu on a file showing Rename, Duplicate, Copy to, Move to, Permissions, and Delete entries" />
</Frame>
Right-click anywhere on a row, not only on its name, to open the menu.
| Right-click target | Admin entries (with stack edit) | Viewer entries |
|---|---|---|
| Folder | New File, New Folder, Rename, Duplicate, Copy to…, Move to…, Delete | No write entries |
| File | Rename, Duplicate, Copy to…, Move to…, Permissions, Delete | Permissions (read-only) |
**Move to…** is absent for the protected stack files (compose / docker-compose / `.env`) at the stack root, which cannot be moved out of the stack directory.
The Permissions dialog opens for everyone; only users with stack edit permission can save changes.
## Troubleshooting
<AccordionGroup>
<Accordion title="The Files tab is missing">
The signed-in role lacks **stack read** permission for this stack, so the tab is not offered; opening it directly (for example via a bookmarked link) redirects you back to the Anatomy tab. Ask an admin to grant the permission on your account, or sign in with a role that already has it.
</Accordion>
<Accordion title="No browsable stack volumes detected">
The stack declares no volume Sencho can reach, so the explorer opens on the stack source. This happens when the stack uses only anonymous or `tmpfs` mounts, when its bind mounts point at host paths that are not mounted into the Sencho container, or when its named volumes have not been created yet. Deploy the stack so its named volumes exist, or mount the relevant host path into the Sencho container, then reopen the tab.
</Accordion>
<Accordion title="A volume is shown but cannot be browsed">
The root carries a short explanation. A bind mount to a host path Sencho cannot reach needs that path mounted into the Sencho container. A bind that targets a single file rather than a directory has no tree to browse. A mount that targets a protected host location (the filesystem root, kernel/OS paths, Sencho's own runtime binaries, or the Docker socket) or that overlaps an area Sencho manages at runtime is intentionally not browsable. A named volume that cannot be resolved on the owning node is not browsable until it exists, or until a transient Docker connectivity issue clears.
</Accordion>
<Accordion title="Saving a file in a named volume returns a permission error">
Sencho writes to named volumes as an unprivileged helper user. A file owned by `root` or a specific application user may not be writable by that helper. Adjust the file's ownership or permissions on the host, or edit the file from a context that owns it, then try again.
</Accordion>
<Accordion title="A volume is read-only">
Every service that mounts the volume declares it read-only (for example `./config:/config:ro`). The explorer browses it but disables the edit, upload, delete, and rename controls. Change the mount to read/write in the compose file and redeploy if you need to edit its contents from Sencho.
</Accordion>
<Accordion title="Cannot delete compose.yaml or .env">
The five canonical stack files (`compose.yaml`, `compose.yml`, `docker-compose.yaml`, `docker-compose.yml`, `.env`) are protected because removing them mid-life breaks the stack. To delete a stack entirely, use **Delete stack** in the stack toolbar's overflow menu rather than removing these files individually.
</Accordion>
<Accordion title="Save returns 'file changed elsewhere'">
Another writer (a teammate, a deploy hook, or an out-of-band edit on the host) saved the file after you loaded it. Use the diff to compare your edits against the current version, then save again. Your typed buffer is kept so you do not have to retype your changes.
</Accordion>
<Accordion title="Upload fails with no obvious reason">
A full disk on the host backing the target directory surfaces as a generic write failure, not a dedicated error code. Check the volume mounted as `COMPOSE_DIR` (defaults to `/app/compose` inside the Sencho container) or the volume backing the target, free up space, or move it to a larger disk, then retry the upload.
</Accordion>
<Accordion title="File shows as 'Binary file detected' but it is a text file">
The binary detection heuristic flagged null bytes or a high proportion of non-printable characters in the file's first kilobytes. This can happen with certain encodings, non-UTF-8 byte sequences, or unusual line-ending mixes. Click **Open as text anyway** on the binary panel to re-fetch the file as UTF-8 text and edit it in the inline viewer. Files above the 2 MB preview limit still need a Download.
</Accordion>
<Accordion title="Upload fails with a 413 error">
The file exceeds the 25 MB per-file upload limit. Compress or split the file before uploading, or transfer it via `scp` or `rsync` directly to the stack directory on the host.
</Accordion>
<Accordion title="Cannot delete a folder">
The first delete attempt is non-recursive. If the folder is not empty, the dialog switches to `This folder is not empty. Delete everything inside?`. Click **Delete all** to remove the folder and its contents. If the error persists, check whether a running container has an open file handle inside that directory; stop the relevant service and retry.
</Accordion>
<Accordion title="File saved but the container still reads the old content">
The container caches the file at startup, or the mount is configured read-only inside the container. Restart the service after saving so the container picks up the new file.
</Accordion>
<Accordion title="Download button is missing">
The Download button appears in the right-pane action bar only when a file is selected in the tree. Click any text or binary file to open it, and the button activates.
</Accordion>
<Accordion title="A large directory shows a 'N total entries' footer">
Each directory render is capped at 1000 entries to keep the tree responsive. Use the filter input above the tree to narrow the list, or drop into a host shell with `cd` into the stack directory if you need to work with entries past the cap.
</Accordion>
<Accordion title="Write controls are missing">
Upload, create, rename, copy, move, chmod save, and delete require **stack edit** permission. Viewer accounts can browse, preview text files, inspect permissions in read-only mode, and download (including a bulk download).
</Accordion>
<Accordion title="A bulk action reports that some items failed">
Bulk delete and move process each item independently and report how many succeeded. The items that failed stay selected so you can see which they were; open one to find out why (a common cause is a permission error on a named-volume file, or a name collision at the destination), fix it, and retry on the still-selected items.
</Accordion>
<Accordion title="Bulk download is refused as too large">
A bulk download is packed into a single archive capped at 5000 files and 1 GiB of uncompressed content. A selection that would exceed either limit is refused before the download starts. Narrow the selection, or download large folders in smaller batches.
</Accordion>
<Accordion title="A bulk action is rejected for having too many items">
Bulk actions (download, move, delete) accept at most 100 selected paths per request. Select fewer items, or repeat the action across a few smaller batches.
</Accordion>
</AccordionGroup>