feat(extension): derive the version as unpadded CalVer (milestone 318 step 8)

`1.0.<minutes since 2020>` -> `YYYY.M.D.HHMM` UTC, from the commit time of
the newest change to a packaged extension file. Same clock and same commit as
before; readable instead of opaque, and the same value the rest of the family
derives.

The hold on this step was two questions about AMO, and Mozilla's own docs
answer both:

    ^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$

  1. four all-numeric segments -> ACCEPTED ({0,3} more after the first).
  2. leading zeros              -> REJECTED. A segment is the single digit
     `0` or starts 1-9, so `08` and `0201` are refused. MDN says it in prose
     too: "Non-zero numbers must not include a leading zero."

So the documented fallback applies, extension only: the same numbers rendered
without the family's zero-padding. `2026.08.29.0201` and `2026.8.29.201` are
one value in two renderings — rule 148 defines comparison as numeric per
segment, under which they are equal — so nothing already published is
reordered, and left-padding each segment recovers the family string exactly.
HHMM stays one segment because AMO allows at most four.

The transition is safe in the other direction too: 2026 > 1, so every CalVer
outranks every published 1.0.x. build.yml's downgrade guard confirms it.

Also in scope:

* MAJOR.MINOR is gone. `cmd_major_minor`, `cmd_patch` and VERSION_EPOCH go
  with it, the committed version in manifest.json / package.json is now
  wholly inert, and ci.yml's MAJOR.MINOR-agreement check is retired rather
  than left running beside a fact that stopped existing (rule 22).
* ci.yml's `extension-version` lane now asserts Mozilla's regex verbatim
  instead of a loose `^[0-9]+(\.[0-9]+)*$` — which would have passed the
  padded shape. It also asserts YYYY.M.D.HHMM, because AMO would accept a
  regression to `1.0.<minutes>` while that orders below everything signed
  since. Checking here is the point: AMO 409s on re-signing, so a version it
  rejects is burned and cannot be reused.
* `artifacts.sh version extension` delegates to packaging.sh, so the two
  cannot answer differently. The direction matches the existing one —
  artifacts.sh already asks packaging.sh for the extension's path set.

#3156 is what makes this commit safe to make: packaging.sh is in web's path
set, so the web revision moves with the extension version and build-web
rebuilds instead of republishing an image bundling the previous XPI.

Scribe #3138.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

.forgejo/workflows/ci.yml

@@ -2,6 +2,7 @@ name: CI
 
 # CI lanes per FabledRulebook/forgejo.md "CI philosophy":
 #   - lint: ruff only, no dep install — fast-fail for the common lint bounce.
+#   - extension-version: the derived version resolves and is a shape AMO takes.
 #   - backend-lint-and-test: `pytest -m "not integration"`, no service containers.
 #   - frontend-build: vitest unit + vite build.
 #   - integration: pgvector + redis service containers; alembic + `pytest -m integration`.
@@ -34,13 +35,92 @@ jobs:
       - name: Ruff lint
         # agent/ included so the GPU-agent is linted before its image is built
         # (build.yml only `docker build`s it — this is where it gets checked).
-        run: ruff check backend/ tests/ alembic/ agent/
+        # scripts/ likewise: release_notes.py runs only on a tag push, so a
+        # syntax or import error there would otherwise surface at the one
+        # moment nobody wants to debug a workflow.
+        run: ruff check backend/ tests/ alembic/ agent/ scripts/
       - name: Agent syntax check
         # The agent's runtime deps (torch/transformers/ultralytics) aren't in the
         # CI image, so we can't import it — but compileall parses every module,
         # catching syntax errors before the image build.
         run: python -m compileall -q agent/fc_agent
 
+  # The extension version is DERIVED, not hand-maintained (milestone 271 step
+  # 4): build.yml computes it from the commit TIME of the newest packaged
+  # extension change and stamps it into manifest.json / package.json at build
+  # time. The guard that used to live here — "packaged files changed but nobody
+  # bumped the version" — was therefore checking a fact that had stopped
+  # existing. Worse than useless: it would have failed this lane on every real
+  # extension change, demanding a bump that decides nothing. Retired 2026-08-27
+  # rather than left running beside the new mechanism (rule 22).
+  #
+  # Two things are still worth asserting, and this is the only lane that can:
+  # the extension.yml suite runs on node:24-slim, which is exactly why
+  # version.spec.js sticks to packaging.sh's git-free subcommands.
+  #   1. the derivation actually resolves on this commit
+  #   2. the derived string is one AMO will accept, checked against Mozilla's
+  #      own published grammar rather than a loose "digits and dots"
+  #
+  # The MAJOR.MINOR-agreement check that used to be (2) is gone with milestone
+  # 318 step 8: the committed version no longer seeds anything, so there is no
+  # hand-set part left for the two files to disagree about.
+  #
+  # Deliberately NOT checked here: that the derived value beats what has already
+  # been signed. That guard belongs in build.yml, where it compares against the
+  # real ext-* releases. Comparing against origin/main here would be wrong —
+  # dev legitimately derives a LOWER value whenever main is ahead on the
+  # extension, and a lane that fails for being behind is a lane people learn to
+  # ignore.
+  extension-version:
+    runs-on: python-ci
+    container:
+      image: git.fabledsword.com/bvandeusen/ci-python:3.14
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          # The derivation needs real history: a depth-1 clone sees one commit
+          # and produces a wrong, too-low value RATHER THAN FAILING. Checking
+          # that here is half the point of the lane.
+          fetch-depth: 0
+      - name: Extension version derives cleanly
+        run: |
+          set -eu
+          # busybox sh on the act_runner — no bashisms (family rule).
+          VERSION=$(sh extension/scripts/packaging.sh version)
+          echo "derived: $VERSION"
+
+          # Mozilla's published grammar for AMO, transcribed verbatim from
+          # MDN's manifest.json/version page:
+          #
+          #     ^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$
+          #
+          # Not the looser `^[0-9]+(\.[0-9]+)*$` this lane used to carry. That
+          # one passes `2026.08.29.0201`, which AMO REJECTS — a segment must be
+          # the single digit 0 or start 1-9 — and it also passes five segments,
+          # where AMO allows four. Both would surface as a failed sign with the
+          # version already burned: AMO 409s on re-signing, so a rejected value
+          # cannot be reclaimed and cannot be reused. This lane is the cheap
+          # place to find out. (#3138, milestone 318 step 8.)
+          if ! echo "$VERSION" | grep -qE '^(0|[1-9][0-9]{0,8})(\.(0|[1-9][0-9]{0,8})){0,3}$'; then
+            echo "ERROR: derived version '$VERSION' is not a version AMO accepts."
+            echo "AMO's grammar: ^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$"
+            echo "Most likely cause: a zero-padded segment (08, 0201). The rest"
+            echo "of the family pads; the extension must not — see packaging.sh."
+            exit 1
+          fi
+
+          # ...and the shape this project actually derives. AMO would happily
+          # take `1.0.3500147` too, so the grammar check alone would not notice
+          # a regression to the pre-318 shape — which orders BELOW everything
+          # signed since, and is unrecoverable once Firefox has the higher one.
+          if ! echo "$VERSION" | grep -qE '^20[0-9][0-9]\.[0-9]{1,2}\.[0-9]{1,2}\.[0-9]{1,4}$'; then
+            echo "ERROR: derived version '$VERSION' is not YYYY.M.D.HHMM."
+            echo "Rule 148's CalVer is what build.yml signs; the old"
+            echo "1.0.<minutes> shape would order below every ext-2026.* release."
+            exit 1
+          fi
+          echo "OK: derived version $VERSION"
+
   backend-lint-and-test:
     runs-on: python-ci
     container:
@@ -52,6 +132,13 @@ jobs:
       SECRET_KEY: ci_unit_test_placeholder
     steps:
       - uses: actions/checkout@v4
+        with:
+          # Full history for tests/test_artifact_identity.py, which derives
+          # each artifact's revision to check the identity scheme. On a
+          # depth-1 clone that derivation either fails or returns the tip sha
+          # — so the lane would go green while asserting nothing, which is
+          # the one outcome worse than a red one.
+          fetch-depth: 0
 
       # Cache step removed 2026-05-26: act_runner's cache backend has been
       # broken on this homelab runner since 2026-05-15 (first as request-

.forgejo/workflows/extension.yml

@@ -1,5 +1,5 @@
 name: extension
-# Lint-only workflow. The sign-and-publish dance moved into build.yml's
+# Lint + unit tests. The sign-and-publish dance moved into build.yml's
 # `sign-extension` job (2026-05-25) — `:latest` now always bundles the XPI
 # because sign-extension runs as a build-web dependency in the SAME workflow,
 # eliminating the prior race between build.yml and a separate extension.yml.
@@ -10,10 +10,20 @@ on:
     paths:
       - 'extension/**'
       - '.forgejo/workflows/extension.yml'
+      # test/version.spec.js asserts things ABOUT the other two workflows —
+      # that neither inlines the packaged-file set, and that build.yml derives
+      # the shipped version rather than reading it out of the repo. A
+      # workflow-only edit can therefore break this suite, so it has to trigger
+      # it. build.yml joined the list at milestone 271 step 5, when the spec
+      # started asserting against it.
+      - '.forgejo/workflows/ci.yml'
+      - '.forgejo/workflows/build.yml'
   pull_request:
     branches: [main]
     paths:
       - 'extension/**'
+      - '.forgejo/workflows/ci.yml'
+      - '.forgejo/workflows/build.yml'
   workflow_dispatch:
 
 jobs:
@@ -23,7 +33,55 @@ jobs:
       image: node:24-bookworm-slim
     steps:
       - uses: actions/checkout@v4
-      - name: Install web-ext
-        run: cd extension && npm install --no-save --no-audit --no-fund
+      # Not --no-save: vitest and web-ext are both real devDependencies now,
+      # and the suite needs vitest resolvable from node_modules.
+      - name: Install dev dependencies
+        run: cd extension && npm install --no-audit --no-fund
       - name: Lint
         run: cd extension && npm run lint
+      # Pure-logic specs over lib/url.js and lib/platforms.js plus manifest /
+      # package version-consistency checks. No browser, no network.
+      - name: Unit tests
+        run: cd extension && npm run test:unit
+
+      # Everything else about packaging is asserted against our own declaration
+      # of what ships. This is the only check that asks web-ext what it ACTUALLY
+      # put in the archive. Until now that was an unverified assumption about
+      # glob semantics — and a fragile one: `test/**` reaches web-ext intact
+      # only because callers `set -f` first, so losing that quoting would
+      # silently start shipping dev files with no other signal.
+      - name: Verify XPI contents
+        run: |
+          set -eu
+          command -v unzip >/dev/null 2>&1 || { apt-get update -qq && apt-get install -y -qq unzip; }
+          cd extension
+          npm run build
+          ZIP=$(ls web-ext-artifacts/*.zip | head -1)
+          echo "=== packaged entries in $ZIP ==="
+          unzip -Z1 "$ZIP" | sort
+          echo "=== end ==="
+          ENTRIES=$(unzip -Z1 "$ZIP")
+          fail=0
+          # Must NOT ship: repo infrastructure with no business in a user's browser.
+          for pat in 'test/' 'scripts/' 'vitest.config.js' 'package.json' 'package-lock.json' 'README.md' 'node_modules/' 'web-ext-artifacts/'; do
+            if echo "$ENTRIES" | grep -q "^$pat"; then
+              echo "ERROR: '$pat' was packaged into the XPI but must not be"
+              fail=1
+            fi
+          done
+          # Must ship: if an exclusion pattern ever over-matches, the extension
+          # breaks at runtime rather than at build time, so assert presence too.
+          for req in 'manifest.json' 'lib/url.js' 'lib/api.js' 'lib/platforms.js' 'lib/cookies.js'; do
+            if ! echo "$ENTRIES" | grep -q "^$req$"; then
+              echo "ERROR: '$req' is missing from the XPI"
+              fail=1
+            fi
+          done
+          for dir in 'background/' 'popup/' 'options/' 'content/' 'icons/'; do
+            if ! echo "$ENTRIES" | grep -q "^$dir"; then
+              echo "ERROR: nothing from '$dir' was packaged"
+              fail=1
+            fi
+          done
+          [ "$fail" -eq 0 ] || exit 1
+          echo "XPI contents verified."

.forgejo/workflows/release.yml

@@ -0,0 +1,80 @@
+name: Release
+
+# A `v*` tag publishes a changelog. It does NOT build anything.
+#
+# Milestone 318 step 2 removed the tag trigger from build.yml: by the time
+# anyone tags a commit, `main` has already built and published it, and a
+# rebuild would re-push `:c-<sha>` — which rule 145 forbids even when the
+# source matches, since image configs carry timestamps and "same source" does
+# not mean "same manifest". That left the tag with no consequence at all.
+#
+# This is the consequence it has instead. Step 6 put the derived version in the
+# Settings footer, so an operator can say WHICH build they are running; this
+# says what is IN it that was not in the one they ran last month. Both halves
+# of one question (note #3127 §5).
+#
+# Nothing here runs on a schedule and nothing auto-tags on merge. Release tags
+# are bookmarks — cut one when you will want to point at that day by name,
+# otherwise don't (note #3127 §0). FC went twelve weeks between v26.06.04.0 and
+# the next one and nothing was wrong. A schedule would turn an optional
+# bookmark back into ceremony, which is the thing this milestone is removing.
+#
+# Cutting the tag is an explicit operator action under rule 2 ("`main` — never
+# without explicit request", which since 2026-08-28 covers PR, merge and tag
+# alike). This lane only decides what happens once they do.
+#
+# Requires repo secret RELEASE_TOKEN with the `write:release` scope — the same
+# PAT build.yml uses for the ext-<version> XPI asset cache.
+
+on:
+  push:
+    tags: ['v*']
+  # So a release body can be regenerated after the fact — the publisher PATCHes
+  # an existing release rather than falling through on a conflict, so re-running
+  # this on a tag rewrites the body instead of silently keeping the first one
+  # (note #3127 §6.7).
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: 'Tag to (re)publish notes for'
+        required: true
+
+jobs:
+  changelog:
+    runs-on: python-ci
+    container:
+      image: git.fabledsword.com/bvandeusen/ci-python:3.14
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          # Load-bearing twice over: the previous release is found by walking
+          # ancestry back through the tag graph, and the cross-check against
+          # the derived web version calls artifacts.sh, which reads commit
+          # times. A shallow clone would find no previous tag and emit the
+          # entire history as the changelog — plausible-looking and wrong.
+          fetch-depth: 0
+          ref: ${{ github.event.inputs.tag || github.ref }}
+
+      # The `:c-<sha>` rollback refs are only real if `main` built this commit.
+      # The script checks that against origin/main and downgrades the claim to
+      # "unverified" when it cannot resolve one; fetching it here means that
+      # downgrade stays an actual signal instead of firing on every release.
+      - name: Make main's history resolvable
+        run: git fetch --no-tags --quiet origin +main:refs/remotes/origin/main || true
+
+      # TAG goes through the environment, not through `${{ }}` inside the
+      # run block. The value is operator-supplied, and an expression expanded
+      # into a shell line is expanded BEFORE the shell sees it — there is no
+      # quoting that makes that safe. On a tag push it is empty and the script
+      # falls back to GITHUB_REF.
+      - name: Publish the derived changelog
+        env:
+          RELEASE_TOKEN: ${{ secrets.RELEASE_TOKEN }}
+          TAG: ${{ github.event.inputs.tag }}
+        run: |
+          set -eu
+          if [ -n "${TAG:-}" ]; then
+            python3 scripts/release_notes.py "$TAG"
+          else
+            python3 scripts/release_notes.py
+          fi

Dockerfile

@@ -47,6 +47,29 @@ RUN chmod +x entrypoint.sh
 
 COPY --from=frontend-builder /build/dist ./frontend/dist
 
+# Which channel this image belongs to — `dev` or `main` (milestone 271 step 7).
+# build.yml passes it; /api/extension/manifest reports it beside the version so
+# an operator can tell which channel an install came from without the channel
+# ever touching the version string.
+#
+# Empty by default, deliberately: a locally-built image then reports NO channel
+# rather than claiming to be one, and the manifest omits the field entirely —
+# indistinguishable from an image built before the field existed, which is
+# exactly the shape every reader already has to handle.
+#
+# Declared LAST on purpose. An ARG/ENV invalidates every layer below it, and
+# these are the values that differ between builds of otherwise identical
+# source — put them any earlier and the two channels could never share a
+# cached pip install.
+#
+# FC_VERSION is what the instance reports about itself in the UI. Since
+# milestone 318 stopped publishing version image tags, that self-report is
+# the only answer to "which build is this?" — nothing else names it.
+ARG FC_CHANNEL=""
+ENV FC_CHANNEL=${FC_CHANNEL}
+ARG FC_VERSION=""
+ENV FC_VERSION=${FC_VERSION}
+
 EXPOSE 8080
 
 ENTRYPOINT ["./entrypoint.sh"]

README.md

@@ -6,7 +6,50 @@ Combines what was [ImageRepo](https://git.fabledsword.com/bvandeusen/ImageRepo)
 
 ## Status
 
-Pre-v1. Not yet functional.
+In production. `main` is continuously deployed — every merge to `main` builds
+and publishes `:latest` images, so whatever is on `main` is what is running.
+Day-to-day work happens on `dev`, which publishes `:dev` images.
+
+## Versions and tags
+
+Three image tags exist, and no others:
+
+| Tag | Branch | Meaning |
+| --- | --- | --- |
+| `:latest` | `main` | Production. Moves on every merge. |
+| `:c-<sha>` | `main` | Immutable — the rollback unit, all three images together. |
+| `:dev` | `dev` | The rolling test channel. Moves on every push. |
+
+There are deliberately **no version tags**. Nothing pins one, and a per-build
+name nobody reads is upkeep for a model FC does not run (family rule 145; the
+reasoning is note #3127 §5). Rolling back is `docker pull …:c-<sha>`.
+
+Each artifact still has a version, derived rather than chosen: the commit time
+of the newest change to that artifact's *own* shipped files, as
+`YYYY.MM.DD.HHMM` UTC (rule 148). Four artifacts, four independent versions —
+a push touching only `agent/` re-versions the agent and leaves web and ml
+alone, and CI skips the builds whose content did not move.
+
+Because no registry name carries it, the running instance's own report is the
+only answer to "which build is this?". The foot of Settings shows
+`FabledCurator 2026.08.29.0201 · dev`, and `/api/health` returns the same two
+fields.
+
+Release tags are optional bookmarks — FC went twelve weeks without one and
+nothing was wrong. Pushing `v<version>` publishes a Forgejo release listing the
+commits since the previous tag; it builds no image.
+
+## What's in here
+
+Five deployable pieces, built by `.forgejo/workflows/build.yml`:
+
+| Piece | Built from | Image | Role |
+| --- | --- | --- | --- |
+| **Web / workers** | `Dockerfile` | `fabledcurator` | Quart API + the built Vue SPA in one image. `entrypoint.sh` picks the role: `web`, `worker`, `scheduler`. The `maintenance-long` service is a second `worker` pinned to the long-running maintenance queue. |
+| **ML worker** | `Dockerfile.ml` | `fabledcurator-ml` | Same app, plus `requirements-ml.txt` — tagging and embedding models that run in-container. |
+| **GPU agent** | `agent/Dockerfile` | `fabledcurator-agent` | Optional desktop-GPU worker (`agent/`). Leases jobs over **HTTP only** — never touches the database or Redis. Run it for a burst, stop it to reclaim the card. See `agent/README.md`. |
+| **Firefox extension** | `extension/` | signed XPI | MV3 extension: pushes platform session cookies into FC and adds a creator as a Source in one click. AMO-signed on both `dev` and `main` (one signature per extension change, shared by the two channels), bundled into that channel's web image and served from Settings → Maintenance. See `extension/README.md`. |
+| **Data** | — | `pgvector/pgvector:pg16`, `redis:7-alpine` | Postgres with pgvector for embeddings; Redis as the Celery broker. |
 
 ## Quick start
 
@@ -29,22 +72,42 @@ docker compose -f docker-compose.yml up -d
 # (skips the override so containers pull registry images)
 ```
 
+The GPU agent is deployed separately, on the machine with the card —
+`agent/docker-compose.yml`, not this stack.
+
 ## Deployment posture
 
 FabledCurator is designed to run inside a self-hosted homelab environment over plain HTTP. If you want TLS, terminate it at your reverse proxy. The app does not generate certificates, redirect to HTTPS, or set HSTS.
 
 ## CI / Forgejo setup
 
-The repo's workflows expect:
+Four workflows: `ci.yml` (lint, extension-version check, backend unit tests,
+frontend build, integration), `extension.yml` (extension lint, vitest, XPI
+content verification), `build.yml` (sign + publish), and `release.yml`, which
+runs only on a `v*` tag and publishes a changelog without building anything.
 
-- **Runner label `python-ci`** — a Forgejo runner with Python 3.14, ruff, and Node 22 pre-installed. Both `ci.yml` and `build.yml` use this label. The runner image (`runner-base:python-ci`) is built from `CI-Runner/CI-python/` in the operator's workspace; `make push` from that directory builds and pushes a new image when toolchain pins change.
-- **Repo secret `RELEASE_TOKEN`** — a Forgejo PAT with the following scopes:
+**The toolchain each job runs in is its `container.image`, not its `runs-on`
+label.** `runs-on: python-ci` only schedules the job onto a runner; every job
+then names the image it actually wants. `ci-requirements.md` is the current,
+authoritative list of images and per-job installs — read that rather than a
+copy here, so the two can't drift.
+
+The repo expects one secret:
+
+- **`RELEASE_TOKEN`** — a Forgejo PAT with:
   - `write:package` + `read:package` — for `docker push` to `git.fabledsword.com`
-  - `write:release` — for future release-cutting workflows
-  - `write:issue` — for future issue-management automation
+  - `write:release` — for the `ext-<version>` releases that cache the signed XPI
+  - `write:issue` — for issue-management automation
 
   Generate at https://git.fabledsword.com/user/settings/applications. The injected `GITHUB_TOKEN` cannot be used because it lacks `write:package`.
 
+AMO signing additionally needs `MOZILLA_AMO_JWT_KEY` / `MOZILLA_AMO_JWT_SECRET`.
+It runs on **both** channels and is cached per version: because the version is
+derived from commit time, `dev` and `main` derive the same number for the same
+source, so `main` finds `dev`'s signature already cached and makes no second AMO
+call. That cache is why signing must be one-shot — AMO rejects a re-signed
+version.
+
 ## License
 
 Personal project; use at your own discretion.

agent/fc_agent/app.py

@@ -21,7 +21,7 @@ log = logging.getLogger("fc_agent.app")
 # Bump on every agent change. The page embeds this and /status reports it; the UI
 # warns to reload when they differ — so a stale browser-cached page can't be
 # mistaken for "the new image didn't deploy". (Belt-and-braces with no-store.)
-VERSION = "2026-07-02.6 · sleep mode: an empty queue sheds to one downloader and backs the lease poll off to 15 min"
+VERSION = "2026-07-17.1 · idle model-unload: after ~5 min idle the GPU models release their VRAM and reload on the next job (env IDLE_UNLOAD_SECONDS, 0=off) · sleep mode sheds to one downloader"
 
 logbuf.install()
 cfg = Config.from_env()
@@ -334,9 +334,12 @@ _PAGE = """<!doctype html><html><head><meta charset=utf-8>
   waited.textContent=s.transient||0
   // Instantaneous pool state → demoted to the sub-line, where its jumpiness reads
   // as live churn rather than a "broken" headline metric.
+  // '=== false' (not falsy) so a stale page that doesn't send models_loaded shows
+  // nothing; when the idle monitor unloads, the VRAM meter drops alongside this.
   pipe.textContent='downloaders '+(s.downloaders!=null?s.downloaders:'—')+' · consumers '+(s.consumers!=null?s.consumers:'—')+' · on GPU '+(s.active||0)
    +' · net '+(s.net_mb_s!=null?s.net_mb_s.toFixed(1):'—')+' MB/s'
    +(s.bandwidth_limit_mb_s>0?(' / cap '+s.bandwidth_limit_mb_s):'')
+   +(s.models_loaded===false?' · GPU models unloaded (idle — reload on next job)':'')
   if(document.activeElement!==bw && s.bandwidth_limit_mb_s!=null) bw.value=s.bandwidth_limit_mb_s
   // Buffer occupancy bar (also driven here so it tracks the /status cadence).
   if(s.buffer!=null && s.buffer_max){ const p=Math.round(100*s.buffer/s.buffer_max)

agent/fc_agent/config.py

@@ -51,6 +51,12 @@ class Config:
     bandwidth_limit_mb_s: float  # aggregate download cap in MEGABYTES/s across
     #                              all downloaders + video streams (0 = unlimited);
     #                              tunable live from the agent UI
+    idle_unload_seconds: float   # after this long with the GPU idle (nothing in
+    #                              flight, queue empty or Stopped), unload the
+    #                              SigLIP embedder + YOLO proposers to free their
+    #                              VRAM; they reload lazily on the next job. A
+    #                              24/7 agent otherwise squats on ~5GB doing
+    #                              nothing. 0 disables (keep models warm forever).
 
     @classmethod
     def from_env(cls) -> Config:
@@ -87,4 +93,8 @@ class Config:
             # link to ~1-1.5 MB/s per stream, browser included). Raise it (or 0)
             # from the agent UI on wired/faster networks.
             bandwidth_limit_mb_s=float(os.environ.get("BANDWIDTH_LIMIT_MB_S", "8")),
+            # 5 min: long enough that a lull between job bursts doesn't thrash the
+            # (few-second) reload, short enough that an agent left running with an
+            # empty queue hands its VRAM back promptly.
+            idle_unload_seconds=float(os.environ.get("IDLE_UNLOAD_SECONDS", "300")),
         )

agent/fc_agent/detectors.py

@@ -170,6 +170,13 @@ class YoloProposer:
             ))
         return out
 
+    def unload(self) -> None:
+        """Drop the loaded YOLO so its VRAM can be reclaimed; detect() reloads it
+        lazily on the next job. Leaves _ok untouched — a healthy proposer comes
+        back, but one that self-disabled on a fault stays off."""
+        with self._lock:
+            self._model = None
+
 
 class Proposers:
     """The agent's proposer set, built from config. Each detector is optional —
@@ -216,3 +223,11 @@ class Proposers:
 
     def panels(self, image):
         return self._top(self._panel, image, self.cfg.max_panels)
+
+    def unload(self) -> None:
+        """Release every loaded proposer's YOLO (idle VRAM reclaim). The worker
+        also drops its reference to this Proposers and rebuilds a fresh one via
+        _proposers_for on the next job, so this is belt-and-braces."""
+        for p in (self._person, self._anatomy, self._panel):
+            if p is not None:
+                p.unload()

agent/fc_agent/embedder.py

@@ -75,3 +75,18 @@ class CropEmbedder:
         pooled = out.pooler_output if hasattr(out, "pooler_output") else out
         arr = pooled.float().cpu().numpy().astype(np.float32)
         return [row.reshape(-1).tolist() for row in arr]
+
+    def unload(self) -> bool:
+        """Drop the loaded model so its VRAM can be reclaimed — the idle monitor
+        calls this after a spell with no work so an idle agent doesn't squat on
+        the card; the next embed() reloads it lazily (a few seconds). Held under
+        BOTH the load and inference locks so it can never race a concurrent load
+        or an in-flight forward pass. Returns True if a model was actually
+        released (the caller then runs one empty_cache() to hand the freed blocks
+        back to the driver)."""
+        with self._load_lock, self._infer_lock:
+            if self._model is None:
+                return False
+            self._model = None
+            self._processor = None
+            return True

agent/fc_agent/worker.py

@@ -57,6 +57,15 @@ MAX_BACKOFF_SECONDS = 60.0
 # up on their own.
 IDLE_POLL_MAX_SECONDS = 900.0
 
+# Idle VRAM reclaim (operator 2026-07-17): the SigLIP embedder + YOLO proposers
+# load lazily and then stay warm for fast job bursts — but a 24/7 agent with an
+# empty queue would otherwise squat on that VRAM (~5GB on the operator's card)
+# indefinitely while doing nothing. So a monitor unloads them after
+# cfg.idle_unload_seconds with the GPU genuinely idle (nothing in flight, buffer
+# drained); they reload lazily on the next job. This is just how often the
+# monitor wakes to check — it bounds how soon past the threshold the unload fires.
+IDLE_UNLOAD_CHECK_INTERVAL = 30.0
+
 # A job whose fetch dies transiently this many times IN ONE SESSION stops being
 # handed back and is failed instead. Transient handbacks (release) burn no
 # attempts on the server, so a poisoned transfer — an original that stalls the
@@ -268,6 +277,11 @@ class Worker:
         self._proposers_sig = None    # detector-config signature the current
         #                               proposers were built for (#134)
         self._proposers_lock = threading.Lock()
+        # Monotonic time of the last GPU activity (a consumer finishing a job).
+        # The idle monitor unloads the warm models once this goes stale by
+        # cfg.idle_unload_seconds — see _idle_unload_loop.
+        self._last_gpu_activity = time.monotonic()
+        threading.Thread(target=self._idle_unload_loop, daemon=True).start()
 
     # --- held-lease bookkeeping --------------------------------------------
     def _hold(self, job_ids) -> None:
@@ -608,6 +622,9 @@ class Worker:
             "net_mb_s": round(self._net_mb_s, 1),       # observed aggregate rate
             "bw_capped": self._bw_capped,   # autoscaler holding at the cap (UI hint)
             "idle": self._idle,             # queue empty → poll backed off (UI hint)
+            # Whether the GPU models are currently resident (False after an idle
+            # unload freed their VRAM) — a plain bool read, UI hint only.
+            "models_loaded": self._embedder is not None or self._proposers is not None,
         }
 
     def _bump(self, *, processed=0, downloaded=0, errors=0, active=0, transient=0):
@@ -788,6 +805,9 @@ class Worker:
                     self._bump(processed=1)
             finally:
                 self._bump(active=-1)
+                # Mark the GPU busy-until-now so the idle monitor starts its
+                # unload countdown from when work actually stopped, not before.
+                self._last_gpu_activity = time.monotonic()
 
     def _ensure_embedder(self, model_name: str):
         if self._embedder is not None:
@@ -845,6 +865,61 @@ class Worker:
                 self._proposers_sig = sig
             return self._proposers
 
+    def _unload_models(self) -> bool:
+        """Release the GPU-resident models (SigLIP embedder + YOLO proposers) so an
+        idle agent hands their VRAM back instead of squatting on the card. They
+        reload lazily on the next job (_ensure_embedder / _proposers_for) — a
+        few seconds' cost paid only when work actually resumes. Dropping the
+        shared instances under their build locks means a concurrent job either
+        sees the old instance (before) or rebuilds a fresh one (after); the idle
+        monitor only calls this with nothing in flight, so no inference is using
+        them. Returns True if anything was released."""
+        released = False
+        with self._embedder_lock:
+            if self._embedder is not None:
+                self._embedder.unload()
+                self._embedder = None
+                released = True
+        with self._proposers_lock:
+            if self._proposers is not None:
+                self._proposers.unload()
+                self._proposers = None
+                self._proposers_sig = None
+                released = True
+        if released:
+            try:
+                import torch
+                if torch.cuda.is_available():
+                    # torch's caching allocator holds freed blocks; hand them back
+                    # to the driver so nvidia-smi actually reflects the drop.
+                    torch.cuda.empty_cache()
+            except Exception:   # noqa: BLE001 — torch absent / CPU-only → nothing to free
+                pass
+        return released
+
+    def _idle_unload_loop(self) -> None:
+        """Unload the warm GPU models after a stretch of inactivity so a 24/7
+        agent with an empty queue doesn't hold ~5GB of VRAM doing nothing. Fires
+        only when nothing is in flight (active == 0 AND the buffer is drained) and
+        no job has completed for cfg.idle_unload_seconds — a window long enough
+        that a brief lull between bursts doesn't thrash reload/unload. Covers BOTH
+        sleep mode (queue empty, pipeline still running) and a full Stop; the
+        models reload lazily on the next job. idle_unload_seconds <= 0 disables it."""
+        idle_after = self.cfg.idle_unload_seconds
+        if idle_after <= 0:
+            return
+        while True:
+            time.sleep(IDLE_UNLOAD_CHECK_INTERVAL)
+            if self._embedder is None and self._proposers is None:
+                continue                       # nothing loaded → nothing to free
+            if self._active != 0 or not self._buffer.empty():
+                continue                       # work in flight → keep them warm
+            if time.monotonic() - self._last_gpu_activity < idle_after:
+                continue                       # not idle long enough yet
+            if self._unload_models():
+                log.info("idle %.0fs — unloaded GPU models, freed VRAM "
+                         "(reload on next job)", idle_after)
+
     def _consume(self, job: dict, frames: list, stop_evt: threading.Event) -> bool:
         """Detect + embed the decoded frames and submit the result. Returns True
         when the job was completed (→ count it processed), False otherwise: a

backend/app/api/admin.py

@@ -459,6 +459,22 @@ async def trigger_prune_missing_files():
     return _queued(async_result)
 
 
+@admin_bp.route("/maintenance/reclaim-attachments", methods=["POST"])
+async def trigger_reclaim_attachments():
+    """Reclaim orphaned attachments (#3068). Body {"dry_run": bool}: dry_run
+    (the DEFAULT here) projects the orphan rows and unreferenced store blobs
+    without touching either; dry_run=false deletes the rows then unlinks every
+    blob no surviving row references. Maintenance queue; operator-triggered
+    only — never an unattended sweep, since the apply unlinks files. Returns the
+    Celery task id — poll /maintenance/task-result/<id> for the summary."""
+    from ..tasks.admin import reclaim_orphaned_attachments_task
+
+    body = await request.get_json(silent=True) or {}
+    dry_run = bool(body.get("dry_run", True))  # default to the SAFE preview
+    async_result = reclaim_orphaned_attachments_task.delay(dry_run=dry_run)
+    return _queued(async_result)
+
+
 @admin_bp.route("/maintenance/dedup-videos", methods=["POST"])
 async def trigger_dedup_videos():
     """Tier-1 video dedup (#871). Body {"dry_run": bool}: dry_run=true previews

backend/app/api/extension.py

@@ -6,12 +6,14 @@ from __future__ import annotations
 
 import asyncio
 import hashlib
+import hmac
 import re
 from pathlib import Path
 
 from quart import Blueprint, jsonify, request
 from sqlalchemy import select
 
+from ..build_info import FC_CHANNEL as _FC_CHANNEL
 from ..extensions import get_session
 from ..models import AppSetting
 from ..services.extension_service import (
@@ -30,6 +32,15 @@ XPI_DIR = Path("/app/frontend/dist/extension")
 
 _XPI_VERSION_RE = re.compile(r"fabledcurator-(?P<version>[\w.-]+)\.xpi$")
 
+# Which channel this image belongs to — "dev" or "main" — baked in at build
+# time (milestone 271 step 7). Read from build_info rather than the environment
+# a second time: /api/health reports the same value, and two independent
+# `os.environ.get` calls are two things that can drift.
+#
+# Still bound as a module-level name here, so tests monkeypatch
+# `extension.FC_CHANNEL` exactly as they did before, same as XPI_DIR above.
+FC_CHANNEL = _FC_CHANNEL
+
 
 async def _ext_key_required(session) -> bool:
     """Unlike /api/credentials (which accepts the browser path with no
@@ -41,7 +52,15 @@ async def _ext_key_required(session) -> bool:
     stored = (await session.execute(
         select(AppSetting.value).where(AppSetting.key == "extension_api_key")
     )).scalar_one_or_none()
-    return stored is not None and supplied == stored
+    if stored is None:
+        return False
+    # compare_digest, not `==`: the stored key is a shared secret, and a
+    # short-circuiting compare leaks its prefix through timing. Costs nothing
+    # here — it is not that this route is exposed (#3072). Compared as BYTES:
+    # compare_digest's str form rejects non-ASCII with TypeError, and this
+    # header is attacker-supplied, so a str compare would turn a junk key into
+    # a 500 instead of a 403.
+    return hmac.compare_digest(supplied.encode("utf-8"), stored.encode("utf-8"))
 
 
 def _extract_version(xpi_name: str) -> str:
@@ -124,13 +143,30 @@ def _read_manifest_sync() -> dict | None:
         return None
     versioned.sort(key=lambda p: p.stat().st_mtime)
     latest = versioned[-1]
-    return {
+    info = {
         "installed": True,
         "version": _extract_version(latest.name),
         "xpi_url": f"/extension/{latest.name}",
         "latest_url": "/extension/fabledcurator-latest.xpi",
         "sha256": _sha256(latest),
     }
+    # The channel goes BESIDE the version, never inside it. A `-dev` suffix is
+    # what silently disabled the dev channel in the sibling project this design
+    # comes from: the comparator returned nothing for a non-integer segment, so
+    # every dev version compared equal and "no update available" became
+    # indistinguishable from "I cannot read this version".
+    #
+    # Omitted rather than defaulted when unset. Absence already has a meaning
+    # every reader must handle — an image built before this field existed says
+    # exactly the same thing by not having the key — so a blank channel reuses
+    # that path instead of inventing a second "unknown" spelling.
+    #
+    # Reported verbatim, not validated against {"dev", "main"}: if an image
+    # declares something else, showing what it actually claims is more useful
+    # to whoever is debugging it than dropping the value on the floor.
+    if FC_CHANNEL:
+        info["channel"] = FC_CHANNEL
+    return info
 
 
 @extension_bp.route("/manifest", methods=["GET"])

backend/app/api/gpu.py

@@ -256,9 +256,7 @@ async def lease():
         if not await _agent_authed(session):
             return jsonify({"error": "unauthorized"}), 401
         jobs = await GpuJobService(session).lease(agent_id, batch_size=batch)
-        ml = (
-            await session.execute(select(MLSettings).where(MLSettings.id == 1))
-        ).scalar_one()
+        ml = await MLSettings.load(session)
         # image rows for url/mime in one shot
         ids = [j.image_record_id for j in jobs]
         imgs = {

backend/app/api/health.py

@@ -1,5 +1,20 @@
-"""Health endpoint — no DB or Redis touch; just liveness."""
+"""Health endpoint — no DB or Redis touch; liveness, plus the build's identity.
+
+The identity rides here rather than on a route of its own because it answers
+at the same cost: two module constants, no I/O, nothing that can be slow or
+fail. It is also already fetched app-wide — TopNav calls `refreshHealth` on
+mount — so a separate endpoint would mean a second request for two strings.
+
+Both fields are OMITTED when unset rather than sent empty. See build_info.
+"""
+
+from ..build_info import FC_CHANNEL, FC_VERSION
 
 
 async def get_health():
-    return {"status": "ok"}, 200
+    body = {"status": "ok"}
+    if FC_VERSION:
+        body["version"] = FC_VERSION
+    if FC_CHANNEL:
+        body["channel"] = FC_CHANNEL
+    return body, 200

backend/app/api/ml_admin.py

@@ -4,6 +4,7 @@ from quart import Blueprint, jsonify, request
 
 from ..extensions import get_session
 from ..models import MLSettings
+from ..services.ml.heads import AUTO_APPLY_THRESHOLD_MAX, AUTO_APPLY_THRESHOLD_MIN
 
 ml_admin_bp = Blueprint("ml_admin", __name__, url_prefix="/api/ml")
 
@@ -83,48 +84,21 @@ async def embedder_models():
 
 @ml_admin_bp.route("/settings", methods=["GET"])
 async def get_settings():
-    from sqlalchemy import select
-
     async with get_session() as session:
-        s = (
-            await session.execute(select(MLSettings).where(MLSettings.id == 1))
-        ).scalar_one()
-    return jsonify(
-        {
-            "cpu_embed_enabled": s.cpu_embed_enabled,
-            "video_frame_interval_seconds": s.video_frame_interval_seconds,
-            "video_max_frames": s.video_max_frames,
-            "embedder_model_version": s.embedder_model_version,
-            "head_min_positives": s.head_min_positives,
-            "head_auto_apply_precision": s.head_auto_apply_precision,
-            "head_auto_apply_enabled": s.head_auto_apply_enabled,
-            "head_auto_apply_min_positives": s.head_auto_apply_min_positives,
-            "ccip_match_threshold": s.ccip_match_threshold,
-            "ccip_auto_apply_enabled": s.ccip_auto_apply_enabled,
-            "ccip_auto_apply_threshold": s.ccip_auto_apply_threshold,
-            "presentation_auto_apply_enabled": s.presentation_auto_apply_enabled,
-            "presentation_auto_apply_threshold": s.presentation_auto_apply_threshold,
-            "presentation_conflict_threshold": s.presentation_conflict_threshold,
-            "process_auto_apply_enabled": s.process_auto_apply_enabled,
-            "process_auto_apply_threshold": s.process_auto_apply_threshold,
-            "process_conflict_threshold": s.process_conflict_threshold,
-            "embedder_model_name": s.embedder_model_name,
-            **{f: getattr(s, f) for f in _DETECTOR_FIELDS},
-        }
-    )
+        s = await MLSettings.load(session)
+    # Table-driven off _EDITABLE (which PATCH also writes) so a new settings field
+    # can never be silently absent from GET — the split that historically dropped
+    # fields. _EDITABLE already includes *_DETECTOR_FIELDS.
+    return jsonify({f: getattr(s, f) for f in _EDITABLE})
 
 
 @ml_admin_bp.route("/settings", methods=["PATCH"])
 async def patch_settings():
-    from sqlalchemy import select
-
     body = await request.get_json()
     if not isinstance(body, dict):
         return jsonify({"error": "body must be an object"}), 400
     async with get_session() as session:
-        s = (
-            await session.execute(select(MLSettings).where(MLSettings.id == 1))
-        ).scalar_one()
+        s = await MLSettings.load(session)
 
         # Merge the patch over current values, then validate the result as a
         # whole — the store-floor invariant couples three fields, so they
@@ -154,24 +128,24 @@ def _validate(p: dict) -> str | None:
     # Head training (#114).
     if int(p["head_min_positives"]) < 1:
         return "head_min_positives must be >= 1"
-    if not (0.5 <= float(p["head_auto_apply_precision"]) <= 0.999):
-        return "head_auto_apply_precision must be between 0.5 and 0.999"
+    if not (AUTO_APPLY_THRESHOLD_MIN <= float(p["head_auto_apply_precision"]) <= AUTO_APPLY_THRESHOLD_MAX):
+        return f"head_auto_apply_precision must be between {AUTO_APPLY_THRESHOLD_MIN} and {AUTO_APPLY_THRESHOLD_MAX}"
     if int(p["head_auto_apply_min_positives"]) < 1:
         return "head_auto_apply_min_positives must be >= 1"
-    if not (0.5 <= float(p["ccip_match_threshold"]) <= 0.999):
-        return "ccip_match_threshold must be between 0.5 and 0.999"
-    if not (0.5 <= float(p["ccip_auto_apply_threshold"]) <= 0.999):
-        return "ccip_auto_apply_threshold must be between 0.5 and 0.999"
+    if not (AUTO_APPLY_THRESHOLD_MIN <= float(p["ccip_match_threshold"]) <= AUTO_APPLY_THRESHOLD_MAX):
+        return f"ccip_match_threshold must be between {AUTO_APPLY_THRESHOLD_MIN} and {AUTO_APPLY_THRESHOLD_MAX}"
+    if not (AUTO_APPLY_THRESHOLD_MIN <= float(p["ccip_auto_apply_threshold"]) <= AUTO_APPLY_THRESHOLD_MAX):
+        return f"ccip_auto_apply_threshold must be between {AUTO_APPLY_THRESHOLD_MIN} and {AUTO_APPLY_THRESHOLD_MAX}"
     # Presentation chrome auto-hide (#141). Auto-apply runs high (hiding is
     # consequential); the conflict cut is a plain probability [0,1].
-    if not (0.5 <= float(p["presentation_auto_apply_threshold"]) <= 0.999):
-        return "presentation_auto_apply_threshold must be between 0.5 and 0.999"
+    if not (AUTO_APPLY_THRESHOLD_MIN <= float(p["presentation_auto_apply_threshold"]) <= AUTO_APPLY_THRESHOLD_MAX):
+        return f"presentation_auto_apply_threshold must be between {AUTO_APPLY_THRESHOLD_MIN} and {AUTO_APPLY_THRESHOLD_MAX}"
     if not (0.0 <= float(p["presentation_conflict_threshold"]) <= 1.0):
         return "presentation_conflict_threshold must be between 0 and 1"
     # Process auto-apply (#1464). wip/editor stay VISIBLE so a false apply is
     # low-harm (excludes-from-training + a review flag), but keep the same bar.
-    if not (0.5 <= float(p["process_auto_apply_threshold"]) <= 0.999):
-        return "process_auto_apply_threshold must be between 0.5 and 0.999"
+    if not (AUTO_APPLY_THRESHOLD_MIN <= float(p["process_auto_apply_threshold"]) <= AUTO_APPLY_THRESHOLD_MAX):
+        return f"process_auto_apply_threshold must be between {AUTO_APPLY_THRESHOLD_MIN} and {AUTO_APPLY_THRESHOLD_MAX}"
     if not (0.0 <= float(p["process_conflict_threshold"]) <= 1.0):
         return "process_conflict_threshold must be between 0 and 1"
     # Embedder model swap (#1190): both must be non-empty. Changing them means a

backend/app/api/settings.py

@@ -66,34 +66,9 @@ _EXTDL_TOGGLE_FIELDS = (
 async def get_import_settings():
     async with get_session() as session:
         row = await ImportSettings.load(session)
-    return jsonify({
-        "min_width": row.min_width,
-        "min_height": row.min_height,
-        "skip_transparent": row.skip_transparent,
-        "transparency_threshold": row.transparency_threshold,
-        "skip_single_color": row.skip_single_color,
-        "single_color_threshold": row.single_color_threshold,
-        "single_color_tolerance": row.single_color_tolerance,
-        "phash_threshold": row.phash_threshold,
-        "download_rate_limit_seconds": row.download_rate_limit_seconds,
-        "download_validate_files": row.download_validate_files,
-        "download_schedule_default_seconds": row.download_schedule_default_seconds,
-        "download_event_retention_days": row.download_event_retention_days,
-        "download_failure_warning_threshold": row.download_failure_warning_threshold,
-        "series_suggest_enabled": row.series_suggest_enabled,
-        "series_suggest_threshold": row.series_suggest_threshold,
-        "extdl_mega_enabled": row.extdl_mega_enabled,
-        "extdl_gdrive_enabled": row.extdl_gdrive_enabled,
-        "extdl_mediafire_enabled": row.extdl_mediafire_enabled,
-        "extdl_dropbox_enabled": row.extdl_dropbox_enabled,
-        "extdl_pixeldrain_enabled": row.extdl_pixeldrain_enabled,
-        "translation_enabled": row.translation_enabled,
-        "interpreter_base_url": row.interpreter_base_url,
-        "translation_target_lang": row.translation_target_lang,
-        "translation_min_confidence": row.translation_min_confidence,
-        "wip_title_tagging_enabled": row.wip_title_tagging_enabled,
-        "wip_soft_title_tagging_enabled": row.wip_soft_title_tagging_enabled,
-    })
+    # Table-driven off _EDITABLE_FIELDS (which PATCH also writes) so a new field
+    # can't be silently absent from GET.
+    return jsonify({f: getattr(row, f) for f in _EDITABLE_FIELDS})
 
 
 @settings_bp.route("/settings/import", methods=["PATCH"])

backend/app/build_info.py

@@ -0,0 +1,30 @@
+"""What this build IS — stamped at image build time, not configurable.
+
+Deliberately separate from `config.py`. Those are operator settings, read from
+the environment and meant to be changed. These describe the artifact itself and
+are baked in by CI (the `FC_VERSION` / `FC_CHANNEL` build args); an operator
+setting them by hand is not a supported thing to do, it is just how a value
+gets from the build into the running process.
+
+**Absent rather than empty when unknown.** A locally-built image has no version,
+and neither did any image predating the field — one spelling of "cannot say",
+which every reader already has to handle, instead of a second one to
+special-case (note #3127 §7).
+
+**Why this matters more than it used to.** Milestone 318 stopped publishing
+version image tags, so a running instance's self-report is now the *only*
+answer to "which build is this?" — there is no registry name left to check it
+against. A wrong value here has nothing to contradict it. That is why the UI
+renders `unknown` rather than a blank or a plausible default: an empty footer
+reads as "no version", which is a different and false claim.
+
+The channel lives BESIDE the version and is never folded into it (rule 149).
+A `-dev` suffix would be parsed by the extension's comparator as a segment
+worth 0, making every dev build compare equal to every other — issue #2993's
+exact failure.
+"""
+
+import os
+
+FC_VERSION = os.environ.get("FC_VERSION", "").strip()
+FC_CHANNEL = os.environ.get("FC_CHANNEL", "").strip()

backend/app/models/__init__.py

@@ -27,7 +27,7 @@ from .patreon_seen_media import PatreonSeenMedia
 from .pixiv_failed_media import PixivFailedMedia
 from .pixiv_seen_media import PixivSeenMedia
 from .post import Post
-from .post_attachment import PostAttachment
+from .post_attachment import PostAttachment, attachment_download_url
 from .presentation_review import PresentationReview
 from .series_chapter import SeriesChapter
 from .series_page import SeriesPage
@@ -58,6 +58,7 @@ __all__ = [
     "SubscribeStarSeenMedia",
     "Post",
     "PostAttachment",
+    "attachment_download_url",
     "PresentationReview",
     "SeriesChapter",
     "SeriesPage",

backend/app/models/ml_settings.py

@@ -10,6 +10,7 @@ from sqlalchemy import (
     Integer,
     String,
     func,
+    select,
 )
 from sqlalchemy.orm import Mapped, mapped_column
 
@@ -212,3 +213,14 @@ class MLSettings(Base):
     updated_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), nullable=False, server_default=func.now()
     )
+
+    @classmethod
+    async def load(cls, session) -> MLSettings:
+        """The singleton settings row (id=1), via an async session. Mirrors
+        ImportSettings.load — the shared singleton-loader pattern."""
+        return (await session.execute(select(cls).where(cls.id == 1))).scalar_one()
+
+    @classmethod
+    def load_sync(cls, session) -> MLSettings:
+        """The singleton settings row (id=1), via a sync session."""
+        return session.execute(select(cls).where(cls.id == 1)).scalar_one()

backend/app/models/post_attachment.py

@@ -65,3 +65,15 @@ class PostAttachment(Base):
     captured_at: Mapped[datetime] = mapped_column(
         DateTime(timezone=True), nullable=False, server_default=func.now()
     )
+
+
+def attachment_download_url(attachment_id: int) -> str:
+    """The path that streams this attachment's bytes.
+
+    Both serializers that expose an attachment to the frontend
+    (`provenance_service`, `post_feed_service`) built this literal themselves,
+    so changing the route in `api/attachments.py` meant two edits and only one
+    would be remembered (#3072). `test_attachment_download_url` pins it against
+    the app's registered rule, so the drift is caught rather than trusted to.
+    """
+    return f"/api/attachments/{attachment_id}/download"

backend/app/services/cleanup_service.py

@@ -48,6 +48,47 @@ log = logging.getLogger(__name__)
 _VIDEO_DURATION_UNKNOWN = -1.0
 
 
+# -- artist-cascade predicates (rule 93: ONE definition, preview + apply) ---
+# project_artist_cascade (preview) and delete_artist_cascade (apply) both build
+# their queries from these. The preview used to re-derive its own — which is how
+# it came to count images and stay silent about posts and attachments while the
+# apply destroyed both. Same failure shape as the 2026-06-08 fandom-tag
+# deletion, where a re-implemented delete predicate diverged from the preview's.
+# Returned as condition LISTS spread into `.where(*conds)`, matching
+# _unused_tag_conditions / _bare_post_conditions below.
+
+
+def _artist_images_conditions(artist_id: int) -> list:
+    """Images the cascade deletes (rows AND their on-disk files)."""
+    return [ImageRecord.artist_id == artist_id]
+
+
+def _artist_posts_conditions(artist_id: int) -> list:
+    """Posts the cascade destroys. The apply never names these — post.artist_id
+    is ondelete=CASCADE, so Postgres takes them when the artist row goes — which
+    is exactly why the preview has to name them: an artist whose posts are
+    body-only (no images) otherwise previews as `images: 0` and reads as an
+    empty artist, while every captured body/description/external-link set is
+    destroyed."""
+    return [Post.artist_id == artist_id]
+
+
+def _artist_attachments_conditions(artist_id: int) -> list:
+    """Attachments the cascade deletes. Matched by artist_id OR by the owning
+    post's artist: artist_id is nullable (_capture_attachment leaves it NULL
+    when no artist resolved), so neither arm alone covers every row. The
+    sha-addressed blobs are NOT unlinked (one blob backs many rows) — these are
+    row counts, and the bytes are not part of this operation's footprint."""
+    return [
+        or_(
+            PostAttachment.artist_id == artist_id,
+            PostAttachment.post_id.in_(
+                select(Post.id).where(*_artist_posts_conditions(artist_id))
+            ),
+        )
+    ]
+
+
 def project_artist_cascade(session: Session, *, slug: str) -> dict:
     """Read-only projection of what delete_artist_cascade would touch.
 
@@ -56,12 +97,17 @@ def project_artist_cascade(session: Session, *, slug: str) -> dict:
           "artist": {"id": int, "name": str, "slug": str},
           "projected": {
             "images": int,
+            "posts": int,         # hard-deleted by the post.artist_id CASCADE
+            "attachments": int,   # rows deleted; the sha-addressed blobs stay
             "sources": int,
             "thumbs": int,        # images with a thumbnail_path set
             "import_tasks": int,  # ImportTask rows referencing the artist's images
             "bytes_on_disk": int, # SUM(image_record.size_bytes) — column is NOT NULL
           },
         }
+    Every count is built from the shared `_artist_*_conditions` predicates the
+    apply uses, so the two halves cannot drift (rule 93).
+
     Raises LookupError if slug not found. No mutations.
     """
     from ..models.import_task import ImportTask
@@ -73,36 +119,49 @@ def project_artist_cascade(session: Session, *, slug: str) -> dict:
     if artist is None:
         raise LookupError(f"artist slug not found: {slug!r}")
 
+    images_conds = _artist_images_conditions(artist.id)
+
     images_count = session.execute(
-        select(func.count(ImageRecord.id))
-        .where(ImageRecord.artist_id == artist.id)
+        select(func.count(ImageRecord.id)).where(*images_conds)
     ).scalar_one()
+    posts_count = session.execute(
+        select(func.count(Post.id))
+        .where(*_artist_posts_conditions(artist.id))
+    ).scalar_one()
+    attachments_count = session.execute(
+        select(func.count(PostAttachment.id))
+        .where(*_artist_attachments_conditions(artist.id))
+    ).scalar_one()
+    # Sources have no shared predicate: the apply never queries them either, it
+    # gets them from the Artist.sources ORM cascade. Counted directly here.
     sources_count = session.execute(
         select(func.count(Source.id))
         .where(Source.artist_id == artist.id)
     ).scalar_one()
     thumbs_count = session.execute(
         select(func.count(ImageRecord.id))
-        .where(ImageRecord.artist_id == artist.id)
+        .where(*images_conds)
         .where(ImageRecord.thumbnail_path.is_not(None))
     ).scalar_one()
     import_tasks_count = session.execute(
         select(func.count(ImportTask.id))
         .where(
             ImportTask.result_image_id.in_(
-                select(ImageRecord.id).where(ImageRecord.artist_id == artist.id)
+                select(ImageRecord.id).where(*images_conds)
             )
         )
     ).scalar_one()
     bytes_on_disk = session.execute(
         select(func.coalesce(func.sum(ImageRecord.size_bytes), 0))
-        .where(ImageRecord.artist_id == artist.id)
+        .where(*images_conds)
     ).scalar_one()
 
     return {
         "artist": {"id": artist.id, "name": artist.name, "slug": artist.slug},
         "projected": {
             "images": images_count,
+            "posts": posts_count,
+            "attachments": attachments_count,
             "sources": sources_count,
             "thumbs": thumbs_count,
             "import_tasks": import_tasks_count,
@@ -277,6 +336,10 @@ def delete_artist_cascade(
     series_page / tag_suggestion_rejection from ImageRecord delete,
     and source / post / download_event / etc. from Artist delete
     (via Artist.sources cascade="all, delete-orphan").
+
+    The artist's post_attachment rows are cleared EXPLICITLY before the
+    artist row goes — see the comment at that step; leaving them to the
+    cascade aborts the whole delete on a unique violation.
     """
     artist = session.get(Artist, artist_id)
     if artist is None:
@@ -287,11 +350,22 @@ def delete_artist_cascade(
                 "files_deleted": 0,
                 "thumbs_deleted": 0,
                 "import_tasks_nulled": 0,
+                "posts_deleted": 0,
+                "attachments_deleted": 0,
                 "files_failed": 0,
             },
         }
     artist_info = {"id": artist.id, "name": artist.name, "slug": artist.slug}
 
+    # Counted BEFORE the delete: Postgres takes these via the post.artist_id
+    # CASCADE when the artist row goes, so afterwards there is nothing left to
+    # count. Reported so the summary can be checked against the preview's
+    # `posts` — the parity rule 93 asks for is only testable if both halves
+    # actually state the number.
+    posts_deleted = session.execute(
+        select(func.count(Post.id)).where(*_artist_posts_conditions(artist.id))
+    ).scalar_one()
+
     images_deleted = 0
     files_deleted = 0
     thumbs_deleted = 0
@@ -300,7 +374,7 @@ def delete_artist_cascade(
     while True:
         rows = session.execute(
             select(ImageRecord)
-            .where(ImageRecord.artist_id == artist.id)
+            .where(*_artist_images_conditions(artist.id))
             .limit(500)
         ).scalars().all()
         if not rows:
@@ -323,6 +397,28 @@ def delete_artist_cascade(
     # source_path_prefix matching that's out of scope here.
     import_tasks_nulled = 0
 
+    # Clear the artist's attachments BEFORE the artist row, or the delete below
+    # aborts. Deleting an artist CASCADEs to Post (post.artist_id is
+    # ondelete=CASCADE), which SET NULLs post_attachment.post_id — and
+    # `uq_post_attachment_null_post_sha` is a partial UNIQUE on sha256 ALONE
+    # WHERE post_id IS NULL, so any two of this artist's attachments sharing a
+    # sha collapse onto one another and raise. That is an ORDINARY shape, not a
+    # corrupt one: _capture_attachment deliberately writes one row per post over
+    # a single sha-addressed blob (a creator who attaches the same pdf to two
+    # posts has two rows), and a pre-existing filesystem-import row with the same
+    # sha and a NULL post_id collides on its own. Migration 0043 reasoned only
+    # about upgrade-time safety and never about this later SET NULL.
+    # _repoint_post_links guards the identical collision class in the reconcile
+    # path; this is its artist-cascade counterpart.
+    #
+    # Which rows count as the artist's — and why the blobs are left on disk —
+    # is _artist_attachments_conditions, shared with the preview.
+    attachments_deleted = session.execute(
+        delete(PostAttachment)
+        .where(*_artist_attachments_conditions(artist.id))
+    ).rowcount or 0
+    session.commit()
+
     session.delete(artist)
     session.commit()
 
@@ -333,6 +429,8 @@ def delete_artist_cascade(
             "files_deleted": files_deleted,
             "thumbs_deleted": thumbs_deleted,
             "import_tasks_nulled": import_tasks_nulled,
+            "posts_deleted": posts_deleted,
+            "attachments_deleted": attachments_deleted,
             "files_failed": files_failed,
         },
     }
@@ -1494,3 +1592,155 @@ def purge_gated_previews(
         "ledger_cleared": ledger_cleared,
         "posts_deleted": posts_deleted,
     }
+
+
+# -- orphaned attachment reclamation ---------------------------------------
+# PostAttachment's two FKs are both ON DELETE SET NULL, so a deleted post or
+# artist leaves the row behind rather than taking it. Nothing ever pruned those
+# rows, and nothing has ever unlinked a file under the attachment store — so
+# both rows and bytes accumulated permanently and were invisible to every
+# existing diagnostic.
+#
+# Why this is a DISK->DB reconciliation rather than a row sweep: the store is
+# sha-addressed and idempotent (attachment_store.store), so ONE blob backs MANY
+# rows. Deleting a row therefore does not free its blob, and — since the artist
+# cascade now deletes its attachment rows outright — a freed blob has no DB
+# pointer left to find it by. Walking the store and asking "does any row still
+# reference this sha?" catches orphans from every cause, including ones no
+# future delete path will think to report.
+
+# A blob is written by attachment_store.store BEFORE its row is inserted and
+# committed, so a just-stored file legitimately has no referencing row for a
+# moment. Same guard, same reasoning as ORPHAN_TEMP_MIN_AGE_HOURS in
+# tasks/maintenance.py: never judge a file younger than this.
+_ATTACHMENT_ORPHAN_MIN_AGE_HOURS = 6
+
+# Wall-clock budget for the store walk (rule 89). A library with a large
+# attachment store shouldn't be able to run this past its soft time limit; on
+# exhaustion it reports partial=True and the operator re-runs to finish.
+_ATTACHMENT_RECLAIM_BUDGET_SECONDS = 900
+
+# The store names files `<sha256><ext>`. Parse the sha as the first 64 chars
+# rather than via Path.stem: store() takes the extension straight from the
+# source filename, and a URL-encoded basename yields a multi-dot "suffix"
+# (see [[reference_url_encoded_basename_suffix]]) that would make stem eat part
+# of the sha. Validating the 64 chars as hex also skips anything else in the
+# tree that isn't a stored blob.
+_SHA256_HEX_LEN = 64
+
+
+def _orphan_attachment_conditions() -> list:
+    """PostAttachment rows belonging to nothing: both FKs nulled by a deleted
+    post AND a deleted artist. A row with post_id NULL but an artist_id is the
+    deliberate filesystem-import case (importer._capture_attachment writes it
+    that way) and is NOT an orphan — it is still attributed."""
+    return [
+        PostAttachment.post_id.is_(None),
+        PostAttachment.artist_id.is_(None),
+    ]
+
+
+def _is_sha_named(name: str) -> bool:
+    """True when `name` starts with a 64-char lowercase-hex sha256."""
+    if len(name) < _SHA256_HEX_LEN:
+        return False
+    head = name[:_SHA256_HEX_LEN]
+    return all(c in "0123456789abcdef" for c in head)
+
+
+def reclaim_orphaned_attachments(
+    session: Session, *, images_root: Path, dry_run: bool = False,
+) -> dict:
+    """Prune unattributed PostAttachment rows, then unlink store blobs that no
+    surviving row references.
+
+    Returns (same discovery keys either way, so the UI renders one shape):
+        {"rows": int,            # orphan rows found / deleted
+         "files": int,           # unreferenced blobs found / unlinked
+         "bytes": int,           # their total size
+         "scanned": int,         # blobs examined
+         "skipped_recent": int,  # blobs under the min-age guard
+         "files_failed": int,    # unlink raised (apply only)
+         "partial": bool}        # walk hit the time budget
+
+    dry_run computes exactly what the apply would do and mutates nothing — the
+    surviving-sha set is derived by NEGATING the same orphan predicate the
+    delete uses, so the preview cannot disagree with the apply (rule 93).
+    """
+    started = time.monotonic()
+    orphan_conds = _orphan_attachment_conditions()
+
+    if dry_run:
+        rows = session.execute(
+            select(func.count(PostAttachment.id)).where(*orphan_conds)
+        ).scalar_one()
+    else:
+        rows = session.execute(
+            delete(PostAttachment).where(*orphan_conds)
+        ).rowcount or 0
+        session.commit()
+
+    # Shas that still have a home. In the apply path the orphan rows are already
+    # gone, so `NOT orphan` is redundant but harmless; in the dry-run path it is
+    # what makes the projection honest about blobs the delete would free. One
+    # predicate, one query, both modes.
+    surviving_shas = set(session.execute(
+        select(PostAttachment.sha256).where(~and_(*orphan_conds)).distinct()
+    ).scalars())
+
+    root = Path(images_root) / "attachments"
+    cutoff = (
+        datetime.now(UTC).timestamp()
+        - _ATTACHMENT_ORPHAN_MIN_AGE_HOURS * 3600
+    )
+    files = 0
+    freed_bytes = 0
+    scanned = 0
+    skipped_recent = 0
+    files_failed = 0
+    partial = False
+
+    if root.is_dir():
+        for path in root.rglob("*"):
+            if time.monotonic() - started >= _ATTACHMENT_RECLAIM_BUDGET_SECONDS:
+                partial = True
+                break
+            # .partial staging files belong to cleanup_orphaned_temp_files —
+            # leave them alone rather than racing an in-flight store().
+            if path.suffix in (".part", ".partial") or not path.is_file():
+                continue
+            if not _is_sha_named(path.name):
+                continue
+            scanned += 1
+            sha = path.name[:_SHA256_HEX_LEN]
+            if sha in surviving_shas:
+                continue
+            try:
+                st = path.stat()
+                if st.st_mtime >= cutoff:
+                    skipped_recent += 1
+                    continue
+                size = st.st_size
+                if not dry_run:
+                    path.unlink()
+                files += 1
+                freed_bytes += size
+            except OSError as exc:
+                files_failed += 1
+                log.warning("reclaim_orphaned_attachments: %s: %s", path, exc)
+
+    if not dry_run and (rows or files):
+        log.info(
+            "attachment reclaim: %d orphan row(s) deleted, %d blob(s) unlinked "
+            "(%d bytes), %d failed, partial=%s",
+            rows, files, freed_bytes, files_failed, partial,
+        )
+    return {
+        "rows": rows,
+        "files": files,
+        "bytes": freed_bytes,
+        "scanned": scanned,
+        "skipped_recent": skipped_recent,
+        "files_failed": files_failed,
+        "partial": partial,
+    }

backend/app/services/credential_service.py

@@ -181,7 +181,7 @@ def _augment_cookies(platform: str, netscape: str) -> str:
     """Delegate to the platform's `augment_cookies` hook if one is
     registered (subscribestar, hentaifoundry, etc. — see
     `services/platforms/<name>.py`). No-op when the platform doesn't
-    register a hook (Patreon, DeviantArt). Centralizing the
+    register a hook (Patreon, Discord). Centralizing the
     quirks-per-platform in the platforms package means adding a new
     platform's cookie quirks doesn't require touching this file."""
     info = PLATFORMS.get(platform)

backend/app/services/download_backends.py

@@ -31,9 +31,8 @@ from .pixiv_ingester import PixivIngester
 from .subscribestar_ingester import SubscribeStarIngester
 
 # Platforms whose download + verify go through the native ingester rather than
-# gallery-dl. gallery-dl still serves the rest (hentaifoundry, discord,
-# deviantart — the latter slated for retirement, not migration) until they
-# migrate too.
+# gallery-dl. gallery-dl still serves the rest (hentaifoundry, discord) until
+# they migrate too.
 NATIVE_INGESTER_PLATFORMS = frozenset({"patreon", "subscribestar", "pixiv"})
 
 # Mirrors patreon_resolver._CAMPAIGNS_URL — surfaced in resolution-failure

backend/app/services/extension_service.py

@@ -55,12 +55,6 @@ _PLATFORM_PATTERNS: list[tuple[str, re.Pattern[str]]] = [
         r"^https?://(?:www\.)?hentai-foundry\.com/user/(?P<slug>[^/?#]+)",
         re.IGNORECASE,
     )),
-    ("deviantart", re.compile(
-        r"^https?://(?:www\.)?deviantart\.com/"
-        r"(?!home$|watch\b|tag\b|browse\b)"
-        r"(?P<slug>[^/?#]+)/?$",
-        re.IGNORECASE,
-    )),
     ("pixiv", re.compile(
         r"^https?://(?:www\.)?pixiv\.net/(?:en/)?users/(?P<slug>\d+)",
         re.IGNORECASE,

backend/app/services/gallery_dl.py

@@ -299,8 +299,9 @@ class GalleryDLService:
     # (services/patreon_ingester.py), not gallery-dl.
     PLATFORM_DEFAULTS = {
         # subscribestar removed — native-ingester platform now (#71); pixiv
-        # removed likewise (#129). The remaining entries are the gallery-dl
-        # platforms not yet migrated.
+        # removed likewise (#129); deviantart removed at #3069 as a dropped
+        # platform, not a migrated one. The remaining entries are the
+        # gallery-dl platforms not yet migrated.
         "hentaifoundry": {
             "content_types": ["all"],
             "directory": [],
@@ -316,15 +317,6 @@ class GalleryDLService:
             "reactions": False,
             "threads": True,
         },
-        "deviantart": {
-            "content_types": ["all"],
-            "directory": [],
-            "filename": "{index:>03}_{title[:50]}.{extension}",
-            "flat": True,
-            "original": True,
-            "mature": True,
-            "metadata": True,
-        },
     }
 
     def __init__(

backend/app/services/image_tag_apply.py

@@ -0,0 +1,51 @@
+"""Bulk, idempotent writes to the ``image_tag`` association table.
+
+Three writers attach tags to images in bulk: the WIP-title backfill
+(`wip_title.apply_wip_image_tags`), the concept-head auto-apply sweep and the
+system-tag auto-apply sweep (both in `ml/heads.py`). The two sweeps used to
+issue ONE INSERT PER ROW from inside their per-image loop — fine in steady
+state, but a first pass over a back-catalogue is tens of thousands of
+individual round-trips (#3072). All three share this one chunked multi-row
+insert now.
+
+Sync only: every caller runs on a sync ``Session`` (the Celery task path). No
+async service writes image_tag in bulk, so there is no async sibling to keep in
+step — unlike `db_helpers.get_or_create`, which does have one.
+"""
+
+from __future__ import annotations
+
+from sqlalchemy.dialects.postgresql import insert as pg_insert
+from sqlalchemy.orm import Session
+
+from ..models.tag import image_tag
+
+# 5000 rows x 3 bound params = 15000, comfortably inside Postgres' 65535-param
+# ceiling for a single statement. Raising this past ~21000 rows would exceed it.
+INSERT_CHUNK = 5000
+
+
+def insert_image_tags(
+    session: Session, rows: list[dict], *, chunk: int = INSERT_CHUNK
+) -> None:
+    """Attach ``rows`` to their images, skipping any tag already on one.
+
+    Each row is ``{"image_record_id": int, "tag_id": int, "source": str}``.
+    Does NOT commit — the caller owns the transaction.
+
+    ON CONFLICT DO NOTHING against the (image_record_id, tag_id) primary key,
+    so an existing tag keeps its ORIGINAL ``source``: re-running a sweep can
+    never re-stamp a tag the operator applied by hand as machine-applied.
+
+    Returns nothing on purpose. psycopg reports ``rowcount`` -1 for a multi-row
+    ON CONFLICT DO NOTHING insert (it runs via an executemany path), so a count
+    taken from the statement would be a lie rather than an approximation.
+    Callers that need an accurate count derive it themselves — see
+    `wip_title.apply_wip_image_tags`' pre-SELECT, and the sweeps' `skip` sets.
+    """
+    for start in range(0, len(rows), chunk):
+        session.execute(
+            pg_insert(image_tag)
+            .values(rows[start:start + chunk])
+            .on_conflict_do_nothing(index_elements=["image_record_id", "tag_id"])
+        )

backend/app/services/ml/character_prototypes.py

@@ -150,9 +150,7 @@ def refresh_character_prototypes(
     """Incrementally refresh the prototype store. `full=True` rebuilds every
     character regardless of the gate/fingerprints (nightly reconcile). Returns
     {skipped, rebuilt, removed}; commits."""
-    settings = session.execute(
-        select(MLSettings).where(MLSettings.id == 1)
-    ).scalar_one()
+    settings = MLSettings.load_sync(session)
     sig = _global_signature(session)
     if not full and settings.ccip_ref_signature == sig:
         return {"skipped": True, "rebuilt": 0, "removed": 0}
@@ -204,9 +202,7 @@ def retract_auto_applied_ccip(session: Session) -> int:
     n_retracted."""
     import numpy as np
 
-    settings = session.execute(
-        select(MLSettings).where(MLSettings.id == 1)
-    ).scalar_one()
+    settings = MLSettings.load_sync(session)
     if not settings.ccip_auto_apply_enabled:
         return 0
     thr = float(settings.ccip_auto_apply_threshold)

backend/app/services/ml/heads.py

@@ -23,6 +23,7 @@ from datetime import UTC, datetime
 from typing import Any
 
 from sqlalchemy import delete, exists, func, select
+from sqlalchemy.dialects.postgresql import insert as pg_insert
 from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.orm import Session
 
@@ -40,8 +41,10 @@ from ...models import (
     TagSuggestionRejection,
 )
 from ...models.tag import CHROME_SYSTEM_TAGS, PROCESS_SYSTEM_TAGS, image_tag
+from ..image_tag_apply import insert_image_tags
 from .training_data import (
     _AUTO_SOURCES,
+    _applied_or_rejected,
     _auto_apply_point,
     _hygiene_excluded_ids,
     _ids_with_tag,
@@ -61,6 +64,14 @@ MIN_POSITIVES_FLOOR = 8     # hard floor; settings.head_min_positives can raise
 _UNLABELED_POOL = 4000
 _EXAMPLES_MIN = 8           # need at least this many embedded +/- to fit a head
 
+# Auto-apply / match confidence operating range. Every graduated auto-apply or
+# CCIP-match threshold the operator can set lives in this band, and the head
+# precision target is clamped to it: below 0.5 "auto-apply" is meaningless, and
+# 1.0 is unachievable so 0.999 is the ceiling. One source shared by the service
+# clamp (_normalize_params) and the API validator (ml_admin._validate).
+AUTO_APPLY_THRESHOLD_MIN = 0.5
+AUTO_APPLY_THRESHOLD_MAX = 0.999
+
 # Only these tag kinds get heads (the surfaced suggestion categories).
 _HEAD_KINDS = (TagKind.general, TagKind.character)
 # tag.kind -> the suggestion category the rail groups under.
@@ -78,6 +89,38 @@ _CATEGORY = {TagKind.general: "general", TagKind.character: "character"}
 _SYSTEM_TAG_SUGGEST_FLOOR = 0.65
 
 
+def _sigmoid(z, np):
+    """Logistic sigmoid 1/(1+e^-z): the head score→probability transform. One home
+    for what was inlined at every scoring site (suggest, both sweeps, retract)."""
+    return 1.0 / (1.0 + np.exp(-z))
+
+
+def _conflict_scores(Xn, Wc, bc, np):
+    """The presentation conflict signal (#141): per row, the MAX content-head
+    probability and WHICH head produced it. Shared by the system-tag sweep's guard-2
+    and the soft-wip audit — both ask "does this ALSO look like real content?"."""
+    cprobs = _sigmoid(Xn @ Wc.T + bc, np)
+    return cprobs.max(axis=1), cprobs.argmax(axis=1)
+
+
+def _insert_presentation_review(
+    session, *, image_record_id, tag_id, conflict_tag_id, conflict_score, mode,
+):
+    """Single-source the ring-loud PresentationReview row shape so the two writers
+    (system-tag sweep guard-2 + soft-wip audit) can't drift on columns or `mode` —
+    they share the (image_record_id, tag_id) composite PK, so a divergent `mode`
+    would be a silent first-writer-wins bug."""
+    session.execute(
+        pg_insert(PresentationReview)
+        .values(
+            image_record_id=image_record_id, tag_id=tag_id,
+            conflict_tag_id=conflict_tag_id, conflict_score=conflict_score,
+            mode=mode,
+        )
+        .on_conflict_do_nothing()
+    )
+
+
 class HeadTrainingAlreadyRunning(Exception):
     """Raised by start_head_training_run when a run is already in flight."""
 
@@ -103,9 +146,7 @@ def start_head_training_run(session: Session, params: dict[str, Any]) -> int:
 
 
 def _settings(session: Session) -> MLSettings:
-    return session.execute(
-        select(MLSettings).where(MLSettings.id == 1)
-    ).scalar_one()
+    return MLSettings.load_sync(session)
 
 
 def _normalize_params(session: Session, params: dict[str, Any] | None) -> dict[str, Any]:
@@ -124,7 +165,7 @@ def _normalize_params(session: Session, params: dict[str, Any] | None) -> dict[s
     except (TypeError, ValueError):
         cv_folds = DEFAULT_CV_FOLDS
     try:
-        precision_target = min(max(float(params.get("precision_target", s.head_auto_apply_precision)), 0.5), 0.999)
+        precision_target = min(max(float(params.get("precision_target", s.head_auto_apply_precision)), AUTO_APPLY_THRESHOLD_MIN), AUTO_APPLY_THRESHOLD_MAX)
     except (TypeError, ValueError):
         precision_target = s.head_auto_apply_precision
     return {
@@ -536,7 +577,7 @@ async def score_image(
     norms[norms == 0] = 1.0
     Xn = X / norms
     Z = Xn @ heads["W"].T + heads["b"]                  # (B, H)
-    probs_bag = 1.0 / (1.0 + np.exp(-Z))                # (B, H)
+    probs_bag = _sigmoid(Z, np)                         # (B, H)
     probs = probs_bag.max(axis=0)                       # (H,) best over the bag
     # ARGMAX beside the max: WHICH bag row won each head → the region that grounds
     # the tag (bag_meta[win]); None when the whole-image vector won (#1206).
@@ -614,9 +655,7 @@ async def ground_applied_tag(
 
 
 async def _settings_async(session: AsyncSession) -> MLSettings:
-    return (
-        await session.execute(select(MLSettings).where(MLSettings.id == 1))
-    ).scalar_one()
+    return await MLSettings.load(session)
 
 
 # --- Earned auto-apply (sync, ml worker) ---------------------------------
@@ -687,7 +726,6 @@ def auto_apply_sweep(
     embeddings in chunks; commits per chunk on a real run. Returns
     {n_applied, concepts:[{tag_id,name,applied,scanned,threshold}]}."""
     import numpy as np
-    from sqlalchemy.dialects.postgresql import insert as pg_insert
 
     settings = _settings(session)
     rows = _auto_apply_heads(
@@ -704,18 +742,7 @@ def auto_apply_sweep(
     names = [r.name for r in rows]
 
     # Skip images that already carry, or have rejected, each tag.
-    skip = {tid: set() for tid in tag_ids}
-    for tid in tag_ids:
-        for (iid,) in session.execute(
-            select(image_tag.c.image_record_id).where(image_tag.c.tag_id == tid)
-        ):
-            skip[tid].add(iid)
-        for (iid,) in session.execute(
-            select(TagSuggestionRejection.image_record_id).where(
-                TagSuggestionRejection.tag_id == tid
-            )
-        ):
-            skip[tid].add(iid)
+    skip = _applied_or_rejected(session, tag_ids)
 
     applied = [0] * len(rows)
     scanned = 0
@@ -729,8 +756,12 @@ def auto_apply_sweep(
         if not cids:
             continue
         Xn = _l2norm(np.vstack([emb[i] for i in cids]).astype(np.float32), np)
-        probs = 1.0 / (1.0 + np.exp(-(Xn @ W.T + b)))   # (N, H)
+        probs = _sigmoid(Xn @ W.T + b, np)              # (N, H)
         scanned += len(cids)
+        # Collected across every head, then written as ONE insert below. Was an
+        # insert per applied tag from inside this loop, which on a first sweep
+        # over a back-catalogue is tens of thousands of round-trips (#3072).
+        pending: list[dict] = []
         for h in range(len(rows)):
             tid = tag_ids[h]
             for idx in np.where(probs[:, h] >= thr[h])[0]:
@@ -740,12 +771,12 @@ def auto_apply_sweep(
                 skip[tid].add(iid)
                 applied[h] += 1
                 if not dry_run:
-                    session.execute(
-                        pg_insert(image_tag)
-                        .values(image_record_id=iid, tag_id=tid, source="head_auto")
-                        .on_conflict_do_nothing()
-                    )
+                    pending.append({
+                        "image_record_id": iid, "tag_id": tid,
+                        "source": "head_auto",
+                    })
         if not dry_run:
+            insert_image_tags(session, pending)
             session.commit()
         run.last_progress_at = datetime.now(UTC)
         session.commit()
@@ -840,7 +871,6 @@ def system_tag_auto_apply_sweep(
     enabled flag is set. numpy-only (no sklearn). Returns {n_applied, n_flagged,
     concepts}."""
     import numpy as np
-    from sqlalchemy.dialects.postgresql import insert as pg_insert
 
     cfg = _SWEEP_MODES[mode]
     settings = _settings(session)
@@ -869,18 +899,7 @@ def system_tag_auto_apply_sweep(
     valued = _valued_image_ids(session)
 
     # Skip images that already carry, or have rejected, each presentation tag.
-    skip = {tid: set() for tid in pres_tag_ids}
-    for tid in pres_tag_ids:
-        for (iid,) in session.execute(
-            select(image_tag.c.image_record_id).where(image_tag.c.tag_id == tid)
-        ):
-            skip[tid].add(iid)
-        for (iid,) in session.execute(
-            select(TagSuggestionRejection.image_record_id).where(
-                TagSuggestionRejection.tag_id == tid
-            )
-        ):
-            skip[tid].add(iid)
+    skip = _applied_or_rejected(session, pres_tag_ids)
 
     applied = [0] * len(pres)
     n_flagged = 0
@@ -895,12 +914,15 @@ def system_tag_auto_apply_sweep(
         if not cids:
             continue
         Xn = _l2norm(np.vstack([emb[i] for i in cids]).astype(np.float32), np)
-        probs = 1.0 / (1.0 + np.exp(-(Xn @ Wp.T + bp)))       # (N, P)
+        probs = _sigmoid(Xn @ Wp.T + bp, np)                  # (N, P)
         if Wc is not None:
-            cprobs = 1.0 / (1.0 + np.exp(-(Xn @ Wc.T + bc)))  # (N, C)
-            max_c = cprobs.max(axis=1)
-            arg_c = cprobs.argmax(axis=1)
+            max_c, arg_c = _conflict_scores(Xn, Wc, bc, np)   # (N,), (N,)
         scanned += len(cids)
+        # Same batching as auto_apply_sweep (#3072): collect the chunk's rows
+        # and write them once, below. The PresentationReview rows stay per-row —
+        # they FK to image_record/tag, not to image_tag, so writing the tags
+        # after them is safe, and a flagged conflict is rare by construction.
+        pending: list[dict] = []
         for p in range(len(pres)):
             tid = pres_tag_ids[p]
             for idx in np.where(probs[:, p] >= thr)[0]:
@@ -910,31 +932,25 @@ def system_tag_auto_apply_sweep(
                 skip[tid].add(iid)
                 applied[p] += 1
                 if not dry_run:
-                    session.execute(
-                        pg_insert(image_tag)
-                        .values(
-                            image_record_id=iid, tag_id=tid,
-                            source=source,
-                        )
-                        .on_conflict_do_nothing()
-                    )
+                    pending.append({
+                        "image_record_id": iid, "tag_id": tid,
+                        "source": source,
+                    })
                 # Guard 2: also looks like real content → still apply, but flag it
                 # for the review strip instead of silently marking (chrome hides,
                 # process stays visible — either way the operator gets a heads-up).
                 if Wc is not None and float(max_c[idx]) >= conflict_thr:
                     n_flagged += 1
                     if not dry_run:
-                        session.execute(
-                            pg_insert(PresentationReview)
-                            .values(
-                                image_record_id=iid, tag_id=tid,
-                                conflict_tag_id=conf_tag_ids[int(arg_c[idx])],
-                                conflict_score=float(max_c[idx]),
-                                mode=mode,
-                            )
-                            .on_conflict_do_nothing()
+                        _insert_presentation_review(
+                            session,
+                            image_record_id=iid, tag_id=tid,
+                            conflict_tag_id=conf_tag_ids[int(arg_c[idx])],
+                            conflict_score=float(max_c[idx]),
+                            mode=mode,
                         )
         if not dry_run:
+            insert_image_tags(session, pending)
             session.commit()
 
     concepts = [
@@ -956,7 +972,6 @@ def soft_wip_conflict_audit(session: Session, dry_run: bool = False) -> dict:
     NOT remove the tag; the operator decides. No-op when there are no content heads.
     numpy-only. Returns {n_scanned, n_flagged}."""
     import numpy as np
-    from sqlalchemy.dialects.postgresql import insert as pg_insert
 
     from ..wip_title import WIP_TITLE_SOFT_SOURCE, resolve_wip_tag_id
 
@@ -993,22 +1008,17 @@ def soft_wip_conflict_audit(session: Session, dry_run: bool = False) -> dict:
             continue
         scanned += len(cids)
         Xn = _l2norm(np.vstack([emb[i] for i in cids]).astype(np.float32), np)
-        cprobs = 1.0 / (1.0 + np.exp(-(Xn @ Wc.T + bc)))
-        max_c = cprobs.max(axis=1)
-        arg_c = cprobs.argmax(axis=1)
+        max_c, arg_c = _conflict_scores(Xn, Wc, bc, np)
         for k in range(len(cids)):
             if float(max_c[k]) >= conflict_thr:
                 n_flagged += 1
                 if not dry_run:
-                    session.execute(
-                        pg_insert(PresentationReview)
-                        .values(
-                            image_record_id=cids[k], tag_id=wip_id,
-                            conflict_tag_id=conf_tag_ids[int(arg_c[k])],
-                            conflict_score=float(max_c[k]),
-                            mode="process",
-                        )
-                        .on_conflict_do_nothing()
+                    _insert_presentation_review(
+                        session,
+                        image_record_id=cids[k], tag_id=wip_id,
+                        conflict_tag_id=conf_tag_ids[int(arg_c[k])],
+                        conflict_score=float(max_c[k]),
+                        mode="process",
                     )
         if not dry_run:
             session.commit()
@@ -1062,7 +1072,7 @@ def retract_auto_applied_heads(session: Session) -> int:
             continue
         Xn = _l2norm(np.vstack([emb[i] for i in cids]).astype(np.float32), np)
         w = np.asarray(weights, dtype=np.float32)
-        probs = 1.0 / (1.0 + np.exp(-(Xn @ w + float(bias))))
+        probs = _sigmoid(Xn @ w + float(bias), np)
         below = [cids[k] for k in np.where(probs < float(thr))[0]]
         for iid in below:
             session.execute(

backend/app/services/ml/training_data.py

@@ -94,6 +94,24 @@ def _rejected_ids(session: Session, tag_id: int) -> list[int]:
     ]
 
 
+def _applied_or_rejected(session: Session, tag_ids) -> dict[int, set[int]]:
+    """Per-tag skip set for the auto-apply sweeps: every image that ALREADY carries
+    the tag (ANY source — not just training positives) OR has rejected it. A sweep
+    never re-applies to these. Shared by auto_apply_sweep + system_tag_auto_apply_sweep
+    (heads.py) and scheduled_ccip_auto_apply (tasks/ml.py). Callers mutate the returned
+    sets in-place to also dedupe within a single run."""
+    skip: dict[int, set[int]] = {}
+    for tid in tag_ids:
+        ids = {
+            r[0] for r in session.execute(
+                select(image_tag.c.image_record_id).where(image_tag.c.tag_id == tid)
+            ).all()
+        }
+        ids.update(_rejected_ids(session, tid))
+        skip[tid] = ids
+    return skip
+
+
 def _sample_unlabeled(session: Session, exclude: set[int], limit: int) -> list[int]:
     """Random image ids (with an embedding) NOT carrying the tag. Concepts are
     sparse, so an untagged image is almost always a true negative."""

backend/app/services/patreon_resolver.py

@@ -91,48 +91,46 @@ def _sync_lookup(vanity: str, cookies_path: str | None) -> str | None:
     )
 
 
-def _lookup_via_api(vanity: str, cookies_path: str | None) -> str | None:
+def _campaigns_api_first(vanity: str, cookies_path: str | None) -> dict | None:
+    """The first `data` object from Patreon's campaigns API filtered by vanity
+    (`?filter[vanity]=<vanity>&fields[campaign]=name`), or None on any failure
+    (network / non-200 / non-JSON / empty). The single request shape shared by
+    _lookup_via_api (plucks the campaign id) and resolve_display_name (plucks the
+    display name)."""
     jar = _load_cookie_jar(cookies_path)
-    headers = {
-        "User-Agent": _USER_AGENT,
-        "Accept": "application/vnd.api+json",
-    }
-    params = {
-        "filter[vanity]": vanity,
-        "fields[campaign]": "name",
-    }
     try:
         resp = requests.get(
             _CAMPAIGNS_URL,
-            params=params,
-            headers=headers,
+            params={"filter[vanity]": vanity, "fields[campaign]": "name"},
+            headers={"User-Agent": _USER_AGENT, "Accept": "application/vnd.api+json"},
             cookies=jar,
             timeout=_TIMEOUT_SECONDS,
         )
     except requests.RequestException as exc:
         log.warning("Patreon campaigns API request failed for vanity=%s: %s", vanity, exc)
         return None
-
     if resp.status_code != 200:
         log.warning(
             "Patreon campaigns API returned HTTP %d for vanity=%s",
             resp.status_code, vanity,
         )
         return None
-
     try:
         payload = resp.json()
     except ValueError as exc:
         log.warning("Patreon campaigns API returned non-JSON for vanity=%s: %s", vanity, exc)
         return None
+    data = payload.get("data") if isinstance(payload, dict) else None
+    if not isinstance(data, list) or not data or not isinstance(data[0], dict):
+        return None
+    return data[0]
 
-    if not isinstance(payload, dict):
+
+def _lookup_via_api(vanity: str, cookies_path: str | None) -> str | None:
+    first = _campaigns_api_first(vanity, cookies_path)
+    if first is None:
         return None
-    data = payload.get("data")
-    if not isinstance(data, list) or not data:
-        return None
-    first = data[0] if isinstance(data[0], dict) else None
-    campaign_id = first.get("id") if first else None
+    campaign_id = first.get("id")
     if not isinstance(campaign_id, str) or not campaign_id:
         return None
     log.info("Resolved Patreon vanity=%s → campaign_id=%s", vanity, campaign_id)
@@ -144,24 +142,10 @@ def resolve_display_name(vanity: str, cookies_path: str | None) -> str | None:
     (`fields[campaign]=name`), used to name the Artist at add-time (#130). None
     on any failure — the caller falls back to the vanity handle. Sync: call from
     an executor."""
-    jar = _load_cookie_jar(cookies_path)
-    try:
-        resp = requests.get(
-            _CAMPAIGNS_URL,
-            params={"filter[vanity]": vanity, "fields[campaign]": "name"},
-            headers={"User-Agent": _USER_AGENT, "Accept": "application/vnd.api+json"},
-            cookies=jar,
-            timeout=_TIMEOUT_SECONDS,
-        )
-        if resp.status_code != 200:
-            return None
-        data = resp.json().get("data")
-    except (requests.RequestException, ValueError) as exc:
-        log.warning("Patreon name lookup failed for vanity=%s: %s", vanity, exc)
+    first = _campaigns_api_first(vanity, cookies_path)
+    if first is None:
         return None
-    if not isinstance(data, list) or not data or not isinstance(data[0], dict):
-        return None
-    name = (data[0].get("attributes") or {}).get("name")
+    name = (first.get("attributes") or {}).get("name")
     return name.strip() if isinstance(name, str) and name.strip() else None
 
 

backend/app/services/platforms/__init__.py

@@ -8,9 +8,10 @@ PLATFORMS below. Sidecar parsing, cookie materialization, and
 
 Lifted from GallerySubscriber's
 ~/Nextcloud/Projects/GallerySubscriber/backend/app/api/platforms.py
-and ~/.../extension/lib/platforms.js. Six platforms; auth_type and
+and ~/.../extension/lib/platforms.js. Five platforms; auth_type and
 URL patterns match GS exactly so the existing browser extension
-hits FC unmodified.
+hits FC unmodified. deviantart was dropped at #3069 (2026-08-27) —
+FC downloaders are art-dedicated services only.
 """
 
 from .base import (
@@ -18,7 +19,6 @@ from .base import (
     DEFAULT_EXTERNAL_POST_ID_KEYS,
     PlatformInfo,
 )
-from .deviantart import INFO as _DEVIANTART
 from .discord import INFO as _DISCORD
 from .hentaifoundry import INFO as _HENTAIFOUNDRY
 from .patreon import INFO as _PATREON
@@ -33,7 +33,6 @@ PLATFORMS: dict[str, PlatformInfo] = {
         _HENTAIFOUNDRY,
         _DISCORD,
         _PIXIV,
-        _DEVIANTART,
     )
 }
 

backend/app/services/platforms/base.py

@@ -63,7 +63,7 @@ class PlatformInfo:
     # Synthesize a post permalink from sidecar data. Required when
     # gallery-dl's `url` field is the file/CDN URL rather than the post
     # permalink (subscribestar/pixiv/hf/discord). None = trust the bare
-    # `url` field (patreon, deviantart).
+    # `url` field (patreon).
     derive_post_url: Callable[[dict], str | None] | None = None
 
     # Post-process the materialized cookies.txt for gallery-dl. Used by

backend/app/services/platforms/deviantart.py

@@ -1,23 +0,0 @@
-"""DeviantArt — no exercised quirks yet.
-
-No operator-owned DeviantArt archive existed at the 2026-05-27 sidecar
-audit, so we don't know yet whether DA's gallery-dl sidecars are
-well-behaved or have their own quirks. When DA gets exercised for the
-first time, add `derive_post_url` / `augment_cookies` here as needed.
-"""
-
-from .base import GD_DEFAULTS, PlatformInfo
-
-INFO = PlatformInfo(
-    key="deviantart",
-    name="DeviantArt",
-    description="Download artwork from DeviantArt artists",
-    auth_type="cookies",
-    requires_auth=False,
-    url_pattern=r"^https?://(www\.)?deviantart\.com/",
-    url_examples=[
-        "https://www.deviantart.com/example-artist",
-        "https://www.deviantart.com/example-artist/gallery",
-    ],
-    default_config={**GD_DEFAULTS, "content_types": ["gallery"]},
-)

backend/app/services/post_feed_service.py

@@ -24,6 +24,7 @@ from ..models import (
     Post,
     PostAttachment,
     Source,
+    attachment_download_url,
 )
 from ..utils.html_sanitize import (
     extract_img_srcs,
@@ -360,7 +361,7 @@ class PostFeedService:
                 "ext": att.ext,
                 "mime": att.mime,
                 "size_bytes": att.size_bytes,
-                "download_url": f"/api/attachments/{att.id}/download",
+                "download_url": attachment_download_url(att.id),
             })
         return out
 

backend/app/services/provenance_service.py

@@ -16,6 +16,7 @@ from ..models import (
     Post,
     PostAttachment,
     Source,
+    attachment_download_url,
 )
 from ..utils.html_sanitize import sanitize_post_html
 
@@ -53,7 +54,7 @@ def _attachment_dict(a: PostAttachment) -> dict:
         "original_filename": a.original_filename,
         "size_bytes": a.size_bytes,
         "ext": a.ext,
-        "download_url": f"/api/attachments/{a.id}/download",
+        "download_url": attachment_download_url(a.id),
     }
 
 

backend/app/services/wip_title.py

@@ -20,10 +20,10 @@ family gains one member.
 import re
 
 from sqlalchemy import select
-from sqlalchemy.dialects.postgresql import insert as pg_insert
 from sqlalchemy.orm import Session
 
 from ..models.tag import WIP_SYSTEM_TAG, Tag, image_tag
+from .image_tag_apply import insert_image_tags
 
 # image_tag.source stamped on title-heuristic WIP tags — distinct from the other
 # apply sources so provenance stays legible and a future undo can target only these.
@@ -113,13 +113,9 @@ def apply_wip_image_tags(
         to_insert = [iid for iid in chunk if iid not in already]
         if not to_insert:
             continue
-        session.execute(
-            pg_insert(image_tag)
-            .values([
-                {"image_record_id": iid, "tag_id": tag_id, "source": source}
-                for iid in to_insert
-            ])
-            .on_conflict_do_nothing(index_elements=["image_record_id", "tag_id"])
-        )
+        insert_image_tags(session, [
+            {"image_record_id": iid, "tag_id": tag_id, "source": source}
+            for iid in to_insert
+        ])
         inserted += len(to_insert)
     return inserted

backend/app/tasks/admin.py

@@ -409,3 +409,31 @@ def rescan_series_suggestions_task(self, after_post_id: int = 0) -> dict:
         )
         rescan_series_suggestions_task.delay(summary["resume_after_id"])
     return summary
+
+
+@celery.task(
+    name="backend.app.tasks.admin.reclaim_orphaned_attachments_task",
+    bind=True,
+    autoretry_for=(OperationalError, DBAPIError),
+    retry_backoff=15, retry_backoff_max=180, max_retries=1,
+    # The service stops walking at its own 900s budget and reports partial, so
+    # these limits are the backstop for a wedged filesystem (NFS stall), not the
+    # expected exit. Comfortably above the budget so a normal run always returns
+    # its summary rather than being killed mid-walk.
+    soft_time_limit=1200, time_limit=1500,  # 20 min / 25 min
+)
+def reclaim_orphaned_attachments_task(self, dry_run: bool = True) -> dict:
+    """Reclaim unattributed PostAttachment rows and the store blobs nothing
+    references any more (#3068). dry_run (the default) returns the projection
+    without touching rows or files; apply deletes the orphan rows, then unlinks
+    every blob no surviving row references.
+
+    Defaults to the SAFE preview — unlike the other tasks here, whose apply is
+    reversible-ish or scoped; this one deletes files. Operator-triggered only,
+    never on a beat: an unattended sweep that unlinks blobs is not something to
+    run without someone reading the projection first."""
+    SessionLocal = _sync_session_factory()
+    with SessionLocal() as session:
+        return cleanup_service.reclaim_orphaned_attachments(
+            session, images_root=IMAGES_ROOT, dry_run=dry_run,
+        )

backend/app/tasks/maintenance.py

@@ -173,6 +173,12 @@ TASK_STUCK_THRESHOLD_MINUTES: dict[str, int] = {
     # task-name override beats the queue threshold whatever queue the row records
     # (it recorded 'default' before the celery_signals fix → download). 65 = 60+5.
     "backend.app.tasks.external.fetch_external_link": 65,
+    # Attachment reclaim walks the whole sha-addressed store; the service caps
+    # itself at a 900s budget and reports partial, but the task's hard limit is
+    # 25 min for a wedged filesystem (NFS stall). Same phantom-flag class as the
+    # external-fetch entry above — without an override a healthy in-flight walk
+    # is swept 'RecoverySweep' at the bare 5-min default. 30 = 25 + 5.
+    "backend.app.tasks.admin.reclaim_orphaned_attachments_task": 30,
 }
 
 
@@ -776,89 +782,62 @@ def recover_stalled_library_audit_runs() -> int:
         return recovered
 
 
+def _recover_stalled_runs(model, *, stall_minutes: int, keep_runs: int, label: str) -> int:
+    """Shared recovery + retention sweep for the head run-tracking tables
+    (HeadTrainingRun / HeadAutoApplyRun, which share the
+    status/last_progress_at/started_at/finished_at/error/id columns): flip 'running'
+    rows with no progress past `stall_minutes` to 'error', then prune to the last
+    `keep_runs` (rule 89). Returns the number recovered. NOTE the two other recover
+    tasks are deliberately NOT folded in — library-audit has no prune tail and
+    backup uses a single started_at cutoff."""
+    SessionLocal = _sync_session_factory()
+    now = datetime.now(UTC)
+    cutoff = now - timedelta(minutes=stall_minutes)
+    with SessionLocal() as session:
+        result = session.execute(
+            update(model)
+            .where(model.status == "running")
+            .where(func.coalesce(model.last_progress_at, model.started_at) < cutoff)
+            .values(
+                status="error", finished_at=now,
+                error=f"stranded by recovery sweep (no progress for {stall_minutes} min)",
+            )
+        )
+        keep = session.execute(
+            select(model.id).order_by(model.id.desc()).limit(keep_runs)
+        ).scalars().all()
+        if keep:
+            session.execute(delete(model).where(model.id.not_in(keep)))
+        session.commit()
+        recovered = result.rowcount or 0
+        if recovered:
+            log.info("%s: recovered %d rows", label, recovered)
+        return recovered
+
+
 @celery.task(name="backend.app.tasks.maintenance.recover_stalled_head_training_runs")
 def recover_stalled_head_training_runs() -> int:
     """Flip HeadTrainingRun rows stuck in 'running' past the stall threshold to
     'error', and prune old runs to the last HEAD_TRAINING_KEEP_RUNS (retention,
     rule 89). Runs every 5 min on the maintenance lane; no-op when idle."""
-    SessionLocal = _sync_session_factory()
-    now = datetime.now(UTC)
-    cutoff = now - timedelta(minutes=HEAD_TRAINING_STALL_THRESHOLD_MINUTES)
-    with SessionLocal() as session:
-        result = session.execute(
-            update(HeadTrainingRun)
-            .where(HeadTrainingRun.status == "running")
-            .where(
-                func.coalesce(
-                    HeadTrainingRun.last_progress_at, HeadTrainingRun.started_at
-                )
-                < cutoff
-            )
-            .values(
-                status="error", finished_at=now,
-                error=(
-                    f"stranded by recovery sweep (no progress for "
-                    f"{HEAD_TRAINING_STALL_THRESHOLD_MINUTES} min)"
-                ),
-            )
-        )
-        keep = session.execute(
-            select(HeadTrainingRun.id).order_by(HeadTrainingRun.id.desc())
-            .limit(HEAD_TRAINING_KEEP_RUNS)
-        ).scalars().all()
-        if keep:
-            session.execute(
-                delete(HeadTrainingRun).where(HeadTrainingRun.id.not_in(keep))
-            )
-        session.commit()
-        recovered = result.rowcount or 0
-        if recovered:
-            log.info(
-                "recover_stalled_head_training_runs: recovered %d rows", recovered
-            )
-        return recovered
+    return _recover_stalled_runs(
+        HeadTrainingRun,
+        stall_minutes=HEAD_TRAINING_STALL_THRESHOLD_MINUTES,
+        keep_runs=HEAD_TRAINING_KEEP_RUNS,
+        label="recover_stalled_head_training_runs",
+    )
 
 
 @celery.task(name="backend.app.tasks.maintenance.recover_stalled_head_auto_apply_runs")
 def recover_stalled_head_auto_apply_runs() -> int:
     """Flip stalled HeadAutoApplyRun 'running' rows to 'error' + prune to the
     last HEAD_AUTO_APPLY_KEEP_RUNS (retention, rule 89). 5-min maintenance lane."""
-    SessionLocal = _sync_session_factory()
-    now = datetime.now(UTC)
-    cutoff = now - timedelta(minutes=HEAD_AUTO_APPLY_STALL_THRESHOLD_MINUTES)
-    with SessionLocal() as session:
-        result = session.execute(
-            update(HeadAutoApplyRun)
-            .where(HeadAutoApplyRun.status == "running")
-            .where(
-                func.coalesce(
-                    HeadAutoApplyRun.last_progress_at, HeadAutoApplyRun.started_at
-                )
-                < cutoff
-            )
-            .values(
-                status="error", finished_at=now,
-                error=(
-                    f"stranded by recovery sweep (no progress for "
-                    f"{HEAD_AUTO_APPLY_STALL_THRESHOLD_MINUTES} min)"
-                ),
-            )
-        )
-        keep = session.execute(
-            select(HeadAutoApplyRun.id).order_by(HeadAutoApplyRun.id.desc())
-            .limit(HEAD_AUTO_APPLY_KEEP_RUNS)
-        ).scalars().all()
-        if keep:
-            session.execute(
-                delete(HeadAutoApplyRun).where(HeadAutoApplyRun.id.not_in(keep))
-            )
-        session.commit()
-        recovered = result.rowcount or 0
-        if recovered:
-            log.info(
-                "recover_stalled_head_auto_apply_runs: recovered %d rows", recovered
-            )
-        return recovered
+    return _recover_stalled_runs(
+        HeadAutoApplyRun,
+        stall_minutes=HEAD_AUTO_APPLY_STALL_THRESHOLD_MINUTES,
+        keep_runs=HEAD_AUTO_APPLY_KEEP_RUNS,
+        label="recover_stalled_head_auto_apply_runs",
+    )
 
 
 # Keep ~6 months of daily head-metric snapshots (enough to see tuning trends).

backend/app/tasks/ml.py

@@ -105,9 +105,7 @@ def embed_image(self, image_id: int) -> dict:
             record = session.get(ImageRecord, image_id)
             if record is None:
                 return {"status": "missing", "image_id": image_id}
-            settings = session.execute(
-                select(MLSettings).where(MLSettings.id == 1)
-            ).scalar_one()
+            settings = MLSettings.load_sync(session)
 
             src = Path(record.path)
             is_vid = _is_video(src)
@@ -488,15 +486,10 @@ def scheduled_ccip_auto_apply() -> str:
     from sqlalchemy import select as sa_select
     from sqlalchemy.dialects.postgresql import insert as pg_insert
 
-    from ..models import ImageRegion, MLSettings, Tag, TagKind, TagSuggestionRejection
+    from ..models import ImageRegion, MLSettings, Tag, TagKind
     from ..models.tag import image_tag
-
-    fig = ("face", "figure")
-
-    def _l2(m):
-        n = np.linalg.norm(m, axis=1, keepdims=True)
-        n[n == 0] = 1.0
-        return m / n
+    from ..services.ml.ccip import _FIGURE_KINDS
+    from ..services.ml.training_data import _applied_or_rejected, _l2norm
 
     SessionLocal = _sync_session_factory()
     with SessionLocal() as session:
@@ -521,7 +514,7 @@ def scheduled_ccip_auto_apply() -> str:
             )
             .join(Tag, Tag.id == image_tag.c.tag_id)
             .where(Tag.kind == TagKind.character)
-            .where(ImageRegion.kind.in_(fig))
+            .where(ImageRegion.kind.in_(_FIGURE_KINDS))
             .where(ImageRegion.ccip_embedding.is_not(None))
             .where(ImageRegion.image_record_id.in_(single))
         ).all()
@@ -532,29 +525,16 @@ def scheduled_ccip_auto_apply() -> str:
         for tid, vec in ref_rows:
             by_char.setdefault(tid, []).append(vec)
         ref_tags = list(by_char)
-        mats = [_l2(np.asarray(by_char[t], dtype=np.float32)) for t in ref_tags]
+        mats = [_l2norm(np.asarray(by_char[t], dtype=np.float32), np) for t in ref_tags]
         allref = np.vstack(mats)                                  # (total, 768)
         seg = np.cumsum([0] + [len(m) for m in mats])[:-1]        # per-char start
 
         # Per character: images that already carry OR rejected the tag — skip.
-        skip = {t: set() for t in ref_tags}
-        for t in ref_tags:
-            for (iid,) in session.execute(
-                sa_select(image_tag.c.image_record_id).where(
-                    image_tag.c.tag_id == t
-                )
-            ):
-                skip[t].add(iid)
-            for (iid,) in session.execute(
-                sa_select(TagSuggestionRejection.image_record_id).where(
-                    TagSuggestionRejection.tag_id == t
-                )
-            ):
-                skip[t].add(iid)
+        skip = _applied_or_rejected(session, ref_tags)
 
         img_ids = list(session.execute(
             sa_select(ImageRegion.image_record_id)
-            .where(ImageRegion.kind.in_(fig), ImageRegion.ccip_embedding.is_not(None))
+            .where(ImageRegion.kind.in_(_FIGURE_KINDS), ImageRegion.ccip_embedding.is_not(None))
             .distinct()
         ).scalars())
 
@@ -566,7 +546,7 @@ def scheduled_ccip_auto_apply() -> str:
                 sa_select(ImageRegion.image_record_id, ImageRegion.ccip_embedding)
                 .where(
                     ImageRegion.image_record_id.in_(chunk),
-                    ImageRegion.kind.in_(fig),
+                    ImageRegion.kind.in_(_FIGURE_KINDS),
                     ImageRegion.ccip_embedding.is_not(None),
                 )
             ).all()
@@ -574,7 +554,7 @@ def scheduled_ccip_auto_apply() -> str:
             for iid, vec in rows:
                 by_img.setdefault(iid, []).append(vec)
             for iid, vecs in by_img.items():
-                q = _l2(np.asarray(vecs, dtype=np.float32))      # (nq, 768)
+                q = _l2norm(np.asarray(vecs, dtype=np.float32), np)  # (nq, 768)
                 colmax = (q @ allref.T).max(axis=0)              # (total,)
                 charmax = np.maximum.reduceat(colmax, seg)       # (n_chars,)
                 for ci in np.where(charmax >= thr)[0]:

ci-requirements.md

@@ -9,14 +9,31 @@ git.fabledsword.com/bvandeusen/ci-python:3.14
 ## Image deps used
 
 - python 3.14
-- ruff (analyzer for `backend/`, `tests/`, `alembic/`)
+- ruff (analyzer for `backend/`, `tests/`, `alembic/`, `agent/`, `scripts/`)
 - node (frontend job: `npm install` + vitest + vite build)
-- docker CLI + buildx (`.forgejo/workflows/build.yml`: build-web, build-ml — Forgejo registry push)
+- docker CLI + buildx (`.forgejo/workflows/build.yml`: build-web, build-ml, build-agent — Fabled-Git registry push, and `imagetools inspect`/`create` for the reuse path)
+
+## Secondary runtime image
+
+node:24-bookworm-slim — `.forgejo/workflows/extension.yml` only.
+
+`.forgejo/workflows/release.yml` runs on `ci-python:3.14` like everything else
+and installs nothing: it needs git and stdlib python, and builds no image.
+
+The extension lane is the one job that does NOT run on `ci-python:3.14`: it
+needs a current Node for `web-ext` and vitest and nothing Python at all. Kept
+on the upstream slim image rather than adding a Node toolchain to `ci-python`,
+per `docs/process.md`'s "add deps to the image when used by >1 project".
 
 ## Per-job tool installs
 
 - `pip install -r requirements.txt pytest pytest-asyncio` — in `backend-lint-and-test` and `integration` jobs
 - `npm install --no-audit --no-fund` — in `frontend-build` job
+- `npm install --no-audit --no-fund` — in `extension.yml`'s `lint` job (web-ext + vitest)
+- `unzip` — in `extension.yml`'s "Verify XPI contents" step, installed via apt
+  only when absent (`node:24-bookworm-slim` may or may not carry it). Debian
+  package, ~2s. Not worth baking into a shared image for a single consumer, per
+  `docs/process.md`'s ">1 project" rule.
 
 ## Notes
 
@@ -26,12 +43,117 @@ git.fabledsword.com/bvandeusen/ci-python:3.14
   "add deps to image when used by >1 project" rule: FC alone is one Python
   project, so the deps live in `requirements.txt` and install per-job.
   Reconsider when a second Fabled-family Python backend lands.
-- Integration uses Forgejo Actions `services:` + socket-discovered bridge IPs
+- Integration uses Fabled-Git Actions `services:` + socket-discovered bridge IPs
   because `act_runner` (swarm-runner v0.6+) puts services on the default
   bridge with no embedded DNS. The pattern is documented in the rulebook's
-  `forgejo.md` "CI philosophy" section and FC's `ci.yml` is the canonical
+  `fabled-git.md` "CI philosophy" section and FC's `ci.yml` is the canonical
   example.
 - No `package-lock.json` is tracked yet (FC's `feedback_no_local_runs`
   memory bans `npm install` locally). Using `npm install` rather than
   `npm ci` until a lockfile lands.
 - No `imagemagick` / `pandoc` per-job installs needed.
+- `extension/`'s vitest specs load `lib/*.js` by evaluating the real file as a
+  classic script (`test/helpers/loadLib.js`) rather than adding `module.exports`
+  shims to production code — the libs ship as `background.scripts`, not ES
+  modules, so the specs exercise exactly the bytes packaged into the XPI.
+- **`extension/scripts/packaging.sh` is the single definition of what ships
+  inside the XPI.** Three consumers read from it rather than keeping their own
+  copy: web-ext's `--ignore-files` (`extension/package.json`), the `git log`
+  pathspec inside the script's own version derivation, and `scripts/artifacts.sh`,
+  which appends the extension's set to web's because the web image bundles the
+  signed XPI. Hand-kept copies of that one fact is what allowed issue #2397, so
+  `extension/test/version.spec.js` asserts no workflow has reintroduced a
+  literal `:(exclude)extension/…`.
+- **Packaged and version-relevant are two different sets** (#3156). `scripts/`
+  is excluded from the XPI and is NOT excluded from the version derivation,
+  because `packaging.sh` decides the version string stamped into the packaged
+  `manifest.json`. The membership test is *"can changing this file change the
+  published bytes?"*, not *"is this file copied in?"* — which is why the script
+  keeps two lists rather than one.
+- **The shipped extension version is derived, not committed.** It is the commit
+  TIME of the newest packaged-extension change, rendered `YYYY.M.D.HHMM` UTC
+  (family rules 148/149 — never a commit count, which orders by branch rather
+  than by recency). `build.yml`'s `sign-extension` computes it and stamps it
+  into `extension/manifest.json` + `package.json` in the working tree before
+  signing; the stamp is never committed. The version in the repo is **wholly
+  inert** — since milestone 318 step 8 there is no hand-set MAJOR.MINOR either.
+- **The extension is the one artifact that does not zero-pad, and that is not a
+  drift** (#3138). Mozilla's grammar for AMO is
+  `^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$` — a segment is the single
+  digit `0` or starts 1-9, and there are at most four. `2026.08.29.0201` is
+  rejected; `2026.8.29.201` is the same value one character narrower per
+  segment, and rule 148 defines comparison as numeric per segment, so nothing is
+  reordered. `ci.yml`'s `extension-version` lane asserts the derived string
+  against that exact regex, plus a `YYYY.M.D.HHMM` shape check that would catch
+  a regression to the pre-318 `1.0.<minutes>` — which AMO would accept and which
+  orders below everything already signed. Checking here is the whole point: AMO
+  409s on re-signing, so a version it rejects is burned and cannot be reused.
+  `scripts/artifacts.sh version extension` **delegates** to `packaging.sh` so
+  the two cannot answer differently.
+- Every job that derives anything checks out with `fetch-depth: 0` — all four
+  `build.yml` jobs, `ci.yml`'s `extension-version` and `backend-lint-and-test`
+  (for `tests/test_artifact_paths.py` and `test_artifact_identity.py`), and
+  `release.yml`, which additionally walks the tag graph. A depth-1 clone sees
+  one commit and derives a wrong, too-low value **rather than failing**, so the
+  full-history checkout is load-bearing rather than incidental.
+- **`scripts/artifacts.sh` is the same shape one level up: one definition per
+  artifact of what it is built from, and the two values derived from it.**
+  `revision` (12 hex of the newest commit touching that set) and `version`
+  (`YYYY.MM.DD.HHMM` UTC, rule 148). Four artifacts, four independent answers,
+  so a push touching only `agent/` leaves web and ml alone.
+  `tests/test_artifact_paths.py` reads each Dockerfile and asserts every COPY
+  source is covered, so adding a COPY without updating the script fails CI.
+- **A file that DECIDES an artifact's identity belongs in its set even though it
+  is copied into nothing** — `packaging.sh` for the extension and web (#3156),
+  and `artifacts.sh` itself for web (#3202), which decides the `FC_VERSION`
+  baked into that image. Only web needs the second entry: every artifact stamps
+  a revision, but a revision has a backstop (a changed derivation stops matching
+  the published label and forces a rebuild) and a version has none, because
+  nothing compares it to anything. `tests/test_artifact_paths.py`'s `DERIVERS`
+  table is the guard.
+- **Builds are skipped when the content is already published.** Each image
+  carries its revision as an `fc.revision` LABEL, and `build.yml` reads that
+  label back off the moving channel tag (`imagetools inspect --format`). Equal
+  to the derived revision means the bytes are already published, so the job
+  repoints the remaining tags at the existing manifest instead of rebuilding.
+  Two things this depends on: an inspect that errors for ANY reason reads as a
+  MISS so no needed build is ever skipped, and the repoint must EXCLUDE the
+  source tag — `imagetools create` wraps its source in a manifest index, and
+  config labels do not resolve through an index, so writing the channel tag
+  from itself destroys the label the next run reads (#3183).
+- **The build pushes exactly ONE tag — the channel's — and every other tag is
+  written registry-side afterwards** (#3190). buildx on this runner pushes the
+  first tag to the registry and then re-pushes the rest through the docker
+  driver, out of a local image store that a registry-direct build never fills;
+  it fails intermittently with `tag does not exist`. On `dev` that only reddens
+  a job, but on `main` it silently skips `:c-<sha>` while `:latest` publishes
+  fine — a missing rollback tag has no consumer that fails, so nothing but the
+  red job would notice until somebody needs to roll back. `imagetools create`
+  has no local store to be absent from, and it is the code the reuse path
+  already ran, so both paths now share one proven route. The cost: `:c-<sha>`
+  is an index rather than a plain image, so `fc.revision` does not resolve
+  through it — nothing reads it there, and the index names the same manifest.
+- **`FC_CHANNEL` and `FC_VERSION` are build args, not runtime settings.**
+  `build.yml` passes them to the web image only — the ml and agent images have
+  nothing to report them to. `/api/health` returns both, the foot of Settings
+  renders them, and `/api/extension/manifest` reports the channel beside the
+  extension version so an install can be traced to a channel. With no version image tags, that
+  self-report is the ONLY answer to "which build is this?" — which is why a
+  missing version renders `unknown` rather than a blank: an empty footer reads
+  as "no version", a different and false claim.
+  Both are declared LAST in the Dockerfile on purpose: an ARG invalidates every
+  layer below it, and these are the values that differ between the dev and main
+  builds of identical source, so placing them earlier would stop the two
+  channels ever sharing a cached `pip install`. Empty by default — a local build
+  then reports nothing rather than claiming a channel it is not on.
+- **The channel is never folded into the version.** A `-dev` suffix makes the
+  extension's per-segment `parseInt` comparator read that segment as 0, so every
+  dev build compares equal to every other — issue #2993 exactly (rule 149).
+  `frontend/test/systemBuild.spec.js` pins the rendered version to the bare
+  number.
+- Callers MUST `set -f` before substituting the script's output. Without it the
+  shell expands `test/**` against the working tree and silently narrows the
+  pattern to whatever files exist at that moment — a failure that looks like
+  nothing until dev files start appearing in the XPI. `test/version.spec.js`
+  asserts every `--ignore-files` consumer sets it, and that no consumer has
+  quietly reinstated a hardcoded list.

extension/README.md

@@ -1,13 +1,14 @@
 # FabledCurator Firefox Extension
 
 Self-hosted Firefox extension that pushes session cookies from supported
-platforms (Patreon, SubscribeStar, Hentai-Foundry, Discord, Pixiv,
-DeviantArt) into FabledCurator, and lets you add a creator as a Source
-from their page in one click.
+platforms (Patreon, SubscribeStar, Hentai-Foundry, Discord, Pixiv)
+into FabledCurator, and lets you add a creator as a Source from their
+page in one click.
 
 ## Install (operator)
 
-The signed XPI is bundled into the FC Docker image. Open FC →
+The signed XPI is bundled into the FC Docker image — `:dev` and
+`:latest` each carry their own channel's build. Open FC →
 Settings → Maintenance → Browser extension → click "Install Firefox
 extension". Firefox shows its native install prompt. After installing,
 open the extension's options page (about:addons → FabledCurator →
@@ -20,6 +21,7 @@ same card.
 cd extension/
 npm install --no-save        # web-ext only
 npm run lint                 # web-ext lint
+npm run test:unit            # vitest — lib/ logic + packaging/version checks
 npm run start                # launches Firefox with extension loaded
 npm run build                # unsigned XPI in web-ext-artifacts/
 ```
@@ -36,10 +38,76 @@ npm run build                # unsigned XPI in web-ext-artifacts/
 - [ ] Subscriptions list: popup → "Sources" tab → list renders
 - [ ] Check now: click play icon on source row → no error toast
 
+## Versioning — the committed number decides nothing
+
+The shipped version is **derived**, not committed. `scripts/packaging.sh
+version` returns `YYYY.M.D.HHMM` in UTC: the commit *time* of the newest change
+to a packaged extension file. `build.yml` computes it and stamps it into both
+`manifest.json` and `package.json` at build time. The stamp is never
+committed — the commit carrying it would itself be a change to the extension,
+which would move the version again.
+
+So:
+
+- **Editing the version does nothing.** All of it is overwritten before web-ext
+  ever reads it. There is no bump to make, and none to forget. There is no
+  hand-set part left either: MAJOR.MINOR went away with milestone 318 step 8.
+- `npm run build` locally produces an XPI labelled with the *committed*
+  version, since nothing stamped it. Fine for loading into a test profile; not
+  what ships.
+
+**Why the extension is the one artifact that does not zero-pad.** Every other
+FC artifact emits rule 148's `YYYY.MM.DD.HHMM`. AMO will not take it: Mozilla's
+grammar for addons.mozilla.org is
+
+```
+^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$
+```
+
+— each segment is the single digit `0` or starts 1-9, so `08` and `0201` are
+rejected, and at most four segments are allowed. The extension therefore emits
+**the same numbers unpadded**: `2026.8.29.201` where the rest of the family
+says `2026.08.29.0201`. Rule 148 already defines comparison as numeric per
+segment, under which the two are equal, so nothing is reordered by the choice
+and left-padding each segment recovers the family string exactly. `ci.yml`'s
+`extension-version` lane checks the derived string against that regex on every
+push — the cheap place to find out, because AMO 409s on re-signing and a
+rejected version is burned for good.
+
+Why commit time and not a commit count: a count is per-branch, so `dev` and
+`main` count different histories of the same code and their versions end up
+ordered by which branch accumulated more commits rather than by which is newer.
+Commit time gives both branches the same number for the same source — which is
+exactly what lets one AMO signature serve both channels (family rule 149, FC
+issue #3092).
+
+## Channels
+
+`dev` and `main` each build and sign their own extension, and an install is
+tied to whichever FC instance it points at — Firefox's static `update_url`
+cannot apply here, since every FC install is a different host, so the extension
+asks its configured backend. **The channel therefore IS the instance.**
+Switching channel means repointing the FC URL in options and reinstalling from
+that host; there is no separate channel setting, and adding one would
+contradict each server build shipping its own extension.
+
+The channel is reported *beside* the version, never inside it:
+`/api/extension/manifest` answers `{"version": "...", "channel": "dev"}`. It is
+optional — an instance that declares none simply omits the key, and the popup,
+the toolbar tooltip and the Settings card all read exactly as they did before
+the field existed. Do not be tempted to make it a `-dev` version suffix: the
+comparator parses each dotted segment with `parseInt`, so a suffixed segment
+reads as 0 and every dev build compares equal to every other, collapsing "no
+update available" and "I cannot read this version" into one answer.
+
 ## Release
 
-Bump `manifest.json` + `package.json` SemVer (both files) and commit
-under `extension/**`. The `.forgejo/workflows/extension.yml` workflow
-runs `web-ext sign` on main, commits the signed XPI to
-`frontend/public/extension/`, and the next FC server build bundles it
-into the Docker image.
+Nothing to do by hand. Push to `dev`: `build.yml` signs the extension if this
+change moved the version, caches the signed XPI as a Forgejo `ext-<version>`
+release, and bundles it into `fabledcurator:dev`. Merging to `main` derives the
+same version, hits that cache, and bundles the byte-identical XPI into
+`:latest` with no second AMO call.
+
+AMO refuses to re-sign a version it has already issued, so signing is one-shot
+per version — which is why the cache exists and why the version must never move
+backwards.

extension/background/background.js

@@ -37,7 +37,16 @@ ensureInitialized().catch(e => console.error('init failed:', e));
 // configured backend for the latest published version and nudge the operator to
 // reinstall the freshly-signed XPI — surfaced as a popup banner (on demand) and
 // a toolbar badge (daily). /api/extension/manifest is public and returns
-// {version, latest_url, sha256}; the XPI is served from the web root (not /api).
+// {version, latest_url, sha256} plus an OPTIONAL {channel} naming which channel
+// that instance serves ("dev"/"main", #3113); the XPI is served from the web
+// root (not /api).
+//
+// The channel IS the instance: Firefox's static update_url cannot apply here
+// because every FC install is a different host, so the extension asks its
+// configured backend — which means switching channel is repointing apiUrl in
+// options and reinstalling from that host. There is no separate channel
+// setting to build, and building one would contradict each server build
+// shipping its own extension.
 
 function versionIsNewer(candidate, current) {
   // Dotted numeric compare so 1.0.10 > 1.0.9 (a plain string compare wouldn't).
@@ -60,13 +69,22 @@ async function checkForUpdateInfo() {
   }
   const currentVersion = browser.runtime.getManifest().version;
   const latestVersion = info && info.version ? info.version : null;
-  // latest_url is served from the web root; strip the /api suffix off baseUrl
-  // (same transform as OPEN_ARTIST_PAGE).
-  const base = (api.baseUrl || '').replace(/\/+$/, '').replace(/\/api$/, '');
+  // Which channel the configured instance serves — reported ALONGSIDE the
+  // version, never folded into it. A `-dev` suffix would have to survive
+  // versionIsNewer's parseInt above, and it wouldn't: the segment would read
+  // as 0 and every dev build would compare equal to every other.
+  //
+  // null is a normal answer, not a failure — an instance built before the
+  // field existed, or one built locally with no channel declared. Nothing
+  // below branches on it except the label.
+  const channel = info && info.channel ? info.channel : null;
+  // latest_url is served from the web root, not the JSON API.
+  const base = api.webRoot();
   return {
     updateAvailable: !!latestVersion && versionIsNewer(latestVersion, currentVersion),
     currentVersion,
     latestVersion,
+    channel,
     xpiUrl: info && info.latest_url ? `${base}${info.latest_url}` : null,
   };
 }
@@ -78,7 +96,11 @@ async function refreshUpdateBadge() {
     await browser.action.setBadgeText({ text: r.updateAvailable ? '↑' : '' });
     if (r.updateAvailable) {
       await browser.action.setBadgeBackgroundColor({ color: '#F4BA7A' });
-      await browser.action.setTitle({ title: `FabledCurator — update available (v${r.latestVersion})` });
+      // Channel first, version second, and the channel dropped entirely when
+      // the instance doesn't report one — so the tooltip reads exactly as it
+      // did before the field existed rather than saying "(unknown ...)".
+      const label = r.channel ? `${r.channel} v${r.latestVersion}` : `v${r.latestVersion}`;
+      await browser.action.setTitle({ title: `FabledCurator — update available (${label})` });
     } else {
       await browser.action.setTitle({ title: 'FabledCurator' });
     }
@@ -211,6 +233,21 @@ browser.webRequest.onBeforeRedirect.addListener(
   { urls: ['https://app-api.pixiv.net/web/v1/users/auth/pixiv/callback*'] },
 );
 
+// Extract → verify → upload one cookie-auth platform. Returns a structured
+// outcome so the two callers (EXPORT_COOKIES single, EXPORT_ALL_COOKIES) shape
+// their own response + skip semantics. Verifies the captured cookies are
+// actually live BEFORE uploading, so a confirmed-stale session doesn't overwrite
+// good FC-side credentials; platforms with no verify config (v.ok === null) fall
+// through to upload.
+async function exportPlatformCookies(key) {
+  const cookies = await extractCookiesForPlatform(key);
+  if (cookies.length === 0) return { status: 'empty' };
+  const v = await verifyCookiesForPlatform(key);
+  if (v.ok === false) return { status: 'stale', reason: v.reason, cookieCount: cookies.length };
+  await api.uploadCredentials(key, 'cookies', toNetscapeFormat(cookies));
+  return { status: 'ok', cookieCount: cookies.length, verified: v.ok === true };
+}
+
 // ---- Message router ----
 
 browser.runtime.onMessage.addListener(async (msg) => {
@@ -255,22 +292,14 @@ browser.runtime.onMessage.addListener(async (msg) => {
       if (!platform) return { error: `Unknown platform: ${key}` };
       try {
         if (platform.authType === 'cookies') {
-          const cookies = await extractCookiesForPlatform(key);
-          if (cookies.length === 0) return { error: 'No cookies found — log in first.' };
-          // Verify the captured cookies are actually live BEFORE
-          // uploading. Skips upload on confirmed-stale sessions so we
-          // don't overwrite FC-side credentials with garbage. Platforms
-          // without a verify config (verify.ok === null) fall through
-          // to upload as before.
-          const v = await verifyCookiesForPlatform(key);
-          if (v.ok === false) {
+          const r = await exportPlatformCookies(key);
+          if (r.status === 'empty') return { error: 'No cookies found — log in first.' };
+          if (r.status === 'stale') {
             return {
-              error: `Captured ${cookies.length} ${platform.name} cookies but they don't appear authenticated (${v.reason}). Log in again in this browser, then retry.`,
+              error: `Captured ${r.cookieCount} ${platform.name} cookies but they don't appear authenticated (${r.reason}). Log in again in this browser, then retry.`,
             };
           }
-          const data = toNetscapeFormat(cookies);
-          await api.uploadCredentials(key, 'cookies', data);
-          return { success: true, cookieCount: cookies.length, verified: v.ok === true };
+          return { success: true, cookieCount: r.cookieCount, verified: r.verified };
         }
         if (key === 'discord') {
           if (!discordToken) return { error: 'Open discord.com to capture a token first.' };
@@ -298,18 +327,10 @@ browser.runtime.onMessage.addListener(async (msg) => {
           continue;
         }
         try {
-          const cookies = await extractCookiesForPlatform(key);
-          if (cookies.length === 0) {
-            results[key] = { skipped: true, reason: 'no cookies' };
-            continue;
-          }
-          const v = await verifyCookiesForPlatform(key);
-          if (v.ok === false) {
-            results[key] = { error: `verify failed: ${v.reason}` };
-            continue;
-          }
-          await api.uploadCredentials(key, 'cookies', toNetscapeFormat(cookies));
-          results[key] = { success: true, cookieCount: cookies.length, verified: v.ok === true };
+          const r = await exportPlatformCookies(key);
+          if (r.status === 'empty') results[key] = { skipped: true, reason: 'no cookies' };
+          else if (r.status === 'stale') results[key] = { error: `verify failed: ${r.reason}` };
+          else results[key] = { success: true, cookieCount: r.cookieCount, verified: r.verified };
         } catch (e) {
           results[key] = { error: e.message };
         }
@@ -346,11 +367,9 @@ browser.runtime.onMessage.addListener(async (msg) => {
       }
 
     case 'OPEN_ARTIST_PAGE': {
-      // apiUrl is configured with the /api suffix (see
-      // options/options.html placeholder); the SPA artist route is
-      // /artist/:slug, served from the same origin. Strip /api so the
-      // browser-level URL hits the Vue router, not the JSON API.
-      const base = (api.baseUrl || '').replace(/\/+$/, '').replace(/\/api$/, '');
+      // The SPA artist route (/artist/:slug) is served from the web root, not
+      // the JSON API — see api.webRoot().
+      const base = api.webRoot();
       const slug = encodeURIComponent(msg.slug || '');
       if (!base || !slug) return { error: 'apiUrl or slug missing' };
       try {

extension/lib/api.js

@@ -11,7 +11,10 @@ class FabledCuratorAPI {
 
   async init() {
     const cfg = await browser.storage.local.get(['apiUrl', 'apiKey']);
-    this.baseUrl = cfg.apiUrl || null;
+    // Normalize on READ, not just on save: configs stored before the options
+    // page started normalizing are missing the `/api` suffix, and this heals
+    // them without the operator having to reopen Settings.
+    this.baseUrl = normalizeApiUrl(cfg.apiUrl) || null;
     this.apiKey = cfg.apiKey || null;
     return this.isConfigured();
   }
@@ -50,6 +53,13 @@ class FabledCuratorAPI {
       } catch {
         message = `HTTP ${response.status}: ${response.statusText}`;
       }
+      // 404/405 from FC almost always means the request never reached the JSON
+      // API — it fell through to the SPA catch-all, which serves HTML on GET
+      // and rejects everything else. Say so, rather than making the operator
+      // decode "Method Not Allowed" on an endpoint that plainly allows POST.
+      if (response.status === 404 || response.status === 405) {
+        message += ` — ${url} isn't the FC API. Check the FC URL in settings.`;
+      }
       const err = new Error(message);
       err.status = response.status;
       throw err;
@@ -96,6 +106,12 @@ class FabledCuratorAPI {
     return this.request('GET', '/extension/manifest');
   }
 
+  // The web/SPA root: where the Vue router (artist pages) and the served XPI
+  // live, NOT the JSON API. Used by OPEN_ARTIST_PAGE + the self-update check.
+  webRoot() {
+    return webRootFromApiUrl(this.baseUrl);
+  }
+
   // Connection test = the cheapest read with auth.
   testConnection() {
     return this.request('GET', '/credentials');

extension/lib/platforms.js

@@ -68,16 +68,6 @@ const PLATFORMS = {
     urlPattern: /^https?:\/\/(www\.)?pixiv\.net/,
     note: 'Click to authenticate via OAuth',
   },
-  deviantart: {
-    name: 'DeviantArt',
-    domains: ['.deviantart.com', 'www.deviantart.com', 'deviantart.com'],
-    authType: 'cookies',
-    color: '#05CC47',
-    urlPattern: /^https?:\/\/(www\.)?deviantart\.com/,
-    // DA's logged-in-only endpoints sit behind their internal _napi
-    // namespace which shifts; skipping verify until a stable check
-    // surfaces. Same posture as SubscribeStar.
-  },
 };
 
 /**
@@ -98,7 +88,6 @@ const PLATFORM_ARTIST_PATTERNS = {
   patreon:       /^https?:\/\/(www\.)?patreon\.com\/(?:cw\/|c\/)?(?!(?:home|search|messages|notifications|library|settings|posts)(?:[\/?#]|$))[^/?#]+/i,
   subscribestar: /^https?:\/\/(www\.)?subscribestar\.(com|adult)\/(?!feed$|messages$|library$)[^/?#]+\/?$/i,
   hentaifoundry: /^https?:\/\/(www\.)?hentai-foundry\.com\/user\/[^/?#]+/i,
-  deviantart:    /^https?:\/\/(www\.)?deviantart\.com\/(?!home$|watch\b|tag\b|browse\b)[^/?#]+\/?$/i,
   pixiv:         /^https?:\/\/(www\.)?pixiv\.net\/(en\/)?users\/\d+/i,
 };
 

extension/lib/url.js

@@ -0,0 +1,32 @@
+/**
+ * Canonical FC endpoint derivation, shared by the background client and the
+ * options page so a URL entered either way behaves identically.
+ *
+ * FC serves two things on one origin: the JSON API under `/api`, and the Vue
+ * SPA from the root. `api.js` builds requests as `${baseUrl}/credentials`, so
+ * the stored base URL has to carry the `/api` suffix.
+ */
+
+/**
+ * Accept what an operator would naturally type — the instance root
+ * (`http://curator.example.com`) or the API root (`.../api`) — and return the
+ * API root either way.
+ *
+ * Worth normalizing rather than validating: a root-form URL doesn't fail
+ * loudly, it lands on the SPA catch-all, which answers `GET /credentials` with
+ * 200 HTML and rejects `POST /credentials` with 405. The operator sees a
+ * working Test Connection and a broken export.
+ */
+function normalizeApiUrl(raw) {
+  const trimmed = (raw || '').trim().replace(/\/+$/, '');
+  if (!trimmed) return '';
+  return /\/api$/i.test(trimmed) ? trimmed : `${trimmed}/api`;
+}
+
+/**
+ * The SPA root — where the Vue router (artist pages) and the served XPI live,
+ * NOT the JSON API. Accepts either input form, same as normalizeApiUrl.
+ */
+function webRootFromApiUrl(raw) {
+  return normalizeApiUrl(raw).replace(/\/api$/i, '');
+}

extension/manifest.json

@@ -1,7 +1,7 @@
 {
   "manifest_version": 3,
   "name": "FabledCurator",
-  "version": "1.0.9",
+  "version": "1.0.11",
   "description": "Export cookies from supported platforms to FabledCurator and add creators as sources in one click.",
 
   "browser_specific_settings": {
@@ -33,7 +33,6 @@
     "*://*.hentai-foundry.com/*",
     "*://*.discord.com/*",
     "*://*.pixiv.net/*",
-    "*://*.deviantart.com/*",
     "*://app-api.pixiv.net/*",
     "*://oauth.secure.pixiv.net/*",
     "*://*/*"
@@ -46,7 +45,7 @@
   },
 
   "background": {
-    "scripts": ["lib/platforms.js", "lib/cookies.js", "lib/api.js", "background/background.js"]
+    "scripts": ["lib/platforms.js", "lib/cookies.js", "lib/url.js", "lib/api.js", "background/background.js"]
   },
 
   "options_ui": {
@@ -61,7 +60,6 @@
         "*://*.subscribestar.com/*",
         "*://*.subscribestar.adult/*",
         "*://*.hentai-foundry.com/*",
-        "*://*.deviantart.com/*",
         "*://*.pixiv.net/*"
       ],
       "js": ["lib/platforms.js", "content/content-script.js"],

extension/options/options.html

@@ -21,9 +21,12 @@
 <body>
   <h1>FabledCurator extension</h1>
 
-  <label for="api-url">FC base URL</label>
-  <input id="api-url" type="url" placeholder="http://curator.example.com/api" />
-  <div class="hint">Find this on FC → Settings → Maintenance → Browser extension.</div>
+  <label for="api-url">FC instance URL</label>
+  <input id="api-url" type="url" placeholder="http://curator.example.com" />
+  <div class="hint">
+    Your FabledCurator address — with or without the trailing <code>/api</code>; both work.
+    Find it on FC → Settings → Maintenance → Browser extension.
+  </div>
 
   <label for="api-key">Extension API key</label>
   <input id="api-key" type="password" placeholder="paste from FC Settings card" />
@@ -36,6 +39,7 @@
 
   <div id="status" class="status" style="display:none;"></div>
 
+  <script src="../lib/url.js"></script>
   <script src="options.js"></script>
 </body>
 </html>

extension/options/options.js

@@ -8,7 +8,7 @@ document.addEventListener('DOMContentLoaded', async () => {
 });
 
 async function save() {
-  const apiUrl = document.getElementById('api-url').value.trim().replace(/\/+$/, '');
+  const apiUrl = normalizeApiUrl(document.getElementById('api-url').value);
   const apiKey = document.getElementById('api-key').value.trim();
   if (!apiUrl || !apiKey) {
     showStatus('Both fields are required.', 'err');
@@ -16,11 +16,14 @@ async function save() {
   }
   await browser.storage.local.set({ apiUrl, apiKey });
   await browser.storage.local.remove(['lastConnectionTest', 'lastConnectionStatus']);
-  showStatus('Saved.', 'ok');
+  // Show what was actually stored — the operator may have typed the instance
+  // root and it was normalized to the API root.
+  document.getElementById('api-url').value = apiUrl;
+  showStatus(`Saved — using ${apiUrl}`, 'ok');
 }
 
 async function test() {
-  const apiUrl = document.getElementById('api-url').value.trim().replace(/\/+$/, '');
+  const apiUrl = normalizeApiUrl(document.getElementById('api-url').value);
   const apiKey = document.getElementById('api-key').value.trim();
   if (!apiUrl || !apiKey) {
     showStatus('Fill both fields first.', 'err');
@@ -31,8 +34,23 @@ async function test() {
       method: 'GET',
       headers: { 'X-Extension-Key': apiKey },
     });
-    if (r.ok) showStatus(`Connected — HTTP ${r.status}.`, 'ok');
-    else showStatus(`HTTP ${r.status}: ${r.statusText}`, 'err');
+    if (!r.ok) {
+      showStatus(`HTTP ${r.status}: ${r.statusText}`, 'err');
+      return;
+    }
+    // A 200 is NOT sufficient. If the URL resolves to the Vue SPA instead of
+    // the JSON API, the catch-all route returns 200 with an HTML document —
+    // which used to report "Connected" on a config that could not POST at all.
+    const contentType = r.headers.get('content-type') || '';
+    if (!contentType.includes('json')) {
+      showStatus(
+        `${apiUrl} answered with ${contentType || 'no content-type'}, not JSON `
+          + '— that looks like the FC web UI rather than its API.',
+        'err',
+      );
+      return;
+    }
+    showStatus(`Connected to ${apiUrl} — HTTP ${r.status}.`, 'ok');
   } catch (e) {
     showStatus(`Cannot reach ${apiUrl}: ${e.message}`, 'err');
   }

extension/package.json

@@ -1,15 +1,18 @@
 {
   "name": "fabledcurator-extension",
-  "version": "1.0.9",
+  "version": "1.0.11",
   "private": true,
   "description": "Firefox extension for FabledCurator",
+  "comment_ignore_files": "The --ignore-files list comes from scripts/packaging.sh, the single source of truth shared with ci.yml's guard and the derived-version patch count. `set -f` is REQUIRED before the substitution: without it the shell globs `test/**` against the working tree and silently narrows the pattern to whatever files happen to exist.",
   "scripts": {
-    "lint":  "web-ext lint --source-dir=. --no-config-discovery --ignore-files package.json package-lock.json web-ext-artifacts node_modules README.md .gitignore",
-    "start": "web-ext run --source-dir=. --no-config-discovery --ignore-files package.json package-lock.json web-ext-artifacts node_modules README.md .gitignore --firefox=firefox",
-    "build": "web-ext build --source-dir=. --no-config-discovery --ignore-files package.json package-lock.json web-ext-artifacts node_modules README.md .gitignore --overwrite-dest",
-    "sign":  "web-ext sign --source-dir=. --no-config-discovery --ignore-files package.json package-lock.json web-ext-artifacts node_modules README.md .gitignore --channel=unlisted --api-key=$WEB_EXT_API_KEY --api-secret=$WEB_EXT_API_SECRET"
+    "lint":  "set -f; web-ext lint --source-dir=. --no-config-discovery --ignore-files $(sh scripts/packaging.sh ignore)",
+    "start": "set -f; web-ext run --source-dir=. --no-config-discovery --ignore-files $(sh scripts/packaging.sh ignore) --firefox=firefox",
+    "build": "set -f; web-ext build --source-dir=. --no-config-discovery --ignore-files $(sh scripts/packaging.sh ignore) --overwrite-dest",
+    "sign":  "set -f; web-ext sign --source-dir=. --no-config-discovery --ignore-files $(sh scripts/packaging.sh ignore) --channel=unlisted --api-key=$WEB_EXT_API_KEY --api-secret=$WEB_EXT_API_SECRET",
+    "test:unit": "vitest run"
   },
   "devDependencies": {
+    "vitest": "^4.0.0",
     "web-ext": "^10.0.0"
   }
 }

extension/popup/popup.js

@@ -2,6 +2,15 @@ document.addEventListener('DOMContentLoaded', init);
 
 const CONNECTION_TEST_INTERVAL = 2 * 60 * 1000;
 
+// A centered muted note div — the loading / empty state shared by the platform
+// and sources lists.
+function mutedNote(text) {
+  const d = document.createElement('div');
+  d.style.cssText = 'text-align:center;padding:18px;color:var(--on-surface-variant);';
+  d.textContent = text;
+  return d;
+}
+
 async function init() {
   try {
     const cfg = await browser.runtime.sendMessage({ type: 'GET_CONFIG' });
@@ -38,10 +47,7 @@ function showSetupRequired() {
 function showPlatformsLoading() {
   const c = document.getElementById('platforms-list');
   c.textContent = '';
-  const d = document.createElement('div');
-  d.style.cssText = 'text-align:center;padding:18px;color:var(--on-surface-variant);';
-  d.textContent = 'Loading platforms…';
-  c.appendChild(d);
+  c.appendChild(mutedNote('Loading platforms…'));
 }
 
 async function testConnectionIfNeeded() {
@@ -75,8 +81,12 @@ async function checkForUpdate() {
 }
 
 function showUpdateBanner(r) {
+  // The channel names itself beside the version, never inside it (#3113).
+  // Absent when the instance doesn't report one, and the banner then reads
+  // exactly as it did before the field existed.
+  const channel = r.channel ? ` (${r.channel})` : '';
   document.getElementById('update-text').textContent =
-    `Update available — v${r.latestVersion} (installed v${r.currentVersion})`;
+    `Update available${channel} — v${r.latestVersion} (installed v${r.currentVersion})`;
   // Opening the signed XPI triggers Firefox's native install prompt.
   document.getElementById('update-btn').addEventListener('click', () => {
     browser.tabs.create({ url: r.xpiUrl });
@@ -183,10 +193,7 @@ async function exportAllCookies() {
 async function loadSources() {
   const c = document.getElementById('sources-list');
   c.textContent = '';
-  const d = document.createElement('div');
-  d.style.cssText = 'text-align:center;padding:18px;color:var(--on-surface-variant);';
-  d.textContent = 'Loading sources…';
-  c.appendChild(d);
+  c.appendChild(mutedNote('Loading sources…'));
   const r = await browser.runtime.sendMessage({ type: 'LIST_SOURCES' });
   c.textContent = '';
   if (r.error) {
@@ -197,10 +204,7 @@ async function loadSources() {
     return;
   }
   if (!r.sources || r.sources.length === 0) {
-    const empty = document.createElement('div');
-    empty.style.cssText = 'text-align:center;padding:18px;color:var(--on-surface-variant);';
-    empty.textContent = 'No sources yet.';
-    c.appendChild(empty);
+    c.appendChild(mutedNote('No sources yet.'));
     return;
   }
   for (const src of r.sources) c.appendChild(createSourceRow(src));

extension/scripts/packaging.sh

@@ -0,0 +1,177 @@
+#!/bin/sh
+# Single source of truth for "what ships inside the XPI", plus the version
+# derived from it.
+#
+# Three consumers used to hand-maintain their own copy of this list, and
+# keeping three copies of one fact in sync by hand is how issue #2397 happened:
+#
+#   1. web-ext's --ignore-files      (extension/package.json's four scripts)
+#   2. the :(exclude) pathspec       (what moves the version — a WIDER
+#                                     set than the ignore list; see
+#                                     NOT_VERSION_RELEVANT)
+#   3. the git-log pathspec          (the derived version, below)
+#
+# They now all read from here. POSIX sh only — CI's run shell is busybox.
+#
+# -f (no pathname expansion) is set for the whole script and is load-bearing:
+# the lists below are iterated with deliberate word-splitting, and without -f
+# the shell would also GLOB them, expanding `test/**` into whatever files
+# happen to exist and corrupting the output. A caller's own `set -f` does not
+# help here — this runs as a separate sh process and does not inherit it.
+# Callers still need their own `set -f` for the substituted result; the two
+# guards protect different expansions.
+set -euf
+
+# Paths under extension/ that are NOT packaged into the XPI.
+#
+# Split by whether git tracks them: node_modules and web-ext-artifacts are
+# build/dependency output that never appears in a commit, so they belong in
+# web-ext's ignore list but would be meaningless in a git pathspec.
+#
+# Directories need BOTH forms. `test/**` matches the files inside, but not the
+# directory entry itself — web-ext writes an entry for the directory too, so
+# with only the glob the XPI ends up carrying empty `test/` and `scripts/`
+# entries (caught by the XPI-content check on 2026-08-03). The bare name alone
+# is not enough either: minimatch's `test` does not match `test/url.spec.js`,
+# so dropping the glob would ship the contents. Keep both.
+NOT_PACKAGED_TRACKED='package.json package-lock.json README.md .gitignore vitest.config.js scripts scripts/** test test/**'
+NOT_PACKAGED_BUILD='web-ext-artifacts node_modules'
+
+# Paths under extension/ that cannot change the SHIPPED BYTES, and so must not
+# move the derived version.
+#
+# Deliberately NOT the same list as NOT_PACKAGED_TRACKED, and the whole
+# difference is `scripts/`. packaging.sh is not packaged into the XPI — but it
+# DECIDES the version string, and build.yml stamps that string into the
+# manifest.json that is packaged. A change to how the version is computed is
+# therefore a change to the shipped bytes.
+#
+# Excluding it was harmless only while every push rebuilt the web image.
+# Milestone 313 step 4 made the rebuild conditional on the derived revision
+# moving, which turned it into a silent failure: a packaging.sh change gives a
+# NEW version, so sign-extension misses its ext-<version> cache and signs —
+# while build-web sees an unmoved revision, reuses the published image, and
+# ships the OLD XPI. An orphaned AMO signature, and an instance quietly serving
+# code the registry says is current.
+#
+# The two directions are not symmetric, which is why this list is the narrower
+# one. Too wide costs a re-sign and a rebuild for a change that ships nothing
+# new. Too narrow serves stale bytes and says nothing.
+NOT_VERSION_RELEVANT='package.json package-lock.json README.md .gitignore vitest.config.js test test/**'
+
+usage() {
+  echo "usage: packaging.sh {ignore|pathspec|version}" >&2
+  exit 2
+}
+
+# web-ext --ignore-files values, space-separated.
+#
+# Callers MUST disable pathname expansion first (`set -f`), or the shell will
+# glob `test/**` against the working tree before web-ext ever sees the pattern
+# and silently narrow it to whatever happens to exist right now.
+cmd_ignore() {
+  echo "$NOT_PACKAGED_TRACKED $NOT_PACKAGED_BUILD"
+}
+
+# git pathspec excluding the tracked files that cannot change the shipped
+# bytes, e.g. :(exclude)extension/package.json :(exclude)extension/test/**
+#
+# This answers "what moves the version?", NOT "what goes in the XPI?" — see
+# NOT_VERSION_RELEVANT for why those differ. cmd_ignore answers the other one.
+# Same `set -f` requirement as above.
+cmd_pathspec() {
+  for entry in $NOT_VERSION_RELEVANT; do
+    printf ':(exclude)extension/%s ' "$entry"
+  done
+  echo
+}
+
+# Strip leading zeros from one segment, leaving at least one digit.
+#
+# This exists for AMO and nothing else. Mozilla's version grammar for
+# addons.mozilla.org is documented as
+#
+#     ^(0|[1-9][0-9]{0,8})([.](0|[1-9][0-9]{0,8})){0,3}$
+#
+# — each segment is either the single digit `0` or starts 1-9, so `08` and
+# `0201` are rejected outright, while `0` itself is fine. MDN states it in
+# prose too: "Non-zero numbers must not include a leading zero."
+#
+# POSIX sh has no trim-loop, hence the while.
+unpad() {
+  s=$1
+  while [ "${#s}" -gt 1 ]; do
+    case "$s" in
+      0*) s=${s#0} ;;
+      *)  break ;;
+    esac
+  done
+  printf '%s' "$s"
+}
+
+# The extension's version: `YYYY.M.D.HHMM`, UTC, derived from the commit TIME
+# of the newest change to a PACKAGED extension file.
+#
+# THE ONE DELIBERATE DEPARTURE FROM THE FAMILY SHAPE, and it is a rendering
+# difference only. Rule 148 says `YYYY.MM.DD.HHMM` zero-padded, and every other
+# FC artifact emits exactly that. AMO's grammar (see unpad) forbids the padding,
+# and AMO is not negotiable: a rejected version is burned, since AMO 409s on
+# re-signing a version it has already seen. So the extension emits THE SAME
+# NUMBERS unpadded — 2026.08.29.0201 and 2026.8.29.201 are one value in two
+# renderings, and rule 148 already specifies comparison as numeric per segment,
+# under which they are equal. Nothing published is reordered by the choice, and
+# left-padding each segment recovers the family string exactly.
+#
+# HHMM is one segment, not two, because AMO allows at most FOUR. Unpadded that
+# reads oddly (00:14 -> `14`, midnight -> `0`) but stays strictly increasing
+# within a day, which is all the ordering needs.
+#
+# Why the commit's time and not the build's:
+#   * MONOTONIC — max() over a set that only ever gains members.
+#   * STABLE while the extension is unchanged, so an unchanged extension keeps
+#     its version, the ext-<version> signature cache still hits, and AMO is
+#     called once per extension CHANGE rather than once per push. Build-time
+#     minutes would re-sign on every push and never let two channels share a
+#     signature.
+#   * SHARED ACROSS CHANNELS — after a merge, `main` sees the same commit and
+#     derives the same number, so `:latest` reuses the signature `:dev` already
+#     produced for byte-identical code. Same code, same version, one signing.
+#   * REPRODUCIBLE — any checkout of a commit yields that commit's version.
+#
+# Never a commit count (family rule 149): a count is per-branch, so `dev` and
+# `main` count different histories of the same code and order by which branch
+# accumulated more commits rather than by which is newer. A squash-merge makes
+# that permanent. Roundtable's 2026-08-24 incident, in a different repo.
+#
+# Requires real history: a depth-1 clone sees one commit and will derive a wrong
+# (too low) value. Every consumer must check out with fetch-depth: 0.
+#
+# Formatted through git rather than date(1): busybox date does not reliably
+# accept `-d @<epoch>`, and git's --date=format-local is available wherever git
+# is. TZ=UTC so the value does not depend on the runner's timezone.
+cmd_version() {
+  root=$(git rev-parse --show-toplevel)
+  # Unquoted on purpose: the pathspec must word-split into separate args.
+  # Globbing is already off script-wide (set -euf above).
+  # shellcheck disable=SC2046
+  sha=$(cd "$root" && git log --format='%ct %H' HEAD -- extension/ $(cmd_pathspec) \
+        | sort -n | tail -1 | cut -d' ' -f2)
+  if [ -z "$sha" ]; then
+    echo "packaging.sh: no commit touches a packaged extension file" >&2
+    exit 1
+  fi
+  padded=$(cd "$root" && TZ=UTC git show -s --format=%cd \
+           --date='format-local:%Y.%m.%d.%H%M' "$sha")
+  # Rebinding the function's own positional params, which are unused here.
+  # shellcheck disable=SC2046
+  set -- $(echo "$padded" | tr '.' ' ')
+  echo "$(unpad "$1").$(unpad "$2").$(unpad "$3").$(unpad "$4")"
+}
+
+[ $# -ge 1 ] || usage
+case "$1" in
+  ignore)   cmd_ignore ;;
+  pathspec) cmd_pathspec ;;
+  version)  cmd_version ;;
+  *)        usage ;;
+esac

extension/test/helpers/loadLib.js

@@ -0,0 +1,29 @@
+import { readFileSync } from 'node:fs'
+import { fileURLToPath } from 'node:url'
+import path from 'node:path'
+
+const LIB_DIR = path.join(path.dirname(fileURLToPath(import.meta.url)), '..', '..', 'lib')
+
+/**
+ * Load an extension lib and hand back the globals it declares.
+ *
+ * The files under lib/ are CLASSIC scripts, not ES modules: manifest.json
+ * lists them in `background.scripts` and options.html pulls them in with a
+ * plain <script> tag, so they declare bare functions into a shared scope and
+ * export nothing. Rather than bolt a `module.exports` shim onto production
+ * code that would never run in the browser, evaluate the real file the same
+ * way the browser does — as a script body — and pick the declarations back out.
+ *
+ * This means the specs exercise the exact bytes that get packaged into the
+ * XPI. Only usable for libs that touch no browser APIs at load time
+ * (url.js, platforms.js); cookies.js and api.js reference `browser.*` and
+ * would need stubbing, which is why they aren't loaded this way.
+ *
+ * @param {string} filename e.g. 'url.js'
+ * @param {string[]} names  declarations to return, e.g. ['normalizeApiUrl']
+ */
+export function loadLib(filename, names) {
+  const source = readFileSync(path.join(LIB_DIR, filename), 'utf8')
+  const factory = new Function(`${source}\nreturn { ${names.join(', ')} }`)
+  return factory()
+}

extension/test/platforms.spec.js

@@ -0,0 +1,190 @@
+import { describe, it, expect } from 'vitest'
+import { readFileSync } from 'node:fs'
+import { fileURLToPath } from 'node:url'
+import path from 'node:path'
+import { loadLib } from './helpers/loadLib.js'
+
+const EXT_DIR = path.join(path.dirname(fileURLToPath(import.meta.url)), '..')
+const manifest = JSON.parse(readFileSync(path.join(EXT_DIR, 'manifest.json'), 'utf8'))
+
+const { getPlatformFromUrl, isArtistPage, PLATFORMS, PLATFORM_ARTIST_PATTERNS } = loadLib(
+  'platforms.js',
+  ['getPlatformFromUrl', 'isArtistPage', 'PLATFORMS', 'PLATFORM_ARTIST_PATTERNS']
+)
+
+describe('getPlatformFromUrl', () => {
+  it('identifies each platform from a domain URL', () => {
+    expect(getPlatformFromUrl('https://www.patreon.com/Atole')).toBe('patreon')
+    expect(getPlatformFromUrl('https://subscribestar.adult/someone')).toBe('subscribestar')
+    expect(getPlatformFromUrl('https://www.hentai-foundry.com/user/someone')).toBe('hentaifoundry')
+    expect(getPlatformFromUrl('https://discord.com/channels/@me')).toBe('discord')
+    expect(getPlatformFromUrl('https://www.pixiv.net/en/users/123')).toBe('pixiv')
+  })
+
+  it('accepts http as well as https, with or without www', () => {
+    expect(getPlatformFromUrl('http://patreon.com/Atole')).toBe('patreon')
+    expect(getPlatformFromUrl('https://www.patreon.com/Atole')).toBe('patreon')
+  })
+
+  it('returns null for unrelated hosts', () => {
+    expect(getPlatformFromUrl('https://example.com/patreon.com')).toBe(null)
+    expect(getPlatformFromUrl('https://not-patreon.com/Atole')).toBe(null)
+    expect(getPlatformFromUrl('')).toBe(null)
+  })
+
+  it('returns null for deviantart, retired at #3069', () => {
+    // The 2026-07-05 product decision (FC downloaders = art-dedicated services
+    // only) left deviantart wired for seven weeks. Asserting the negative is
+    // what keeps a partial retirement from being re-completed by accident.
+    expect(getPlatformFromUrl('https://www.deviantart.com/someone')).toBe(null)
+    expect(PLATFORMS.deviantart).toBeUndefined()
+    expect(PLATFORM_ARTIST_PATTERNS.deviantart).toBeUndefined()
+  })
+})
+
+describe('isArtistPage', () => {
+  // Regression cases from issue #1485: the Add-to-FC button vanished once the
+  // operator SUBSCRIBED to a creator, because Patreon serves subscribed users
+  // the /cw/ ("creator workspace") URL and the pattern only matched the bare
+  // root. All three creator URL shapes must match, plus inner pages — the
+  // button matters most exactly when you're subscribed.
+  it('matches all three Patreon creator URL shapes', () => {
+    expect(isArtistPage('https://www.patreon.com/Atole', 'patreon')).toBe(true)
+    expect(isArtistPage('https://www.patreon.com/c/Atole', 'patreon')).toBe(true)
+    expect(isArtistPage('https://www.patreon.com/cw/Atole', 'patreon')).toBe(true)
+  })
+
+  it('matches Patreon creator inner pages', () => {
+    expect(isArtistPage('https://www.patreon.com/cw/Atole/posts', 'patreon')).toBe(true)
+    expect(isArtistPage('https://www.patreon.com/Atole/membership', 'patreon')).toBe(true)
+  })
+
+  it('excludes Patreon navigation pages that are not creators', () => {
+    for (const nav of ['home', 'search', 'messages', 'notifications', 'library', 'settings']) {
+      expect(isArtistPage(`https://www.patreon.com/${nav}`, 'patreon')).toBe(false)
+      expect(isArtistPage(`https://www.patreon.com/${nav}/anything`, 'patreon')).toBe(false)
+    }
+  })
+
+  it('matches SubscribeStar creator roots on both TLDs but not feed pages', () => {
+    expect(isArtistPage('https://subscribestar.adult/someone', 'subscribestar')).toBe(true)
+    expect(isArtistPage('https://subscribestar.com/someone', 'subscribestar')).toBe(true)
+    expect(isArtistPage('https://subscribestar.adult/feed', 'subscribestar')).toBe(false)
+    expect(isArtistPage('https://subscribestar.adult/messages', 'subscribestar')).toBe(false)
+  })
+
+  it('matches Hentai Foundry user pages only', () => {
+    expect(isArtistPage('https://www.hentai-foundry.com/user/someone', 'hentaifoundry')).toBe(true)
+    expect(isArtistPage('https://www.hentai-foundry.com/pictures/popular', 'hentaifoundry')).toBe(
+      false
+    )
+  })
+
+  it('matches Pixiv numeric user pages, with or without the /en/ prefix', () => {
+    expect(isArtistPage('https://www.pixiv.net/users/12345', 'pixiv')).toBe(true)
+    expect(isArtistPage('https://www.pixiv.net/en/users/12345', 'pixiv')).toBe(true)
+    expect(isArtistPage('https://www.pixiv.net/en/artworks/999', 'pixiv')).toBe(false)
+  })
+
+  it('returns false for a platform with no artist pattern (discord)', () => {
+    expect(isArtistPage('https://discord.com/channels/@me', 'discord')).toBe(false)
+  })
+
+  it('returns false for an unknown platform key', () => {
+    expect(isArtistPage('https://www.patreon.com/Atole', 'nope')).toBe(false)
+  })
+})
+
+describe('platform table integrity', () => {
+  it('gives every artist pattern a corresponding platform entry', () => {
+    // A pattern keyed to a platform that no longer exists is dead code that
+    // silently never fires; the reverse (a platform with no pattern) is the
+    // legitimate discord case, so only this direction is an error.
+    for (const key of Object.keys(PLATFORM_ARTIST_PATTERNS)) {
+      expect(Object.keys(PLATFORMS)).toContain(key)
+    }
+  })
+
+  it('gives every platform the fields the popup renders', () => {
+    for (const [key, platform] of Object.entries(PLATFORMS)) {
+      expect(platform.name, `${key}.name`).toBeTruthy()
+      expect(platform.color, `${key}.color`).toMatch(/^#[0-9A-Fa-f]{6}$/)
+      expect(['cookies', 'token'], `${key}.authType`).toContain(platform.authType)
+      expect(platform.urlPattern, `${key}.urlPattern`).toBeInstanceOf(RegExp)
+      expect(Array.isArray(platform.domains), `${key}.domains`).toBe(true)
+      expect(platform.domains.length, `${key}.domains`).toBeGreaterThan(0)
+    }
+  })
+
+  it('keeps every artist URL matched by its own platform pattern too', () => {
+    // isArtistPage is only ever consulted after getPlatformFromUrl resolves a
+    // key, so an artist pattern matching a URL its platform's urlPattern
+    // rejects would be unreachable.
+    const samples = {
+      patreon: 'https://www.patreon.com/cw/Atole',
+      subscribestar: 'https://subscribestar.adult/someone',
+      hentaifoundry: 'https://www.hentai-foundry.com/user/someone',
+      pixiv: 'https://www.pixiv.net/en/users/12345'
+    }
+    for (const [key, url] of Object.entries(samples)) {
+      expect(isArtistPage(url, key), `${key} artist pattern`).toBe(true)
+      expect(getPlatformFromUrl(url), `${key} urlPattern`).toBe(key)
+    }
+  })
+})
+
+describe('manifest.json agrees with the platform table', () => {
+  // #3069: deviantart was dropped from the product in July but survived in
+  // manifest.json until late August, because NOTHING tied the manifest's
+  // domain lists back to PLATFORMS. These two specs are that tie. Both
+  // directions matter: a stale match ships host access the product decided
+  // not to use, and a missing one silently kills the Add-to-FC button.
+  const matches = manifest.content_scripts[0].matches
+  // '*://*.patreon.com/*' -> '.patreon.com', the form PLATFORMS.domains uses.
+  const hostOf = (m) => m.replace(/^\*:\/\/\*/, '').replace(/\/\*$/, '')
+
+  it('injects the content script only on domains a platform claims', () => {
+    for (const m of matches) {
+      const host = hostOf(m)
+      const owner = Object.entries(PLATFORMS).find(
+        ([, p]) => p.domains.includes(host)
+      )
+      expect(owner, `no platform claims content-script match "${m}"`).toBeTruthy()
+      // The content script exists to draw the Add-as-source button, so a
+      // platform with no artist pattern (discord) has no business here.
+      expect(
+        PLATFORM_ARTIST_PATTERNS[owner[0]],
+        `"${m}" injects for ${owner[0]}, which has no artist pattern`
+      ).toBeTruthy()
+    }
+  })
+
+  it('injects on every platform that has an artist pattern', () => {
+    const covered = new Set(
+      matches
+        .map(hostOf)
+        .map((h) => Object.entries(PLATFORMS).find(([, p]) => p.domains.includes(h)))
+        .filter(Boolean)
+        .map(([key]) => key)
+    )
+    for (const key of Object.keys(PLATFORM_ARTIST_PATTERNS)) {
+      expect(covered, `${key} has an artist pattern but no content-script match`).toContain(key)
+    }
+  })
+
+  it('requests no host permission for a domain no platform claims', () => {
+    // '*://*/*' is the deliberate exception: FC is self-hosted at an arbitrary
+    // operator-chosen URL, so the extension cannot enumerate its own backend.
+    // Every OTHER entry is a platform domain and must still have an owner.
+    for (const h of manifest.host_permissions) {
+      if (h === '*://*/*') continue
+      const host = hostOf(h)
+      // pixiv's OAuth/API hosts are pixiv infrastructure, not creator pages,
+      // so they are matched by suffix rather than by the domains list.
+      const claimed = Object.values(PLATFORMS).some(
+        (p) => p.domains.includes(host) || p.domains.some((d) => host.endsWith(d))
+      )
+      expect(claimed, `host permission "${h}" belongs to no platform`).toBe(true)
+    }
+  })
+})

extension/test/url.spec.js

@@ -0,0 +1,93 @@
+import { describe, it, expect } from 'vitest'
+import { loadLib } from './helpers/loadLib.js'
+
+const { normalizeApiUrl, webRootFromApiUrl } = loadLib('url.js', [
+  'normalizeApiUrl',
+  'webRootFromApiUrl'
+])
+
+describe('normalizeApiUrl', () => {
+  // The bug this exists for (issue #2393): the instance root was accepted and
+  // stored verbatim, so every request went to /credentials instead of
+  // /api/credentials. That path is a Vue router route, so the SPA catch-all
+  // answered GET with 200 HTML and rejected POST with 405 — which read as a
+  // backend bug rather than a URL one.
+  it('appends /api to an instance root', () => {
+    expect(normalizeApiUrl('http://curator.traefik.internal')).toBe(
+      'http://curator.traefik.internal/api'
+    )
+  })
+
+  it('leaves an API root alone rather than doubling the suffix', () => {
+    expect(normalizeApiUrl('http://curator.traefik.internal/api')).toBe(
+      'http://curator.traefik.internal/api'
+    )
+  })
+
+  it('is idempotent', () => {
+    const once = normalizeApiUrl('http://curator.example.com')
+    expect(normalizeApiUrl(once)).toBe(once)
+  })
+
+  it('strips trailing slashes before deciding', () => {
+    expect(normalizeApiUrl('http://curator.example.com/')).toBe('http://curator.example.com/api')
+    expect(normalizeApiUrl('http://curator.example.com///')).toBe('http://curator.example.com/api')
+    expect(normalizeApiUrl('http://curator.example.com/api/')).toBe('http://curator.example.com/api')
+  })
+
+  it('trims surrounding whitespace (paste artifacts)', () => {
+    expect(normalizeApiUrl('  http://curator.example.com  ')).toBe(
+      'http://curator.example.com/api'
+    )
+  })
+
+  it('matches the /api suffix case-insensitively', () => {
+    expect(normalizeApiUrl('http://curator.example.com/API')).toBe('http://curator.example.com/API')
+  })
+
+  it('returns empty string for empty/nullish input, never a bare "/api"', () => {
+    // isConfigured() gates on truthiness, so a bogus '/api' here would read as
+    // "configured" and produce a request against the options page's own origin.
+    expect(normalizeApiUrl('')).toBe('')
+    expect(normalizeApiUrl('   ')).toBe('')
+    expect(normalizeApiUrl(null)).toBe('')
+    expect(normalizeApiUrl(undefined)).toBe('')
+  })
+
+  it('does not treat a path merely containing "api" as the suffix', () => {
+    expect(normalizeApiUrl('http://curator.example.com/apiary')).toBe(
+      'http://curator.example.com/apiary/api'
+    )
+  })
+
+  it('preserves a subpath deployment', () => {
+    expect(normalizeApiUrl('http://host.internal/curator')).toBe('http://host.internal/curator/api')
+  })
+})
+
+describe('webRootFromApiUrl', () => {
+  // The SPA root, where the Vue router and the served XPI live. Used by
+  // OPEN_ARTIST_PAGE and the self-update check — NOT the JSON API.
+  it('strips the /api suffix', () => {
+    expect(webRootFromApiUrl('http://curator.example.com/api')).toBe('http://curator.example.com')
+  })
+
+  it('accepts an instance root unchanged', () => {
+    expect(webRootFromApiUrl('http://curator.example.com')).toBe('http://curator.example.com')
+  })
+
+  it('agrees with normalizeApiUrl in both directions', () => {
+    for (const input of ['http://curator.example.com', 'http://curator.example.com/api']) {
+      expect(normalizeApiUrl(webRootFromApiUrl(input))).toBe(normalizeApiUrl(input))
+    }
+  })
+
+  it('preserves a subpath deployment', () => {
+    expect(webRootFromApiUrl('http://host.internal/curator/api')).toBe('http://host.internal/curator')
+  })
+
+  it('returns empty string for empty/nullish input', () => {
+    expect(webRootFromApiUrl('')).toBe('')
+    expect(webRootFromApiUrl(null)).toBe('')
+  })
+})

extension/test/version.spec.js

@@ -0,0 +1,191 @@
+import { describe, it, expect } from 'vitest'
+import { readdirSync, readFileSync } from 'node:fs'
+import { execFileSync } from 'node:child_process'
+import { fileURLToPath } from 'node:url'
+import path from 'node:path'
+
+const EXT_DIR = path.join(path.dirname(fileURLToPath(import.meta.url)), '..')
+const read = (name) => JSON.parse(readFileSync(path.join(EXT_DIR, name), 'utf8'))
+const readText = (...seg) => readFileSync(path.join(EXT_DIR, ...seg), 'utf8')
+
+// Only the git-free subcommands are exercised here: `version` shells out to
+// git, and the extension lane runs on node:24-bookworm-slim which may not ship
+// it. That one is covered where git is guaranteed — ci.yml's extension-version
+// lane and build.yml both run on ci-python.
+const packaging = (cmd) =>
+  execFileSync('sh', [path.join(EXT_DIR, 'scripts', 'packaging.sh'), cmd], {
+    cwd: EXT_DIR,
+    encoding: 'utf8'
+  })
+    .trim()
+    .split(/\s+/)
+    .filter(Boolean)
+
+describe('packaging.sh — the single definition of what ships', () => {
+  it('emits an ignore list and a pathspec that agree on the tracked files', () => {
+    const ignore = packaging('ignore')
+    const pathspec = packaging('pathspec').map((e) => e.replace(':(exclude)extension/', ''))
+
+    // Every git-excluded path must also be hidden from web-ext. The reverse is
+    // not required: node_modules and web-ext-artifacts are build output git
+    // never tracks, so they appear only in the ignore list.
+    for (const entry of pathspec) {
+      expect(ignore, `pathspec has "${entry}" but --ignore-files does not`).toContain(entry)
+    }
+    expect(pathspec.length).toBeGreaterThan(0)
+    expect(ignore).toContain('node_modules')
+  })
+
+  it('emits glob patterns literally, never expanded against the working tree', () => {
+    // The script iterates its lists with deliberate word-splitting, so it must
+    // run with pathname expansion off. Without that, invoking it from a cwd
+    // where test/ exists (exactly how ci.yml and vitest call it) expands
+    // `test/**` into the individual spec files, and the pathspec silently stops
+    // covering anything added later.
+    const pathspec = packaging('pathspec')
+    expect(pathspec).toContain(':(exclude)extension/test/**')
+    expect(pathspec.some((e) => e.includes('.spec.js'))).toBe(false)
+    expect(pathspec.some((e) => e.includes('helpers'))).toBe(false)
+
+    const ignore = packaging('ignore')
+    expect(ignore).toContain('test/**')
+    expect(ignore).toContain('scripts/**')
+    expect(ignore.some((e) => e.includes('.spec.js'))).toBe(false)
+  })
+
+  it('lets packaging.sh move the version, though it never ships in the XPI', () => {
+    // The two lists answer different questions and this is the one place they
+    // disagree. scripts/ is ignored by web-ext — it is repo tooling, not addon
+    // code — but packaging.sh DECIDES the version string, and build.yml stamps
+    // that string into the manifest.json that does ship. So changing how the
+    // version is computed changes the shipped bytes.
+    //
+    // Excluding it from the pathspec was invisible while every push rebuilt the
+    // web image. Milestone 313 step 4 made that rebuild conditional on the
+    // derived revision moving, and the omission turned into a silent failure:
+    // a new version means sign-extension misses its ext-<version> cache and
+    // signs, while build-web sees an unmoved revision, reuses the published
+    // image and ships the OLD XPI. An orphaned signature, and an instance
+    // serving code the registry calls current.
+    const pathspec = packaging('pathspec')
+    expect(
+      pathspec.some((e) => e.startsWith(':(exclude)extension/scripts')),
+      'the pathspec excludes scripts/, so a change to how the version is '
+        + 'derived would not move the version it derives',
+    ).toBe(false)
+
+    // ...and it is still kept out of the package itself. Both must hold: the
+    // tempting "fix" for either half is to make the two lists one again.
+    expect(packaging('ignore')).toContain('scripts')
+  })
+
+  it('keeps its own scripts and specs out of the XPI', () => {
+    // Both are repo infrastructure. web-ext packages everything not ignored, so
+    // omitting either would ship dev tooling to users -- and `test/**` in
+    // particular only survives because callers `set -f` before substituting it.
+    const ignore = packaging('ignore')
+    expect(ignore).toContain('vitest.config.js')
+    // Both forms per directory. The glob covers the contents; the bare name
+    // covers the directory ENTRY, which web-ext writes separately — with only
+    // the glob, the XPI carries an empty `test/` and `scripts/`.
+    for (const dir of ['test', 'scripts']) {
+      expect(ignore, `${dir} contents`).toContain(`${dir}/**`)
+      expect(ignore, `${dir} directory entry`).toContain(dir)
+    }
+  })
+})
+
+describe('consumers delegate rather than keeping their own copy', () => {
+  // These assertions are the actual anti-regression value: it is easy for a
+  // future edit to "simplify" by inlining a literal list again, which silently
+  // reintroduces the drift that issue #2397 was about.
+  it('package.json derives --ignore-files from the script', () => {
+    for (const [name, script] of Object.entries(read('package.json').scripts)) {
+      if (!script.includes('--ignore-files')) continue
+      expect(script, `${name} should call packaging.sh`).toContain('scripts/packaging.sh ignore')
+      expect(script, `${name} must set -f before the substitution`).toMatch(/set -f;/)
+    }
+  })
+
+  // Read from disk rather than listed by hand. The point of this assertion is
+  // that it survives consumers coming and going, and a hardcoded list is the
+  // one part of it that cannot — release.yml (milestone 318 step 7) would have
+  // joined the directory without joining the check.
+  const WORKFLOWS = readdirSync(path.join(EXT_DIR, '..', '.forgejo', 'workflows')).filter((f) =>
+    f.endsWith('.yml')
+  )
+
+  it('no workflow hardcodes the packaged-file set', () => {
+    // ci.yml used to substitute `packaging.sh pathspec` directly, for the
+    // manual-bump guard that milestone 271 step 5 retired. Nothing inlines the
+    // set today, and nothing should start to: a literal :(exclude)extension/...
+    // in a workflow means someone bypassed the shared definition, which is
+    // exactly the drift #2397 was about.
+    expect(
+      WORKFLOWS.length,
+      'no workflows found — the glob is not looking where it thinks'
+    ).toBeGreaterThan(2)
+    for (const wf of WORKFLOWS) {
+      const text = readText('..', '.forgejo', 'workflows', wf)
+      expect(text, `${wf} inlines an :(exclude) literal`).not.toMatch(/:\(exclude\)extension\//)
+    }
+  })
+
+  it('build.yml takes the shipped version from the script, not from the repo', () => {
+    // The version is DERIVED from commit time (#3092, milestone 271 step 4).
+    // Going back to reading the committed value is not a style regression, it
+    // is the bug: a hand-set version makes dev and main sign the same number
+    // for different code, and the ext-<version> cache then serves one channel
+    // the other's XPI.
+    const build = readText('..', '.forgejo', 'workflows', 'build.yml')
+    expect(build).toContain('packaging.sh version')
+    expect(build, 'build.yml re-reads the committed version instead of deriving it')
+      .not.toMatch(/grep[^\n]*'"version"'[^\n]*package\.json/)
+  })
+})
+
+describe('extension version', () => {
+  // Mozilla's published grammar for addons.mozilla.org, transcribed from MDN's
+  // manifest.json/version page. Each segment is the single digit 0 or starts
+  // 1-9 — so no leading zeros — and there are at most four of them.
+  const AMO = /^(0|[1-9][0-9]{0,8})(\.(0|[1-9][0-9]{0,8})){0,3}$/
+
+  it('keeps a committed version AMO would accept, though it ships nothing', () => {
+    // The committed value is wholly inert since milestone 318 step 8: there is
+    // no hand-set MAJOR.MINOR left for packaging.sh to read, and build.yml
+    // stamps the derived string over both files before web-ext sees them.
+    //
+    // It is still asserted, for one reason: `npm run build` locally packages
+    // whatever is committed, so a value AMO would reject turns a local build
+    // into a confusing failure with no CI signal ahead of it. ci.yml checks
+    // the same grammar against the DERIVED value, which is the one AMO sees.
+    for (const file of ['manifest.json', 'package.json']) {
+      expect(read(file).version, `${file} version is not AMO-shaped`).toMatch(AMO)
+    }
+  })
+
+  it('rejects the zero-padded family shape, which is why the extension unpads', () => {
+    // Guards the reason for the exception, not just its result. If this ever
+    // starts passing, someone has loosened the pattern and the next sign burns
+    // an AMO version to find out. (#3138.)
+    expect('2026.08.29.0201').not.toMatch(AMO)
+    expect('2026.8.29.201').toMatch(AMO)
+    // Five segments: AMO allows four.
+    expect('2026.8.29.2.1').not.toMatch(AMO)
+  })
+
+  it('declares manifest v3', () => {
+    expect(read('manifest.json').manifest_version).toBe(3)
+  })
+
+  it('lists every background script that exists, in dependency order', () => {
+    // url.js must load BEFORE api.js: api.js calls normalizeApiUrl at
+    // init()-time, and these are classic scripts sharing one scope, so a
+    // reordering here is a runtime ReferenceError with no build-time signal.
+    const scripts = read('manifest.json').background.scripts
+    for (const rel of scripts) {
+      expect(() => readFileSync(path.join(EXT_DIR, rel)), `missing ${rel}`).not.toThrow()
+    }
+    expect(scripts.indexOf('lib/url.js')).toBeLessThan(scripts.indexOf('lib/api.js'))
+  })
+})

extension/vitest.config.js

@@ -0,0 +1,13 @@
+import { defineConfig } from 'vitest/config'
+
+// Mirrors frontend/vitest.config.js, minus the Vue plugin — the extension has
+// no SFCs and mounts nothing. Pure-logic specs only, so `node` is enough; the
+// libs under test are deliberately the ones with no browser-API surface (see
+// test/helpers/loadLib.js).
+export default defineConfig({
+  test: {
+    environment: 'node',
+    include: ['test/**/*.spec.js'],
+    passWithNoTests: true
+  }
+})

frontend/src/components/artist/ArtistDangerZone.vue

@@ -50,14 +50,19 @@ const projected = ref(null)
 
 const projectedCounts = computed(() => projected.value?.projected || null)
 
-const modalDescription = computed(
-  () => projected.value
-    ? `Artist “${props.artistName}” — `
-        + `${projected.value.projected.images} images, `
-        + `${projected.value.projected.sources} sources, `
-        + `${Math.round(projected.value.projected.bytes_on_disk / 1_048_576)} MiB on disk`
-    : '',
-)
+// `posts` is named here, not left to the counts grid below it: an artist whose
+// posts are body-only previews as `images: 0`, and a summary line that says
+// only "0 images" reads as "this artist is empty" while the apply destroys
+// every captured post body (#3067). Attachments stay in the grid — the grid
+// renders every key, so this line carries only what changes the read.
+const modalDescription = computed(() => {
+  const p = projectedCounts.value
+  return p
+    ? `Artist “${props.artistName}” — ${p.images} images, `
+      + `${p.posts} posts, ${p.sources} sources, `
+      + `${Math.round(p.bytes_on_disk / 1_048_576)} MiB on disk`
+    : ''
+})
 
 async function onClick() {
   loading.value = true

frontend/src/components/common/SettingNumberField.vue

@@ -0,0 +1,56 @@
+<!--
+  Canonical settings number field (DRY pass #161): a compact numeric v-text-field
+  with a built-in clamp to [min,max] on commit. Hand-rolled identically across the
+  ML settings cards (HeadsCard x6, CropProposersCard, VideoEmbeddingCard).
+
+  The clamp is the point: the cards previously sent Number(raw) straight to the
+  API, so an out-of-range value bounced off the API's 400 validator (only
+  TranslationCard clamped). This is now the single home for that clamp.
+
+  Binds `modelValue` (v-model) and emits `change` on blur/enter AFTER clamping, so
+  the parent's save reads the already-clamped value — same as the prior
+  `v-model.number` + `@change=save` pattern.
+-->
+<template>
+  <v-text-field
+    :model-value="modelValue"
+    :label="label"
+    type="number"
+    :min="min"
+    :max="max"
+    :step="step"
+    :disabled="disabled"
+    :density="density" hide-details
+    :style="{ maxWidth }"
+    @update:model-value="v => emit('update:modelValue', v)"
+    @change="onCommit"
+  />
+</template>
+
+<script setup>
+const props = defineProps({
+  modelValue: { type: [Number, String], default: null },
+  label: { type: String, default: '' },
+  min: { type: [Number, String], default: null },
+  max: { type: [Number, String], default: null },
+  step: { type: [Number, String], default: 1 },
+  maxWidth: { type: String, default: '200px' },
+  density: { type: String, default: 'compact' },
+  disabled: { type: Boolean, default: false },
+})
+const emit = defineEmits(['update:modelValue', 'change'])
+
+function onCommit() {
+  // On blur/enter: coerce to a number and clamp to [min,max] so an out-of-range
+  // value never reaches the API. props.modelValue reflects the latest keystroke
+  // (kept in sync by the passthrough above); re-emit the clamped number, then let
+  // the parent persist.
+  let n = Number(props.modelValue)
+  if (!Number.isNaN(n)) {
+    if (props.min !== null && props.min !== '') n = Math.max(Number(props.min), n)
+    if (props.max !== null && props.max !== '') n = Math.min(Number(props.max), n)
+    if (n !== Number(props.modelValue)) emit('update:modelValue', n)
+  }
+  emit('change')
+}
+</script>

frontend/src/components/common/SettingToggleRow.vue

@@ -0,0 +1,42 @@
+<!--
+  Canonical settings toggle row (DRY pass #161): an accent icon + an uppercase
+  .fc-section-h label + a right-aligned switch. Hand-rolled identically in the
+  ML settings cards (HeadsCard x3, CropProposersCard, MLBackfillCard).
+
+  Two-way binds `modelValue` (so the parent switch state stays optimistic) AND
+  emits `change` with the new boolean, so the parent can persist + revert on
+  failure — matching the prior `v-model` + `@update:model-value=handler` pattern.
+-->
+<template>
+  <div class="d-flex align-center mb-1" style="gap: 10px;">
+    <v-icon v-if="icon" size="18" :color="iconColor">{{ icon }}</v-icon>
+    <span class="fc-section-h">{{ label }}</span>
+    <v-switch
+      :model-value="modelValue"
+      :loading="loading"
+      :disabled="disabled"
+      hide-details density="compact" color="success" class="ml-auto"
+      @update:model-value="onSwitch"
+    />
+  </div>
+</template>
+
+<script setup>
+defineProps({
+  modelValue: { type: Boolean, default: false },
+  label: { type: String, default: '' },
+  icon: { type: String, default: '' },
+  // Icon tint. Default accent; pass null for the theme default (e.g. when a row
+  // is off). null (not undefined) so the default doesn't override it.
+  iconColor: { type: String, default: 'accent' },
+  loading: { type: Boolean, default: false },
+  disabled: { type: Boolean, default: false },
+})
+const emit = defineEmits(['update:modelValue', 'change'])
+
+function onSwitch(v) {
+  const b = !!v
+  emit('update:modelValue', b)
+  emit('change', b)
+}
+</script>

frontend/src/components/gallery/GalleryItem.vue

@@ -139,9 +139,9 @@ function onThumbError() { thumbError.value = true }
   position: absolute; top: 8px; left: 8px;
   width: 22px; height: 22px; border-radius: 4px;
   border: 2px solid rgba(232, 228, 216, 0.8);
-  background: rgba(20, 23, 26, 0.45);
+  background: rgba(var(--v-theme-background), 0.45);
   display: grid; place-items: center;
-  color: #14171A; z-index: 11;
+  color: rgb(var(--v-theme-background)); z-index: 11;
 }
 .fc-gallery-item__checkbox.on {
   background: rgb(var(--v-theme-accent));
@@ -152,7 +152,7 @@ function onThumbError() { thumbError.value = true }
   min-width: 22px; height: 22px; padding: 0 5px;
   border-radius: 11px;
   background: rgb(var(--v-theme-accent));
-  color: #14171A; font-size: 12px; font-weight: 700;
+  color: rgb(var(--v-theme-background)); font-size: 12px; font-weight: 700;
   display: grid; place-items: center; z-index: 11;
   pointer-events: none;
 }
@@ -160,7 +160,8 @@ function onThumbError() { thumbError.value = true }
   position: absolute; left: 0; right: 0; bottom: 0;
   padding: 14px 8px 6px;
   background: linear-gradient(
-    to top, rgba(20, 23, 26, 0.78), rgba(20, 23, 26, 0)
+    to top, rgba(var(--v-theme-background), 0.78),
+    rgba(var(--v-theme-background), 0)
   );
   font-size: 12px; line-height: 1.2;
   white-space: nowrap; overflow: hidden; text-overflow: ellipsis;

frontend/src/components/settings/AttachmentReclaimCard.vue

@@ -0,0 +1,127 @@
+<template>
+  <!-- #3068: attachment reclamation. PostAttachment's FKs are both SET NULL, so
+       a deleted post or artist leaves the row behind; and the store is
+       sha-addressed, so one blob backs many rows and deleting a row never freed
+       its file. Nothing swept either. Preview first, then apply (destructive:
+       unlinks files). -->
+  <MaintenanceTile
+    icon="mdi-paperclip-off"
+    title="Reclaim orphaned attachments"
+    blurb="Remove attachment records belonging to nothing, and the files nothing references."
+    destructive
+    :open="applying || previewing"
+  >
+    <p class="text-body-2 mb-3">
+      Attachment records survive the post and artist they belonged to, and the
+      files behind them are shared between records — so a deleted record never
+      freed its file on its own. This finds records attributed to
+      <strong>neither</strong> a post nor an artist, and files in the attachment
+      store that <strong>no remaining record</strong> points at.
+      <strong>Preview</strong> first; <strong>Apply</strong> deletes those
+      records and unlinks those files. Files written in the last few hours are
+      always left alone, so an in-progress download is never caught mid-write.
+    </p>
+
+    <div class="d-flex align-center flex-wrap" style="gap: 12px;">
+      <v-btn
+        color="primary" variant="tonal" rounded="pill"
+        :loading="previewing" :disabled="applying" @click="preview"
+      >
+        <v-icon start>mdi-magnify</v-icon> Preview
+      </v-btn>
+      <v-btn
+        color="error" rounded="pill"
+        :loading="applying"
+        :disabled="previewing || !canApply"
+        @click="confirmOpen = true"
+      >
+        <v-icon start>mdi-paperclip-off</v-icon> Apply
+      </v-btn>
+    </div>
+
+    <v-alert
+      v-if="summary" :type="summaryType" variant="tonal" class="mt-4"
+      density="comfortable"
+    >
+      <span v-if="applied">
+        Deleted {{ summary.rows }} orphaned record(s) and unlinked
+        {{ summary.files }} file(s), reclaiming {{ humanBytes(summary.bytes) }}.
+      </span>
+      <span v-else-if="hasWork">
+        {{ summary.rows }} orphaned record(s) and {{ summary.files }}
+        unreferenced file(s) — {{ humanBytes(summary.bytes) }} reclaimable.
+        Click <strong>Apply</strong> to remove them.
+      </span>
+      <span v-else>Nothing to reclaim — every attachment is accounted for.</span>
+
+      <!-- Both of these change what the numbers MEAN, so they are stated
+           whenever they are non-zero rather than hidden in a tooltip. -->
+      <div v-if="summary.files_failed" class="mt-1 text-caption">
+        {{ summary.files_failed }} file(s) could not be read or removed — see
+        the worker log.
+      </div>
+      <div v-if="summary.partial" class="mt-1 text-caption">
+        Stopped early at the time limit; some of the store was not examined.
+        Run it again to continue.
+      </div>
+    </v-alert>
+
+    <QueueStatusBar queue="maintenance_long" queue-label="Maintenance" />
+
+    <v-dialog v-model="confirmOpen" max-width="440">
+      <v-card>
+        <v-card-title>Reclaim orphaned attachments?</v-card-title>
+        <v-card-text class="text-body-2">
+          This permanently deletes
+          <strong>{{ summary?.rows ?? 0 }}</strong> attachment record(s) and
+          unlinks <strong>{{ summary?.files ?? 0 }}</strong> file(s)
+          ({{ humanBytes(summary?.bytes) }}). Only files that no remaining
+          record points at are removed, so nothing still attached to a post
+          is affected.
+        </v-card-text>
+        <v-card-actions>
+          <v-spacer />
+          <v-btn variant="text" @click="confirmOpen = false">Cancel</v-btn>
+          <v-btn color="error" @click="apply">Reclaim</v-btn>
+        </v-card-actions>
+      </v-card>
+    </v-dialog>
+  </MaintenanceTile>
+</template>
+
+<script setup>
+import { computed, ref } from 'vue'
+
+import { useMaintenanceTask } from '../../composables/useMaintenanceTask.js'
+import { humanBytes } from '../../utils/bytes.js'
+import MaintenanceTile from '../common/MaintenanceTile.vue'
+import QueueStatusBar from './QueueStatusBar.vue'
+
+const confirmOpen = ref(false)
+
+// Walks the whole attachment store, so it can run for minutes on a large
+// library — the service caps itself at 900s and reports `partial`. 150 polls
+// × 2s ≈ 5m of foreground waiting; past that the composable hands off to the
+// task dashboard rather than spinning forever.
+const { previewing, applying, summary, applied, preview, apply: applyTask } = useMaintenanceTask({
+  endpoint: '/api/admin/maintenance/reclaim-attachments',
+  storageKey: 'fc.maint.reclaimAttachments',
+  appliedToast: 'Orphaned attachments reclaimed',
+  maxPolls: 150,
+})
+
+const hasWork = computed(
+  () => !!summary.value && (summary.value.rows > 0 || summary.value.files > 0),
+)
+const canApply = computed(() => hasWork.value && !applied.value)
+const summaryType = computed(() => {
+  if (applied.value) return 'success'
+  return hasWork.value ? 'info' : 'success'
+})
+
+// The confirm dialog gates the destructive apply; close it, then run.
+function apply () {
+  confirmOpen.value = false
+  applyTask()
+}
+</script>

frontend/src/components/settings/BrowserExtensionCard.vue

@@ -4,12 +4,22 @@
       <span v-if="manifest?.installed" class="text-caption fc-muted">
         · Firefox · v{{ manifest.version }}
       </span>
+      <!-- Which channel this instance serves, so it is visible without
+           installing anything. Rendered only when the image declares one: a
+           locally-built image, or one predating the field, says nothing rather
+           than guessing. Never merged into the version string beside it — see
+           the endpoint's note on why a `-dev` suffix breaks the comparator. -->
+      <v-chip
+        v-if="manifest?.channel"
+        size="x-small" variant="tonal" class="ml-2"
+        :color="manifest.channel === 'dev' ? 'warning' : 'info'"
+      >{{ manifest.channel }}</v-chip>
     </CardHeading>
 
     <v-card-text>
       <p class="fc-muted text-body-2">
         Pushes session cookies from supported platforms
-        (patreon, subscribestar, hentaifoundry, discord, pixiv, deviantart)
+        (patreon, subscribestar, hentaifoundry, discord, pixiv)
         into FabledCurator, and lets you add a creator as a source from
         their page in one click.
       </p>

frontend/src/components/settings/CropProposersCard.vue

@@ -15,28 +15,23 @@
     </p>
 
     <div v-for="p in proposers" :key="p.key" class="fc-proposer">
-      <div class="d-flex align-center mb-1" style="gap: 10px;">
-        <v-icon size="18" :color="p.on ? 'accent' : undefined">{{ p.icon }}</v-icon>
-        <span class="fc-section-h">{{ p.label }}</span>
-        <v-switch
-          v-model="p.on" :loading="busy" hide-details density="compact"
-          color="success" class="ml-auto"
-          @update:model-value="v => saveToggle(p, v)"
-        />
-      </div>
+      <SettingToggleRow
+        v-model="p.on" :loading="busy" :icon="p.icon"
+        :icon-color="p.on ? 'accent' : null" :label="p.label"
+        @change="v => saveToggle(p, v)"
+      />
       <p class="fc-muted text-body-2 mb-2">{{ p.help }}</p>
       <div class="d-flex flex-wrap mb-4" style="gap: 12px;">
         <v-text-field
           v-model="p.weights" label="Weights" density="compact" hide-details
           style="min-width: 300px; flex: 1;" :disabled="busy || !p.on"
           placeholder="name | URL | hf_repo::file"
-          @change="save({ [`detector_${p.key}_weights`]: p.weights })"
+          @change="saveField({ [`detector_${p.key}_weights`]: p.weights })"
         />
-        <v-text-field
-          v-model.number="p.conf" label="Confidence" type="number"
-          min="0" max="1" step="0.05" density="compact" hide-details
-          style="max-width: 140px;" :disabled="busy || !p.on"
-          @change="save({ [`detector_${p.key}_conf`]: Number(p.conf) })"
+        <SettingNumberField
+          v-model="p.conf" label="Confidence" :min="0" :max="1" :step="0.05"
+          max-width="140px" :disabled="busy || !p.on"
+          @change="saveField({ [`detector_${p.key}_conf`]: Number(p.conf) })"
         />
       </div>
     </div>
@@ -48,12 +43,12 @@
         storage. Dedupe IoU drops near-duplicate crops before embedding.
       </p>
       <div class="d-flex flex-wrap" style="gap: 12px;">
-        <v-text-field
+        <SettingNumberField
           v-for="c in caps" :key="c.key"
-          v-model.number="c.val" :label="c.label" type="number"
-          :min="c.min" :max="c.max" :step="c.step || 1" density="compact"
-          hide-details style="max-width: 165px;" :disabled="busy"
-          @change="save({ [c.key]: Number(c.val) })"
+          v-model="c.val" :label="c.label"
+          :min="c.min" :max="c.max" :step="c.step || 1"
+          max-width="165px" :disabled="busy"
+          @change="saveField({ [c.key]: Number(c.val) })"
         />
       </div>
     </div>
@@ -61,14 +56,16 @@
 </template>
 
 <script setup>
-import { toast } from '../../utils/toast.js'
 import { onMounted, ref } from 'vue'
 
 import MaintenanceTile from '../common/MaintenanceTile.vue'
+import SettingNumberField from '../common/SettingNumberField.vue'
+import SettingToggleRow from '../common/SettingToggleRow.vue'
+import { useSettingSave } from '../../composables/useSettingSave.js'
 import { useMLStore } from '../../stores/ml.js'
 
 const mlSettings = useMLStore()
-const busy = ref(false)
+const { busy, save } = useSettingSave(mlSettings.patchSettings)
 const proposers = ref([])
 const caps = ref([])
 
@@ -111,31 +108,20 @@ onMounted(async () => {
   caps.value = CAP_DEFS.map(c => ({ ...c, val: s[c.key] ?? 0 }))
 })
 
-async function save(patch, revert) {
-  busy.value = true
-  try {
-    await mlSettings.patchSettings(patch)
-    toast({ text: 'Saved', type: 'success' })
-  } catch (e) {
-    if (revert) revert()
-    toast({ text: `Could not save: ${e.message}`, type: 'error' })
-  } finally {
-    busy.value = false
-  }
+// Field @change → persist with a "Saved" confirmation. SettingNumberField has
+// already clamped numeric values to their [min,max] before this fires.
+function saveField(patch) {
+  save(patch, { successMessage: 'Saved' })
 }
 
-function saveToggle (p, v) {
+async function saveToggle(p, v) {
   // Revert the switch on failure so it never lies about the persisted state.
-  save({ [`detector_${p.key}_enabled`]: !!v }, () => { p.on = !v })
+  const ok = await save({ [`detector_${p.key}_enabled`]: !!v }, { successMessage: 'Saved' })
+  if (!ok) p.on = !v
 }
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
-.fc-section-h {
-  font-size: 13px; font-weight: 700; letter-spacing: 0.03em;
-  text-transform: uppercase; color: rgb(var(--v-theme-on-surface));
-}
 .fc-proposer {
   border-top: 1px solid rgb(var(--v-theme-surface-light)); padding-top: 14px;
 }

frontend/src/components/settings/DangerZoneCard.vue

@@ -112,7 +112,6 @@ async function onCommit() {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 .fc-code {
   background: rgb(var(--v-theme-surface-light));
   border-radius: 4px; padding: 2px 8px;

frontend/src/components/settings/DbMaintenanceCard.vue

@@ -42,7 +42,7 @@
           </tr>
         </tbody>
       </v-table>
-      <p v-else class="text-caption mt-3" style="opacity: 0.6;">
+      <p v-else class="text-caption mt-3 fc-muted">
         No table statistics yet.
       </p>
   </MaintenanceTile>

frontend/src/components/settings/DownloadsActivityPanel.vue

@@ -35,7 +35,7 @@
         All subscription sources healthy.
       </p>
       <p v-else class="text-body-2 mb-0">
-        <b class="fc-bad">{{ failing.length }}</b> failing source(s):
+        <b class="fc-weak">{{ failing.length }}</b> failing source(s):
         <span class="fc-muted">{{ failingNames }}</span>
       </p>
     </v-card-text>
@@ -72,6 +72,4 @@ onUnmounted(() => { if (pollId) clearInterval(pollId) })
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
-.fc-bad { color: rgb(var(--v-theme-error)); }
 </style>

frontend/src/components/settings/GatedPurgeCard.vue

@@ -102,6 +102,7 @@
 import { computed, ref } from 'vue'
 
 import { useMaintenanceTask } from '../../composables/useMaintenanceTask.js'
+import { humanBytes } from '../../utils/bytes.js'
 import MaintenanceTile from '../common/MaintenanceTile.vue'
 import QueueStatusBar from './QueueStatusBar.vue'
 
@@ -122,14 +123,6 @@ const summaryType = computed(() => {
   return summary.value && summary.value.matched > 0 ? 'info' : 'success'
 })
 
-function humanBytes (n) {
-  const b = Number(n || 0)
-  if (b >= 1 << 30) return (b / (1 << 30)).toFixed(1) + ' GB'
-  if (b >= 1 << 20) return (b / (1 << 20)).toFixed(1) + ' MB'
-  if (b >= 1 << 10) return (b / (1 << 10)).toFixed(1) + ' KB'
-  return b + ' B'
-}
-
 // The confirm dialog gates the destructive apply; close it, then run.
 function apply () {
   confirmOpen.value = false

frontend/src/components/settings/GpuActivityPanel.vue

@@ -25,7 +25,7 @@
           <div class="fc-cell__l">done</div>
         </div>
         <div class="fc-cell">
-          <div class="fc-cell__n" :class="q.error ? 'fc-bad' : ''">{{ q.error }}</div>
+          <div class="fc-cell__n" :class="q.error ? 'fc-weak' : ''">{{ q.error }}</div>
           <div class="fc-cell__l">errored</div>
         </div>
       </div>
@@ -95,7 +95,6 @@ onUnmounted(() => { if (pollId) clearInterval(pollId) })
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 .fc-cells { display: flex; gap: 28px; }
 .fc-cell__n {
   font-size: 20px; font-weight: 700; line-height: 1.1;
@@ -105,6 +104,4 @@ onUnmounted(() => { if (pollId) clearInterval(pollId) })
   font-size: 11px; text-transform: uppercase; letter-spacing: 0.04em;
   color: rgb(var(--v-theme-on-surface-variant));
 }
-.fc-good { color: rgb(var(--v-theme-success)); }
-.fc-bad { color: rgb(var(--v-theme-error)); }
 </style>

frontend/src/components/settings/GpuAgentCard.vue

@@ -367,11 +367,6 @@ async function onReprocess() {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
-.fc-section-h {
-  font-size: 13px; font-weight: 700; letter-spacing: 0.03em;
-  text-transform: uppercase; color: rgb(var(--v-theme-on-surface));
-}
 .fc-token {
   display: flex; align-items: center; gap: 4px;
   background: rgb(var(--v-theme-surface-light)); border-radius: 6px;
@@ -390,6 +385,4 @@ async function onReprocess() {
   font-size: 11px; text-transform: uppercase; letter-spacing: 0.04em;
   color: rgb(var(--v-theme-on-surface-variant));
 }
-.fc-good { color: rgb(var(--v-theme-success)); }
-.fc-weak { color: rgb(var(--v-theme-error)); }
 </style>

frontend/src/components/settings/GpuTriageCard.vue

@@ -155,11 +155,6 @@ async function onRecover(it) {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
-.fc-section-h {
-  font-size: 13px; font-weight: 700; letter-spacing: 0.03em;
-  text-transform: uppercase; color: rgb(var(--v-theme-on-surface));
-}
 .fc-queue { display: flex; gap: 24px; }
 .fc-q__n {
   font-size: 20px; font-weight: 700; line-height: 1.1;
@@ -169,8 +164,6 @@ async function onRecover(it) {
   font-size: 11px; text-transform: uppercase; letter-spacing: 0.04em;
   color: rgb(var(--v-theme-on-surface-variant));
 }
-.fc-good { color: rgb(var(--v-theme-success)); }
-.fc-weak { color: rgb(var(--v-theme-error)); }
 .fc-defect {
   display: flex; align-items: center; gap: 12px;
   background: rgb(var(--v-theme-surface-light)); border-radius: 8px;

frontend/src/components/settings/HeadsCard.vue

@@ -95,14 +95,10 @@
 
     <!-- Earned auto-apply -->
     <div class="fc-auto mt-6">
-      <div class="d-flex align-center mb-1" style="gap: 10px;">
-        <v-icon size="18" color="accent">mdi-lightning-bolt</v-icon>
-        <span class="fc-section-h">Auto-apply</span>
-        <v-switch
-          v-model="autoEnabled" :loading="settingBusy" hide-details density="compact"
-          color="success" class="ml-auto" @update:model-value="onToggleAuto"
-        />
-      </div>
+      <SettingToggleRow
+        v-model="autoEnabled" :loading="settingBusy"
+        icon="mdi-lightning-bolt" label="Auto-apply" @change="onToggleAuto"
+      />
       <p class="fc-muted text-body-2 mb-3">
         Graduated heads (⚡, with ≥ {{ autoMinPosInput }} examples) apply their tag
         on their own where they clear {{ Math.round((autoPrecisionInput || 0) * 100) }}%
@@ -111,17 +107,14 @@
       </p>
 
       <div class="d-flex mb-3" style="gap: 12px;">
-        <v-text-field
-          v-model.number="autoPrecisionInput" label="Precision target"
-          type="number" min="0.5" max="0.999" step="0.01" density="compact"
-          hide-details style="max-width: 200px;" :disabled="settingBusy"
+        <SettingNumberField
+          v-model="autoPrecisionInput" label="Precision target"
+          :min="0.5" :max="0.999" :step="0.01" :disabled="settingBusy"
           @change="onSaveSettings"
         />
-        <v-text-field
-          v-model.number="autoMinPosInput" label="Min examples to fire"
-          type="number" min="1" density="compact" hide-details
-          style="max-width: 200px;" :disabled="settingBusy"
-          @change="onSaveSettings"
+        <SettingNumberField
+          v-model="autoMinPosInput" label="Min examples to fire"
+          :min="1" :disabled="settingBusy" @change="onSaveSettings"
         />
       </div>
 
@@ -161,15 +154,11 @@
 
     <!-- Presentation chrome auto-hide (#141) -->
     <div class="fc-auto mt-6">
-      <div class="d-flex align-center mb-1" style="gap: 10px;">
-        <v-icon size="18" color="accent">mdi-image-off-outline</v-icon>
-        <span class="fc-section-h">Hide presentation chrome</span>
-        <v-switch
-          v-model="presentationEnabled" :loading="settingBusy" hide-details
-          density="compact" color="success" class="ml-auto"
-          @update:model-value="onTogglePresentation"
-        />
-      </div>
+      <SettingToggleRow
+        v-model="presentationEnabled" :loading="settingBusy"
+        icon="mdi-image-off-outline" label="Hide presentation chrome"
+        @change="onTogglePresentation"
+      />
       <p class="fc-muted text-body-2 mb-3">
         Auto-hide <code>banner</code> chrome from the gallery once a head has
         learned it (≥ {{ minPositives }} examples) and clears
@@ -180,16 +169,14 @@
         tag), it's flagged for review instead of buried. Every auto-hide is reversible.
       </p>
       <div class="d-flex mb-3" style="gap: 12px;">
-        <v-text-field
-          v-model.number="presentationThresholdInput" label="Hide confidence"
-          type="number" min="0.5" max="0.999" step="0.01" density="compact"
-          hide-details style="max-width: 200px;" :disabled="settingBusy"
+        <SettingNumberField
+          v-model="presentationThresholdInput" label="Hide confidence"
+          :min="0.5" :max="0.999" :step="0.01" :disabled="settingBusy"
           @change="onSavePresentation"
         />
-        <v-text-field
-          v-model.number="presentationConflictInput" label="Flag if content ≥"
-          type="number" min="0" max="1" step="0.05" density="compact"
-          hide-details style="max-width: 200px;" :disabled="settingBusy"
+        <SettingNumberField
+          v-model="presentationConflictInput" label="Flag if content ≥"
+          :min="0" :max="1" :step="0.05" :disabled="settingBusy"
           @change="onSavePresentation"
         />
       </div>
@@ -197,15 +184,11 @@
 
     <!-- Process auto-tagging (#1464): wip / editor screenshot -->
     <div class="fc-auto mt-6">
-      <div class="d-flex align-center mb-1" style="gap: 10px;">
-        <v-icon size="18" color="accent">mdi-progress-wrench</v-icon>
-        <span class="fc-section-h">Auto-tag work-in-progress</span>
-        <v-switch
-          v-model="processEnabled" :loading="settingBusy" hide-details
-          density="compact" color="success" class="ml-auto"
-          @update:model-value="onToggleProcess"
-        />
-      </div>
+      <SettingToggleRow
+        v-model="processEnabled" :loading="settingBusy"
+        icon="mdi-progress-wrench" label="Auto-tag work-in-progress"
+        @change="onToggleProcess"
+      />
       <p class="fc-muted text-body-2 mb-3">
         Auto-tag <code>wip</code> and <code>editor screenshot</code> process art
         once a head has learned them (≥ {{ minPositives }} examples) and clears
@@ -217,16 +200,14 @@
         manual tags, never its own guesses — so it can't run away. Every tag reversible.
       </p>
       <div class="d-flex mb-3" style="gap: 12px;">
-        <v-text-field
-          v-model.number="processThresholdInput" label="Tag confidence"
-          type="number" min="0.5" max="0.999" step="0.01" density="compact"
-          hide-details style="max-width: 200px;" :disabled="settingBusy"
+        <SettingNumberField
+          v-model="processThresholdInput" label="Tag confidence"
+          :min="0.5" :max="0.999" :step="0.01" :disabled="settingBusy"
           @change="onSaveProcess"
         />
-        <v-text-field
-          v-model.number="processConflictInput" label="Flag if content ≥"
-          type="number" min="0" max="1" step="0.05" density="compact"
-          hide-details style="max-width: 200px;" :disabled="settingBusy"
+        <SettingNumberField
+          v-model="processConflictInput" label="Flag if content ≥"
+          :min="0" :max="1" :step="0.05" :disabled="settingBusy"
           @change="onSaveProcess"
         />
       </div>
@@ -256,7 +237,7 @@
               <td class="fc-r fc-mono">{{ c.n_auto_applied }}</td>
               <td class="fc-r fc-mono">{{ c.n_misfires }}</td>
               <td class="fc-r fc-mono" :class="rateClass(c.misfire_rate)">
-                {{ ratePct(c.misfire_rate) }}
+                {{ pct(c.misfire_rate) }}
               </td>
               <td class="fc-r fc-mono">{{ c.n_underfires }}</td>
             </tr>
@@ -272,6 +253,9 @@ import { toast } from '../../utils/toast.js'
 import { computed, onMounted, onUnmounted, ref } from 'vue'
 
 import MaintenanceTile from '../common/MaintenanceTile.vue'
+import SettingNumberField from '../common/SettingNumberField.vue'
+import SettingToggleRow from '../common/SettingToggleRow.vue'
+import { useSettingSave } from '../../composables/useSettingSave.js'
 import { useHeadsStore } from '../../stores/heads.js'
 import { useMLStore } from '../../stores/ml.js'
 
@@ -285,7 +269,9 @@ let pollTimer = null
 const autoEnabled = ref(false)
 const autoPrecisionInput = ref(0.97)
 const autoMinPosInput = ref(30)
-const settingBusy = ref(false)
+// Shared settings-save flow (busy + toast + revert); `settingBusy` gates the
+// toggles/fields, `save` returns ok/false for the optimistic-switch revert.
+const { busy: settingBusy, save } = useSettingSave(mlSettings.patchSettings)
 const autoBusy = ref(false)
 const autoStatus = ref(null)
 const metricsData = ref(null)
@@ -395,81 +381,39 @@ function startAutoPoll() {
 function stopAutoPoll() { if (autoTimer) { clearInterval(autoTimer); autoTimer = null } }
 
 async function onToggleAuto(val) {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({ head_auto_apply_enabled: !!val })
-    toast({ text: val ? 'Auto-apply on' : 'Auto-apply off', type: 'success' })
-  } catch (e) {
-    autoEnabled.value = !val   // revert the switch
-    toast({ text: `Could not update: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  const ok = await save({ head_auto_apply_enabled: !!val },
+    { successMessage: val ? 'Auto-apply on' : 'Auto-apply off', errorPrefix: 'Could not update' })
+  if (!ok) autoEnabled.value = !val   // revert the switch
 }
 async function onSaveSettings() {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({
-      head_auto_apply_precision: Number(autoPrecisionInput.value),
-      head_auto_apply_min_positives: Number(autoMinPosInput.value),
-    })
-  } catch (e) {
-    toast({ text: `Could not save: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  await save({
+    head_auto_apply_precision: Number(autoPrecisionInput.value),
+    head_auto_apply_min_positives: Number(autoMinPosInput.value),
+  })
 }
 
 async function onTogglePresentation(val) {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({ presentation_auto_apply_enabled: !!val })
-    toast({ text: val ? 'Chrome auto-hide on' : 'Chrome auto-hide off', type: 'success' })
-  } catch (e) {
-    presentationEnabled.value = !val   // revert the switch
-    toast({ text: `Could not update: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  const ok = await save({ presentation_auto_apply_enabled: !!val },
+    { successMessage: val ? 'Chrome auto-hide on' : 'Chrome auto-hide off', errorPrefix: 'Could not update' })
+  if (!ok) presentationEnabled.value = !val   // revert the switch
 }
 async function onSavePresentation() {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({
-      presentation_auto_apply_threshold: Number(presentationThresholdInput.value),
-      presentation_conflict_threshold: Number(presentationConflictInput.value),
-    })
-  } catch (e) {
-    toast({ text: `Could not save: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  await save({
+    presentation_auto_apply_threshold: Number(presentationThresholdInput.value),
+    presentation_conflict_threshold: Number(presentationConflictInput.value),
+  })
 }
 
 async function onToggleProcess(val) {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({ process_auto_apply_enabled: !!val })
-    toast({ text: val ? 'WIP auto-tag on' : 'WIP auto-tag off', type: 'success' })
-  } catch (e) {
-    processEnabled.value = !val   // revert the switch
-    toast({ text: `Could not update: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  const ok = await save({ process_auto_apply_enabled: !!val },
+    { successMessage: val ? 'WIP auto-tag on' : 'WIP auto-tag off', errorPrefix: 'Could not update' })
+  if (!ok) processEnabled.value = !val   // revert the switch
 }
 async function onSaveProcess() {
-  settingBusy.value = true
-  try {
-    await mlSettings.patchSettings({
-      process_auto_apply_threshold: Number(processThresholdInput.value),
-      process_conflict_threshold: Number(processConflictInput.value),
-    })
-  } catch (e) {
-    toast({ text: `Could not save: ${e.message}`, type: 'error' })
-  } finally {
-    settingBusy.value = false
-  }
+  await save({
+    process_auto_apply_threshold: Number(processThresholdInput.value),
+    process_conflict_threshold: Number(processConflictInput.value),
+  })
 }
 function onPreview() { startSweep(true) }
 function onApplyNow() { startSweep(false) }
@@ -495,7 +439,6 @@ function sweepConcepts(run) {
     .sort((a, b) => b.applied - a.applied)
 }
 function sweepTotal(run) { return run?.n_applied ?? 0 }
-function ratePct(x) { return x == null ? '—' : `${Math.round(x * 100)}%` }
 function rateClass(x) {
   if (x == null) return ''
   if (x <= 0.03) return 'fc-good'
@@ -526,12 +469,6 @@ function relTime(iso) {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
-
-.fc-section-h {
-  font-size: 13px; font-weight: 700; letter-spacing: 0.03em;
-  text-transform: uppercase; color: rgb(var(--v-theme-on-surface));
-}
 .fc-auto {
   border-top: 1px solid rgb(var(--v-theme-surface-light)); padding-top: 16px;
 }
@@ -588,7 +525,5 @@ function relTime(iso) {
   background: rgb(var(--v-theme-surface-light));
   padding: 1px 6px; border-radius: 999px;
 }
-.fc-good { color: rgb(var(--v-theme-success)); }
 .fc-ok { color: rgb(var(--v-theme-on-surface)); }
-.fc-weak { color: rgb(var(--v-theme-error)); }
 </style>

frontend/src/components/settings/MLBackfillCard.vue

@@ -39,13 +39,14 @@
 import { toast } from '../../utils/toast.js'
 import { onMounted, ref } from 'vue'
 import { useMLStore } from '../../stores/ml.js'
+import { useSettingSave } from '../../composables/useSettingSave.js'
 import MaintenanceTile from '../common/MaintenanceTile.vue'
 import QueueStatusBar from './QueueStatusBar.vue'
 const store = useMLStore()
+const { busy: saving, save } = useSettingSave(store.patchSettings)
 const busy = ref(false)
 const done = ref(false)
 const enabled = ref(true)
-const saving = ref(false)
 onMounted(async () => {
   try {
     await store.loadSettings()
@@ -55,21 +56,12 @@ onMounted(async () => {
   } catch { /* non-fatal */ }
 })
 async function onToggle() {
-  saving.value = true
-  try {
-    await store.patchSettings({ cpu_embed_enabled: enabled.value })
-    toast({
-      text: enabled.value
-        ? 'CPU embedding on — imports queue embeds for the ml-worker'
-        : 'CPU embedding off — the GPU embed backfill owns whole-image embeds',
-      type: 'success',
-    })
-  } catch (e) {
-    toast({ text: `Could not save: ${e.message}`, type: 'error' })
-    enabled.value = !enabled.value
-  } finally {
-    saving.value = false
-  }
+  const ok = await save({ cpu_embed_enabled: enabled.value }, {
+    successMessage: enabled.value
+      ? 'CPU embedding on — imports queue embeds for the ml-worker'
+      : 'CPU embedding off — the GPU embed backfill owns whole-image embeds',
+  })
+  if (!ok) enabled.value = !enabled.value
 }
 async function run() {
   busy.value = true
@@ -80,5 +72,4 @@ async function run() {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 </style>

frontend/src/components/settings/MaintenancePanel.vue

@@ -24,6 +24,7 @@
         the CPU fallback.
       </p>
       <div class="fc-tile-stack">
+        <VideoEmbeddingCard />
         <GpuAgentCard />
         <GpuTriageCard />
         <MLBackfillCard />
@@ -36,7 +37,6 @@
         Suggestion thresholds, trained heads and tag aliases.
       </p>
       <div class="fc-tile-stack">
-        <MLThresholdSliders />
         <CropProposersCard />
         <HeadsCard />
         <AliasTable />
@@ -77,7 +77,7 @@ import ArchiveReextractCard from './ArchiveReextractCard.vue'
 import MissingFileRepairCard from './MissingFileRepairCard.vue'
 import GpuTriageCard from './GpuTriageCard.vue'
 import DbMaintenanceCard from './DbMaintenanceCard.vue'
-import MLThresholdSliders from './MLThresholdSliders.vue'
+import VideoEmbeddingCard from './VideoEmbeddingCard.vue'
 import CropProposersCard from './CropProposersCard.vue'
 import HeadsCard from './HeadsCard.vue'
 import GpuAgentCard from './GpuAgentCard.vue'

frontend/src/components/settings/VideoDedupCard.vue

@@ -78,6 +78,7 @@
 import { computed, ref } from 'vue'
 
 import { useMaintenanceTask } from '../../composables/useMaintenanceTask.js'
+import { humanBytes } from '../../utils/bytes.js'
 import MaintenanceTile from '../common/MaintenanceTile.vue'
 import QueueStatusBar from './QueueStatusBar.vue'
 
@@ -98,14 +99,6 @@ const summaryType = computed(() => {
   return summary.value && summary.value.redundant > 0 ? 'info' : 'success'
 })
 
-function humanBytes (n) {
-  const b = Number(n || 0)
-  if (b >= 1 << 30) return (b / (1 << 30)).toFixed(1) + ' GB'
-  if (b >= 1 << 20) return (b / (1 << 20)).toFixed(1) + ' MB'
-  if (b >= 1 << 10) return (b / (1 << 10)).toFixed(1) + ' KB'
-  return b + ' B'
-}
-
 // The confirm dialog gates the destructive apply; close it, then run.
 function apply () {
   confirmOpen.value = false

frontend/src/components/settings/VideoEmbeddingCard.vue

@@ -12,17 +12,17 @@
       </div>
       <v-row>
         <v-col cols="12" sm="6">
-          <v-text-field
-            v-model.number="local.video_frame_interval_seconds"
-            label="Frame interval (s)" type="number" min="0.5" step="0.5"
-            density="comfortable" hide-details @change="save"
+          <SettingNumberField
+            v-model="local.video_frame_interval_seconds"
+            label="Frame interval (s)" :min="0.5" :step="0.5"
+            density="comfortable" max-width="none" @change="onSave"
           />
         </v-col>
         <v-col cols="12" sm="6">
-          <v-text-field
-            v-model.number="local.video_max_frames"
-            label="Max frames" type="number" min="1" step="1"
-            density="comfortable" hide-details @change="save"
+          <SettingNumberField
+            v-model="local.video_max_frames"
+            label="Max frames" :min="1" :step="1"
+            density="comfortable" max-width="none" @change="onSave"
           />
         </v-col>
       </v-row>
@@ -32,21 +32,23 @@
 </template>
 
 <script setup>
-import { toast } from '../../utils/toast.js'
 import { reactive, watch } from 'vue'
 import { useMLStore } from '../../stores/ml.js'
 import MaintenanceTile from '../common/MaintenanceTile.vue'
+import SettingNumberField from '../common/SettingNumberField.vue'
+import { useSettingSave } from '../../composables/useSettingSave.js'
 
 const store = useMLStore()
+const { save } = useSettingSave(store.patchSettings)
 const local = reactive({})
 watch(() => store.settings, (s) => { if (s) Object.assign(local, s) }, { immediate: true })
 
-async function save() {
-  const patch = {
-    video_frame_interval_seconds: local.video_frame_interval_seconds,
-    video_max_frames: local.video_max_frames
-  }
-  try { await store.patchSettings(patch) }
-  catch (e) { toast({ text: e.message, type: 'error' }) }
+// SettingNumberField clamps interval to ≥0.5 and max-frames to ≥1 before this
+// fires, so an out-of-range value never reaches the API.
+function onSave() {
+  save({
+    video_frame_interval_seconds: Number(local.video_frame_interval_seconds),
+    video_max_frames: Number(local.video_max_frames),
+  })
 }
 </script>

frontend/src/composables/useSettingSave.js

@@ -0,0 +1,33 @@
+import { ref } from 'vue'
+import { toast } from '../utils/toast.js'
+
+// The shared "persist a settings patch" flow for the ML settings cards. Flips a
+// busy flag, calls the store's patch (which rethrows on failure), toasts
+// success/error, and returns true/false so a toggle handler can revert its
+// optimistic switch on failure. Centralises the try/catch/toast the cards each
+// hand-rolled (HeadsCard x6, CropProposersCard, MLBackfillCard) — and where the
+// threshold-clamp drifted; the clamp now lives in <SettingNumberField>.
+//
+// Pass the store's patch fn, e.g. useSettingSave(ml.patchSettings).
+export function useSettingSave(patchFn) {
+  const busy = ref(false)
+
+  // opts.successMessage — toast on success (toggles announce their new state;
+  // silent field-saves omit it). opts.errorPrefix — the failure toast prefix
+  // ("Could not save" default; toggles used "Could not update").
+  async function save(patch, { successMessage = '', errorPrefix = 'Could not save' } = {}) {
+    busy.value = true
+    try {
+      await patchFn(patch)
+      if (successMessage) toast({ text: successMessage, type: 'success' })
+      return true
+    } catch (e) {
+      toast({ text: `${errorPrefix}: ${e.message}`, type: 'error' })
+      return false
+    } finally {
+      busy.value = false
+    }
+  }
+
+  return { busy, save }
+}

frontend/src/stores/system.js

@@ -5,6 +5,18 @@ import { useApi } from '../composables/useApi.js'
 export const useSystemStore = defineStore('system', () => {
   const api = useApi()
   const healthy = ref(null)  // null=unknown, true=ok, false=down
+  // What the instance says it is. Since milestone 318 stopped publishing
+  // version image tags, this is the only answer to "which build is this?" —
+  // there is no registry name left to check it against.
+  //
+  // Three states, and collapsing any two of them would lie:
+  //   buildLoaded=false            we have not asked yet   -> render nothing
+  //   buildLoaded=true, version='' the build cannot say    -> render "unknown"
+  //   buildLoaded=true, version=x  this build is x
+  // A blank footer would read as "no version", which is a different claim.
+  const buildVersion = ref('')
+  const buildChannel = ref('')
+  const buildLoaded = ref(false)
   const stats = ref(null)
   const statsLoading = ref(false)
 
@@ -12,8 +24,17 @@ export const useSystemStore = defineStore('system', () => {
     try {
       const body = await api.get('/api/health')
       healthy.value = body.status === 'ok'
+      // Absent means "cannot say" — the server omits these rather than
+      // sending empty strings, so `?? ''` preserves that rather than
+      // inventing a value for it.
+      buildVersion.value = body.version ?? ''
+      buildChannel.value = body.channel ?? ''
+      buildLoaded.value = true
     } catch {
       healthy.value = false
+      // Deliberately NOT setting buildLoaded: a failed health call tells us
+      // nothing about the build, and claiming "unknown" would present a
+      // network blip as a defective image.
     }
   }
 
@@ -26,5 +47,8 @@ export const useSystemStore = defineStore('system', () => {
     }
   }
 
-  return { healthy, stats, statsLoading, refreshHealth, refreshStats }
+  return {
+    healthy, stats, statsLoading, refreshHealth, refreshStats,
+    buildVersion, buildChannel, buildLoaded,
+  }
 })

frontend/src/styles/app.css

@@ -40,6 +40,25 @@
    emits, so no specificity/reorder fight — no !important needed. */
 .fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 
+/* Section sub-heading in settings cards (DRY pass #161): was redefined
+   identically in 4 cards, and TranslationCard used the class with NO local def
+   so its section headers rendered unstyled. Now one global utility. */
+.fc-section-h {
+  font-size: 13px; font-weight: 700; letter-spacing: 0.03em;
+  text-transform: uppercase; color: rgb(var(--v-theme-on-surface));
+}
+
+/* Status text colours (DRY pass #161): fc-good = success, fc-weak = error,
+   consolidated from the GPU / heads cards. fc-ok is intentionally NOT global —
+   it means on-surface in HeadsCard but success in QueuesTable.
+
+   No `.fc-bad` (#3072): it was defined locally and identically in the Downloads
+   and GPU activity panels, and it is fc-weak under a second name — GpuAgentCard
+   and GpuActivityPanel were colouring the same "errored" count with different
+   class names. Both now use fc-weak. Reach for fc-weak, not a new synonym. */
+.fc-good { color: rgb(var(--v-theme-success)); }
+.fc-weak { color: rgb(var(--v-theme-error)); }
+
 /* Vuetify 4 dropped its global CSS reset (normalisation moved into each
    component). FC's layouts assumed the reset zeroed margins on text elements, so
    restore just that — the "minimal reset" from the v4 upgrade guide — inside

frontend/src/utils/bytes.js

@@ -0,0 +1,17 @@
+// Human-readable byte sizes for maintenance summaries ("2.4 GB reclaimable").
+//
+// Promoted out of the cleanup cards, which had grown byte-identical private
+// copies (VideoDedupCard, GatedPurgeCard) and were about to grow a third for
+// the attachment reclaim. Binary units (1 KB = 1024 B) — these numbers come
+// from st_size / SUM(size_bytes), so they describe disk, not marketing.
+//
+// NOT the same shape as the `formatBytes` helpers in SystemStatsCards,
+// BackupRunsTable and PostCard — those differ in units, precision and
+// zero-handling. Left alone deliberately rather than force-fitted here.
+export function humanBytes (n) {
+  const b = Number(n || 0)
+  if (b >= 1 << 30) return (b / (1 << 30)).toFixed(1) + ' GB'
+  if (b >= 1 << 20) return (b / (1 << 20)).toFixed(1) + ' MB'
+  if (b >= 1 << 10) return (b / (1 << 10)).toFixed(1) + ' KB'
+  return b + ' B'
+}

frontend/src/utils/platformColor.js

@@ -1,8 +1,10 @@
 // Single source of truth for platform → color + icon mapping. Used by
-// PlatformChip and any other GS-style platform-tagged surface. The six
+// PlatformChip and any other GS-style platform-tagged surface. The five
 // platforms FC supports map 1:1 to the GS palette; unknown platforms fall
-// back to grey + mdi-web. Operator-confirmed scope 2026-05-27. The ICONS key
-// set is pinned against backend known_platform_keys() by
+// back to grey + mdi-web — which is deliberately what a retired platform
+// hits: a pre-#3069 deviantart source row still renders, as its raw key on
+// a grey chip. Operator-confirmed scope 2026-05-27. The ICONS key set is
+// pinned against backend known_platform_keys() by
 // tests/test_fe_be_contract.py.
 
 const ICONS = {
@@ -11,7 +13,6 @@ const ICONS = {
   hentaifoundry: 'mdi-palette',
   discord:       'mdi-discord',
   pixiv:         'mdi-alpha-p-box',
-  deviantart:    'mdi-deviantart',
 }
 
 const COLORS = {
@@ -20,7 +21,6 @@ const COLORS = {
   hentaifoundry: 'purple',
   discord:       'indigo',
   pixiv:         'blue',
-  deviantart:    'green',
 }
 
 const LABELS = {
@@ -29,7 +29,6 @@ const LABELS = {
   hentaifoundry: 'HentaiFoundry',
   discord:       'Discord',
   pixiv:         'Pixiv',
-  deviantart:    'DeviantArt',
 }
 
 export function platformIcon(platform) {

frontend/src/views/CleanupView.vue

@@ -19,14 +19,16 @@
     </section>
 
     <section class="fc-section">
-      <h3 class="fc-section__title">Duplicates &amp; posts</h3>
+      <h3 class="fc-section__title">Duplicates &amp; leftovers</h3>
       <p class="fc-section__hint">
-        Tidy post records, duplicates and locked-preview leftovers.
+        Tidy post records, duplicates, locked-preview leftovers and attachments
+        that outlived what they belonged to.
       </p>
       <div class="fc-tile-grid">
         <PostMaintenanceCard />
         <VideoDedupCard />
         <GatedPurgeCard />
+        <AttachmentReclaimCard />
       </div>
     </section>
 
@@ -60,6 +62,7 @@ import SingleColorAuditCard from '../components/cleanup/SingleColorAuditCard.vue
 import PostMaintenanceCard from '../components/settings/PostMaintenanceCard.vue'
 import VideoDedupCard from '../components/settings/VideoDedupCard.vue'
 import GatedPurgeCard from '../components/settings/GatedPurgeCard.vue'
+import AttachmentReclaimCard from '../components/settings/AttachmentReclaimCard.vue'
 import TagMaintenanceCard from '../components/settings/TagMaintenanceCard.vue'
 import DangerZoneCard from '../components/settings/DangerZoneCard.vue'
 </script>

frontend/src/views/ExploreView.vue

@@ -284,7 +284,6 @@ onUnmounted(() => {
 </script>
 
 <style scoped>
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 
 /* Full-height workspace under the sticky top nav. --fc-nav-h is the nav's REAL
    measured height (set by TopNav) — a hardcoded 64px here overflowed the

frontend/src/views/SettingsView.vue

@@ -54,6 +54,21 @@
         <MaintenancePanel />
       </v-window-item>
     </v-window>
+
+    <!-- Which build is this? With no version image tags (milestone 318) the
+         instance's own report is the only answer, so it is shown rather than
+         hidden. The instinct to treat it as information disclosure does not
+         survive contact: the JS bundle and asset hashes fingerprint the build
+         anyway, and "I'm on 2026.08.28.1249" is the single most useful line in
+         a bug report.
+
+         Channel sits BESIDE the version, never inside it (rule 149) — a
+         `-dev` suffix would read as a 0 segment to the extension's comparator
+         and make every dev build compare equal (#2993). -->
+    <div v-if="system.buildLoaded" class="text-caption text-medium-emphasis text-center mt-8">
+      FabledCurator {{ system.buildVersion || 'unknown' }}
+      <span v-if="system.buildChannel"> · {{ system.buildChannel }}</span>
+    </div>
   </v-container>
 </template>
 

frontend/src/views/TagsView.vue

@@ -349,7 +349,6 @@ async function onDeleteTagConfirm() {
 .fc-tags__sentinel {
   display: flex; justify-content: center; padding: 32px 0; min-height: 60px;
 }
-.fc-muted { color: rgb(var(--v-theme-on-surface-variant)); }
 .fc-merge-preview {
   padding: 10px 12px;
   border: 1px solid rgba(var(--v-theme-on-surface), 0.12);

frontend/test/components/browserExtensionCard.spec.js

@@ -0,0 +1,72 @@
+// @vitest-environment happy-dom
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest'
+import { nextTick } from 'vue'
+
+import BrowserExtensionCard from '../../src/components/settings/BrowserExtensionCard.vue'
+import { freshPinia, mountComponent } from '../support/mountComponent.js'
+
+// useApi is a thin fetch wrapper, so the seam is fetch itself (same shape as
+// showcase.spec.js) rather than a module mock.
+function stubApi(manifest) {
+  globalThis.fetch = vi.fn(async (url) => {
+    const payload = String(url).includes('/api/extension/manifest')
+      ? manifest
+      : { key: 'test-key' }
+    return {
+      ok: true, status: 200, statusText: '200',
+      text: async () => JSON.stringify(payload),
+    }
+  })
+}
+
+async function mountCard(manifest) {
+  stubApi(manifest)
+  const w = mountComponent(BrowserExtensionCard, { pinia: freshPinia() })
+  // onMounted fires two fetches (manifest + key) and each resolves through a
+  // chain of microtasks. Yielding to a macrotask drains the whole queue, which
+  // a fixed number of nextTicks would only do by luck.
+  await new Promise((resolve) => setTimeout(resolve, 0))
+  await nextTick()
+  return w
+}
+
+const INSTALLED = {
+  installed: true,
+  version: '1.0.3499884',
+  xpi_url: '/extension/fabledcurator-1.0.3499884.xpi',
+  latest_url: '/extension/fabledcurator-latest.xpi',
+  sha256: 'abc',
+}
+
+describe('BrowserExtensionCard — channel', () => {
+  beforeEach(() => { vi.restoreAllMocks() })
+  afterEach(() => { delete globalThis.fetch })
+
+  it('names the channel the instance reports', async () => {
+    // The point of the whole channel scheme: an operator can tell a dev
+    // instance from a main one without installing anything.
+    const w = await mountCard({ ...INSTALLED, channel: 'dev' })
+    expect(w.text()).toContain('dev')
+  })
+
+  it('shows the version and the channel as SEPARATE text, never merged', async () => {
+    // Regression guard with teeth: the tempting shortcut is a `-dev` version
+    // suffix, and that is precisely what breaks the extension's comparator —
+    // it parses each dotted segment with parseInt, so a suffixed segment reads
+    // as 0 and every dev build compares equal to every other. If someone ever
+    // "simplifies" by folding the channel into the version, the version text
+    // stops being the bare derived number and this fails.
+    const w = await mountCard({ ...INSTALLED, channel: 'dev' })
+    expect(w.text()).toContain('v1.0.3499884')
+    expect(w.text()).not.toContain('1.0.3499884-dev')
+  })
+
+  it('renders no channel when the instance declares none', async () => {
+    // A locally-built image, or one predating the field. The card must read
+    // exactly as it did before the channel existed rather than inventing an
+    // "unknown" badge — absence is a normal answer here, not a fault.
+    const w = await mountCard(INSTALLED)
+    expect(w.text()).toContain('v1.0.3499884')
+    expect(w.findAll('v-chip')).toHaveLength(0)
+  })
+})

frontend/test/systemBuild.spec.js

@@ -0,0 +1,96 @@
+import { describe, it, expect, beforeEach, afterEach, vi } from 'vitest'
+import { setActivePinia, createPinia } from 'pinia'
+import { useSystemStore } from '../src/stores/system.js'
+
+// Which build am I running? Milestone 318 stopped publishing version image
+// tags, so the instance's own report is the ONLY answer — there is no registry
+// name left to check it against. That promotes this from a convenience to the
+// mechanism, and it means the three states below have to stay distinct: a
+// wrong answer here has nothing to contradict it.
+//
+//   not asked yet        -> render nothing
+//   asked, no version    -> render "unknown"
+//   asked, has a version -> render it
+//
+// Collapsing the first two would show "unknown" during every page load, and
+// collapsing either into a blank would read as "no version", which is a
+// different and false claim.
+
+function stubHealth(body, { fail = false } = {}) {
+  globalThis.fetch = vi.fn(async () => {
+    if (fail) throw new Error('network down')
+    return {
+      ok: true, status: 200, statusText: '200',
+      text: async () => JSON.stringify(body),
+    }
+  })
+}
+
+describe('system store — build identity', () => {
+  beforeEach(() => setActivePinia(createPinia()))
+  afterEach(() => { vi.restoreAllMocks(); delete globalThis.fetch })
+
+  it('starts having asked nothing, so the footer renders nothing', () => {
+    const s = useSystemStore()
+    expect(s.buildLoaded).toBe(false)
+  })
+
+  it('reports the version and channel the instance claims', async () => {
+    stubHealth({ status: 'ok', version: '2026.08.28.1249', channel: 'dev' })
+    const s = useSystemStore()
+    await s.refreshHealth()
+
+    expect(s.buildLoaded).toBe(true)
+    expect(s.buildVersion).toBe('2026.08.28.1249')
+    expect(s.buildChannel).toBe('dev')
+  })
+
+  it('keeps the channel OUT of the version string', async () => {
+    // The tempting shortcut is a `-dev` suffix. The extension's comparator
+    // parses each dotted segment with parseInt, so a suffixed segment reads as
+    // 0 and every dev build compares equal to every other — #2993 exactly
+    // (rule 149). If anyone ever "simplifies" by folding them together, the
+    // version stops being the bare derived number and this fails.
+    stubHealth({ status: 'ok', version: '2026.08.28.1249', channel: 'dev' })
+    const s = useSystemStore()
+    await s.refreshHealth()
+
+    expect(s.buildVersion).toBe('2026.08.28.1249')
+    expect(s.buildVersion).not.toContain('dev')
+  })
+
+  it('treats an absent version as "cannot say", not as a value', async () => {
+    // A locally-built image, or one predating the field. The server omits the
+    // key rather than sending an empty string; `?? ''` must preserve that
+    // rather than inventing something. The view renders "unknown" from it.
+    stubHealth({ status: 'ok' })
+    const s = useSystemStore()
+    await s.refreshHealth()
+
+    expect(s.buildLoaded).toBe(true)
+    expect(s.buildVersion).toBe('')
+    expect(s.buildChannel).toBe('')
+  })
+
+  it('reports a version with no channel without inventing one', async () => {
+    stubHealth({ status: 'ok', version: '2026.08.28.1249' })
+    const s = useSystemStore()
+    await s.refreshHealth()
+
+    expect(s.buildVersion).toBe('2026.08.28.1249')
+    expect(s.buildChannel).toBe('')
+  })
+
+  it('does not claim "unknown" when the health call itself failed', async () => {
+    // A network blip says nothing about the build. Marking it loaded here
+    // would present a transient failure as a defective image — and since
+    // nothing else names the build, there would be no second source to
+    // correct the impression.
+    stubHealth(null, { fail: true })
+    const s = useSystemStore()
+    await s.refreshHealth()
+
+    expect(s.healthy).toBe(false)
+    expect(s.buildLoaded).toBe(false)
+  })
+})

scripts/artifacts.sh

@@ -0,0 +1,201 @@
+#!/bin/sh
+# Single definition of WHAT EACH PUBLISHED ARTIFACT IS BUILT FROM, and the
+# version derived from it. Milestone 313; generalises the shape
+# extension/scripts/packaging.sh established for the extension alone.
+#
+# "Built from" is deliberately wider than "copied into". A file that DECIDES an
+# artifact's identity is part of what that artifact is built from even though it
+# never reaches the image — see DERIVER below, and #3156 for the same finding
+# about packaging.sh.
+#
+# Four artifacts, four independent versions. An artifact whose shipped files
+# did not change keeps its version and does not rebuild — that is the whole
+# point, and it is why each path set must match its Dockerfile rather than
+# being a plausible guess. Getting a set wrong is quiet in BOTH directions:
+#
+#   too narrow -> a pin serves stale bytes, because the version did not move
+#                 when the content did. This is the dangerous one.
+#   too wide   -> the artifact re-versions and rebuilds for a change it does
+#                 not ship. Merely wasteful.
+#
+# tests/test_artifact_paths.py asserts every COPY source in each Dockerfile is
+# covered here, so adding a COPY without updating this file fails CI.
+#
+# POSIX sh only — CI's run shell is busybox on some paths.
+#
+# -f (no pathname expansion) is load-bearing for the whole script: the lists
+# below are iterated with deliberate word-splitting, and without it the shell
+# would glob `frontend/test/**` against the working tree and silently narrow
+# the pattern. Callers substituting the output need their own `set -f` too;
+# the two guards protect different expansions.
+set -euf
+
+ROOT=$(git rev-parse --show-toplevel)
+
+# --- what each artifact ships ------------------------------------------------
+#
+# Each set includes its own Dockerfile and requirements: changing a base image
+# or a pin changes the artifact just as surely as changing a source file.
+#
+# web (Dockerfile, context `.`) — the runtime stage copies backend/, alembic/,
+# alembic.ini, entrypoint.sh and requirements.txt; the frontend-builder stage
+# copies frontend/ and the runtime takes its `dist` output.
+#
+# frontend/test is excluded: `npm run build` is vite, which builds from src/,
+# index.html and public/ and never reads test/. It lands in the builder layer
+# but not in `dist`, so it cannot reach the shipped image.
+#
+# The web image ALSO bundles the signed XPI (build.yml downloads it into
+# frontend/public/extension/ before the docker build), so an extension change
+# changes the web image. The extension's packaged set is appended in cmd_paths
+# rather than restated — one definition, per #2397.
+WEB_PATHS='Dockerfile requirements.txt backend alembic alembic.ini entrypoint.sh frontend :(exclude)frontend/test :(exclude)frontend/test/**'
+
+# ml (Dockerfile.ml, context `.`) — no frontend, no extension. Note it copies
+# BOTH requirements-ml.txt and requirements.txt.
+ML_PATHS='Dockerfile.ml requirements-ml.txt requirements.txt backend alembic alembic.ini entrypoint.sh'
+
+# agent (agent/Dockerfile, context `agent`) — copies requirements.txt and
+# fc_agent only. agent/README.md, agent/docker-compose.yml and agent/ruff.toml
+# live in the directory but never reach the image, so they must not re-version
+# it: this is deliberately NOT `agent/`.
+AGENT_PATHS='agent/Dockerfile agent/requirements.txt agent/fc_agent'
+
+# This file. It is copied into no image and it is still part of what the web
+# image is built from, because it DECIDES the FC_VERSION baked into that image
+# (#3202). Same finding as #3156 about packaging.sh, one level up.
+#
+# Why web and nothing else. Every artifact stamps `fc.revision`, but only web
+# also stamps a version (build.yml line ~488 feeds `version web` to the
+# FC_VERSION build arg; ml and agent ask for `revision` alone, and the
+# extension takes its version from packaging.sh). For a revision-only artifact
+# this file needs no entry: any change to how the revision is COMPUTED changes
+# the derived value, which then disagrees with the label on the published image
+# and forces a rebuild. That mechanism is self-correcting because it compares
+# against a string stamped into a real artifact.
+#
+# The version is compared against nothing, so it has no such backstop. Before
+# this entry, a change to cmd_version alone left every artifact's revision
+# untouched, the reuse check hit, the build was skipped, and the published
+# image went on reporting the OLD version format — silently, until some
+# unrelated commit happened to force a rebuild. Milestone 318 step 5 is the
+# worked instance: b3989d0 and 5771fd5 share revision fb2c4d5b80be while the
+# version moved 2026.8.28.1249 -> 2026.08.28.1249. It cost nothing only because
+# FC_VERSION did not exist until one commit later.
+#
+# Named as a file, not as `scripts`: release_notes.py lives beside it and only
+# READS derived values, so it decides nothing and must not re-version anything.
+# A future script that derives an identity belongs here explicitly.
+DERIVER='scripts/artifacts.sh'
+
+
+usage() {
+  echo "usage: artifacts.sh {paths|revision|version} {web|ml|agent|extension}" >&2
+  exit 2
+}
+
+# The extension's packaged set, read from its own definition rather than
+# copied. packaging.sh emits `:(exclude)extension/...` entries, so the bare
+# `extension` include has to come with them.
+ext_paths() {
+  echo "extension $(sh "$ROOT/extension/scripts/packaging.sh" pathspec)"
+}
+
+cmd_paths() {
+  case "$1" in
+    web)       echo "$WEB_PATHS $DERIVER $(ext_paths)" ;;
+    ml)        echo "$ML_PATHS" ;;
+    agent)     echo "$AGENT_PATHS" ;;
+    extension) ext_paths ;;
+    *)         usage ;;
+  esac
+}
+
+# "<unix ts> <sha>" of the newest commit touching this artifact's shipped set.
+# Unquoted on purpose: the pathspec must word-split into separate args.
+# Globbing is already off script-wide.
+newest() {
+  # shellcheck disable=SC2046
+  set -- "$(cd "$ROOT" && git log --format='%ct %H' HEAD -- $(cmd_paths "$1") \
+            | sort -n | tail -1)"
+  if [ -z "$1" ]; then
+    echo "artifacts.sh: no commit touches this artifact's shipped files" >&2
+    exit 1
+  fi
+  echo "$1"
+}
+
+# Formatted through git rather than date(1): busybox date does not reliably
+# accept `-d @<epoch>`, and git's own --date=format-local is available wherever
+# git is. TZ=UTC so the value does not depend on the runner's timezone.
+fmt() {
+  (cd "$ROOT" && TZ=UTC git show -s --format=%cd --date="format-local:$2" "$1")
+}
+
+# The IDENTITY of an artifact's content: the commit its shipped files last
+# changed in. This is what decides whether a build can be skipped.
+#
+# It is published as the `fc.revision` LABEL on the image itself, and read
+# back off the moving channel tag — not as a tag of its own (milestone 318
+# step 3). A tag would be a name minted per build that only one thing reads,
+# which is what rule 145 narrowed against; it would also be prunable under the
+# registry's keep_pattern (#3157), so the cache would silently expire.
+#
+# A published image with no such label reads as a MISS and rebuilds. That is
+# the migration path, not a fault: `imagetools create` copies a manifest and
+# config labels are not manifest annotations, so the reuse path cannot stamp
+# one and there is nothing to backfill. Each artifact pays one rebuild, once.
+cmd_revision() {
+  echo "$(newest "$1")" | cut -d' ' -f2 | cut -c1-12
+}
+
+# The VERSION: `YYYY.MM.DD.HHMM`, zero-padded, UTC. One shape across the whole
+# family (note #3127 §1, rule 148) — the number an instance reports about
+# itself, and, with a `v` in front, the release tag naming the same build.
+#
+# Zero-padded since 2026-08-28. This stripped leading zeros until then, on the
+# reasoning that every segment should read as a plain integer — which never
+# held, since comparison strips them on parse anyway. Padding costs nothing,
+# sorts lexically as well as numerically, and keeps this project emitting the
+# same string as its siblings: unpadded, a `2026.8.28.1432` here sits beside a
+# `2026.08.28.1432` there, two shapes one character apart. Two obviously
+# different formats are safer than two nearly identical ones.
+#
+# Comparison is numeric per dot-segment, so `08` and `8` are equal and nothing
+# already published is reordered by the change.
+#
+# HHMM is not decoration: it is what makes the value unique per build with no
+# lookup. A date alone collides on the second build of a day, and resolving
+# that needs a `.N` suffix, which needs asking the registry what already
+# exists — at which point two lanes derive different answers for one source
+# and the shared-signature property is lost.
+cmd_version() {
+  # The extension is the one artifact this script does not FORMAT, only route.
+  # AMO's version grammar forbids leading zeros, so the extension emits the
+  # same numbers unpadded (#3138) — a rendering exception, documented in
+  # packaging.sh beside the signing step that has to obey it. Delegating keeps
+  # one answer per artifact: `artifacts.sh version extension` and
+  # `packaging.sh version` cannot drift into two.
+  #
+  # The direction is deliberate. artifacts.sh already asks packaging.sh for the
+  # extension's PATH SET (ext_paths above), so the version has to flow the same
+  # way; reversing it would have packaging.sh call back into this script, which
+  # would call packaging.sh for the paths again.
+  if [ "$1" = extension ]; then
+    sh "$ROOT/extension/scripts/packaging.sh" version
+    return
+  fi
+  sha=$(echo "$(newest "$1")" | cut -d' ' -f2)
+  # One git call for the whole string rather than four and a sed. git's
+  # format-local takes the complete format, and doing it in pieces was only
+  # ever there to strip the padding between them.
+  fmt "$sha" '%Y.%m.%d.%H%M'
+}
+
+[ $# -ge 2 ] || usage
+case "$1" in
+  paths)    cmd_paths "$2" ;;
+  revision) cmd_revision "$2" ;;
+  version)  cmd_version "$2" ;;
+  *)        usage ;;
+esac

scripts/release_notes.py

@@ -0,0 +1,312 @@
+"""Publish a Forgejo release whose body is derived from git, not written by hand.
+
+Milestone 318 step 2 took the build consequence away from a `v*` tag: `main`
+has already built and published the commit by the time anyone tags it, and
+rebuilding would re-push `:c-<sha>`, which rule 145 forbids even when the bytes
+match. That left the tag with nothing to do. This gives it the job it has left.
+
+**The half of the question a version string cannot answer.** Step 6 puts
+`2026.08.28.2208` in the Settings footer, so an operator can say which build
+they are running. They still cannot say what is in it that was not in the one
+they ran last month. A dated release carrying the commits since the previous
+one is the object that interprets the identifier (note #3127 §5).
+
+**Derived, so it cannot drift.** The alternative is a hand-maintained
+`CHANGELOG.md`, which goes aspirational the first time someone forgets — and
+nothing ever catches it, because there is no second source to disagree with.
+Every line below comes out of `git log` at publish time.
+
+**Optional by construction.** Release tags are bookmarks: cut one when you will
+want to point at that day by name, otherwise don't. FC went twelve weeks
+without one and nothing was wrong (note #3127 §0). This runs on a tag push and
+on nothing else — deliberately no schedule and no auto-tag on merge, either of
+which would turn an optional bookmark back into ceremony.
+
+## Finding the previous release
+
+`git describe --exclude <this tag>`, which walks ANCESTRY, not a sorted list.
+That is not fussiness: this repo's existing tags are the old `v26.05.22.0`
+shape and the next one will be rule 148's `v2026.08.28.2208`. Lexicographically
+`v2026...` sorts BEFORE `v26...` — every release from here on would report its
+predecessor as itself-or-nothing and emit a changelog covering the entire
+history. Ancestry is immune to the shape change, and it is also the more honest
+question: "what is in this that was not in the last one" IS a reachability
+question.
+
+## Re-runs update, they do not fall through
+
+Note #3127 §6.7: a publisher that POSTs and recovers the id from a `409` never
+rewrites the body, so a re-run silently keeps the first version. Harmless for a
+`v*` tag created once — and wrong the moment anything re-points. This one GETs
+first and PATCHes when the release exists, so it is correct either way rather
+than correct by luck (ThoughtSync #2182 is the same bug).
+"""
+
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import re
+import subprocess
+import sys
+import urllib.error
+import urllib.request
+
+API = "https://git.fabledsword.com/api/v1/repos/bvandeusen/FabledCurator"
+
+IMAGES = (
+    "git.fabledsword.com/bvandeusen/fabledcurator",
+    "git.fabledsword.com/bvandeusen/fabledcurator-ml",
+    "git.fabledsword.com/bvandeusen/fabledcurator-agent",
+)
+
+# Rule 148: `v` + the artifact's own version, zero-padded, no `.N`, no lookup.
+RULE_148 = re.compile(r"^v\d{4}\.\d{2}\.\d{2}\.\d{4}$")
+
+# Past this, the list has stopped being something anyone reads. It is reached
+# in exactly one situation — no previous tag is reachable, so the span is the
+# whole history — which happens on a genuine first release and on a tag cut
+# somewhere `main`'s tags cannot be seen from. Truncating says so; emitting
+# 1100 lines would bury the note explaining why there are 1100 of them.
+MAX_COMMITS = 200
+
+
+def git(*args: str) -> str:
+    return subprocess.run(
+        ["git", *args], capture_output=True, text=True, check=True
+    ).stdout.strip()
+
+
+def git_ok(*args: str) -> str | None:
+    """Run git, returning None instead of raising when it fails.
+
+    Used for the questions that legitimately have no answer — no previous tag,
+    no local `main` — where the absence is information rather than a fault.
+    """
+    try:
+        return git(*args)
+    except subprocess.CalledProcessError:
+        return None
+
+
+def previous_tag(ref: str, tag: str | None) -> str | None:
+    """The most recent `v*` tag reachable from `ref`, excluding `tag` itself.
+
+    `--exclude` rather than `<ref>^` so this is the same call whether or not
+    `ref` is the tag being released — and so it does not blow up on a root
+    commit that has no parent to walk to.
+    """
+    args = ["describe", "--tags", "--abbrev=0", "--match", "v*"]
+    if tag:
+        args += ["--exclude", tag]
+    return git_ok(*args, ref)
+
+
+def commits(previous: str | None, ref: str) -> list[str]:
+    """The subjects between the previous release and this one.
+
+    `--no-merges` because rule 153 merges `dev` into `main` with a plain merge
+    commit, so `main`'s first-parent view is a list of "Merge pull request #N"
+    and nothing else. The work is in the commits under those merges.
+    """
+    span = f"{previous}..{ref}" if previous else ref
+    out = git("log", "--no-merges", "--format=%s (%h)", span)
+    return [line for line in out.split("\n") if line.strip()]
+
+
+def truncate(log: list[str]) -> tuple[list[str], str | None]:
+    if len(log) <= MAX_COMMITS:
+        return log, None
+    return log[:MAX_COMMITS], (
+        f"{len(log)} commits in this span — more than a changelog is for. "
+        f"Listing the newest {MAX_COMMITS}. This usually means no previous "
+        f"`v*` tag was reachable from here."
+    )
+
+
+def render(tag: str, sha: str, previous: str | None, log: list[str], notes: list[str]) -> str:
+    short = sha[:7]
+    parts = []
+
+    if notes:
+        # Anything the derivation could not stand behind goes at the TOP, not
+        # in a footnote. A release that quietly names a build nobody can find
+        # is the failure this whole milestone is about.
+        parts.append("\n".join(f"> **Note:** {n}" for n in notes))
+
+    parts.append(
+        f"Built from `{short}`. The rollback unit is the immutable `:c-` tag "
+        f"(rule 145) — these three move together:\n\n```\n"
+        + "\n".join(f"{image}:c-{short}" for image in IMAGES)
+        + "\n```"
+    )
+
+    heading = f"## Changes since {previous}" if previous else "## Changes"
+    if log:
+        parts.append(heading + "\n\n" + "\n".join(f"- {line}" for line in log))
+    else:
+        parts.append(
+            heading
+            + "\n\n_No non-merge commits since the previous release. This tag "
+            "names the same source under a new name._"
+        )
+
+    span = f"{previous}..{tag}" if previous else tag
+    parts.append(
+        f"---\n\n_Derived at publish time from `git log --no-merges {span}`. "
+        f"Nothing here is hand-maintained._"
+    )
+    return "\n\n".join(parts)
+
+
+def cross_checks(tag: str, sha: str) -> list[str]:
+    """Everything the derivation knows that would make the release a lie.
+
+    Reported rather than enforced. The tag is already pushed by the time this
+    runs, so failing here would leave the operator with a tag and no release
+    and nothing but a red lane to explain it — while the release itself is
+    still the useful object. Say what is wrong, on the release, and publish.
+    """
+    notes = []
+
+    if not RULE_148.match(tag):
+        notes.append(
+            f"`{tag}` is not rule 148's `vYYYY.MM.DD.HHMM` shape. Published "
+            f"anyway — the old `v26.*` tags predate the rule."
+        )
+    else:
+        derived = artifact_version("web")
+        if derived and derived != tag[1:]:
+            notes.append(
+                f"This tag names `{tag[1:]}`, but the web image built from "
+                f"`{sha[:7]}` reports `{derived}`. The Settings footer will not "
+                f"match this release's name."
+            )
+
+    # `:c-<sha>` only exists if `main` built this commit. Checking costs one
+    # git call; claiming it without checking costs a rollback that 404s at the
+    # moment someone needs it.
+    main = git_ok("rev-parse", "--verify", "-q", "refs/remotes/origin/main")
+    if main is None:
+        notes.append(
+            "Could not resolve `origin/main` here, so the `:c-` tags above are "
+            "unverified — they exist only if `main` built this commit."
+        )
+    elif subprocess.run(
+        ["git", "merge-base", "--is-ancestor", sha, main], capture_output=True
+    ).returncode != 0:
+        notes.append(
+            f"`{sha[:7]}` is not on `main`, so no `:c-{sha[:7]}` images were "
+            f"ever published. The refs above will not pull."
+        )
+
+    return notes
+
+
+def artifact_version(artifact: str) -> str | None:
+    """What `artifacts.sh` derives for one artifact in the CURRENT checkout.
+
+    It takes no ref because `artifacts.sh` takes none — it walks history from
+    HEAD. That is right here only because a tag push checks out the tagged
+    commit; calling this after `--dry-run some-other-ref` would compare the
+    tag against the working tree, which is why the mismatch note below is
+    reported and not enforced.
+
+    Returns None rather than raising if the script is missing or unhappy: a
+    cross-check that cannot run should not take the release down with it.
+    """
+    root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+    try:
+        return subprocess.run(
+            ["sh", os.path.join(root, "scripts", "artifacts.sh"), "version", artifact],
+            capture_output=True, text=True, check=True, cwd=root,
+        ).stdout.strip()
+    except (subprocess.CalledProcessError, OSError):
+        return None
+
+
+def api(method: str, path: str, token: str, payload: dict | None = None) -> dict | None:
+    body = json.dumps(payload).encode() if payload is not None else None
+    req = urllib.request.Request(
+        API + path, data=body, method=method,
+        headers={
+            "Authorization": "token " + token,
+            "Content-Type": "application/json",
+        },
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=30) as resp:
+            return json.load(resp)
+    except urllib.error.HTTPError as exc:
+        if exc.code == 404:
+            return None
+        sys.exit(f"release: {method} {path} failed with HTTP {exc.code}: {exc.read()!r}")
+
+
+def publish(tag: str, name: str, body: str, token: str) -> None:
+    existing = api("GET", f"/releases/tags/{tag}", token)
+    if existing:
+        api("PATCH", f"/releases/{existing['id']}", token, {"name": name, "body": body})
+        print(f"release: updated existing release {existing['id']} for {tag}")
+    else:
+        api("POST", "/releases", token, {"tag_name": tag, "name": name, "body": body})
+        print(f"release: created release for {tag}")
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser(description=__doc__)
+    ap.add_argument(
+        "ref", nargs="?", default=None,
+        help="tag or commit to release. Defaults to GITHUB_REF's tag, else HEAD.",
+    )
+    ap.add_argument(
+        "--dry-run", action="store_true",
+        help="render the body to stdout and publish nothing. Needs no token, "
+             "so it also works as a preview before you decide to cut the tag.",
+    )
+    args = ap.parse_args()
+
+    github_ref = os.environ.get("GITHUB_REF", "")
+    if args.ref:
+        ref = args.ref
+    elif github_ref.startswith("refs/tags/"):
+        ref = github_ref[len("refs/tags/"):]
+    else:
+        ref = "HEAD"
+
+    # A tag only if git knows it as one — `HEAD` and a raw sha are refs to
+    # release FROM, never the name to exclude or to publish under.
+    tag = ref if git_ok("rev-parse", "--verify", "-q", f"refs/tags/{ref}") else None
+    sha = git("rev-parse", ref)
+    previous = previous_tag(ref, tag)
+
+    print(f"release: ref={ref} sha={sha[:12]} previous={previous or '<none>'}")
+
+    notes = cross_checks(tag, sha) if tag else [
+        f"Rendered for `{ref}`, which is not a tag. Nothing was published."
+    ]
+    for note in notes:
+        print(f"release: NOTE {note}")
+
+    log = commits(previous, ref)
+    print(f"release: {len(log)} non-merge commits in the span")
+    log, overflow = truncate(log)
+    if overflow:
+        print(f"release: NOTE {overflow}")
+        notes.append(overflow)
+    body = render(tag or ref, sha, previous, log, notes)
+
+    if args.dry_run or not tag:
+        print("--- body ---")
+        print(body)
+        return
+
+    token = os.environ.get("RELEASE_TOKEN") or os.environ.get("TOKEN")
+    if not token:
+        sys.exit("release: no RELEASE_TOKEN in the environment")
+    publish(tag, f"FabledCurator {tag[1:]}", body, token)
+
+
+if __name__ == "__main__":
+    main()

tests/test_api_admin.py

@@ -677,3 +677,34 @@ async def test_reset_content_tagging_apply_requires_confirm_token(client, db):
     )
     assert resp.status_code == 200
     assert (await resp.get_json())["deleted"] == 1
+
+
+@pytest.mark.asyncio
+async def test_trigger_reclaim_attachments_defaults_to_preview(client, monkeypatch):
+    """Unlike the other maintenance triggers, this one's apply unlinks FILES —
+    so an empty body must mean preview, not apply."""
+    from backend.app.tasks import admin as admin_tasks
+
+    calls = []
+    monkeypatch.setattr(
+        admin_tasks.reclaim_orphaned_attachments_task, "delay", _fake_delay(calls)
+    )
+    resp = await client.post("/api/admin/maintenance/reclaim-attachments", json={})
+    assert resp.status_code == 202
+    assert (await resp.get_json())["task_id"] == "task-xyz"
+    assert calls[0][1] == {"dry_run": True}
+
+
+@pytest.mark.asyncio
+async def test_trigger_reclaim_attachments_threads_apply(client, monkeypatch):
+    from backend.app.tasks import admin as admin_tasks
+
+    calls = []
+    monkeypatch.setattr(
+        admin_tasks.reclaim_orphaned_attachments_task, "delay", _fake_delay(calls)
+    )
+    resp = await client.post(
+        "/api/admin/maintenance/reclaim-attachments", json={"dry_run": False},
+    )
+    assert resp.status_code == 202
+    assert calls[0][1] == {"dry_run": False}

tests/test_api_attachments.py

@@ -1,6 +1,6 @@
 import pytest
 
-from backend.app.models import Artist, PostAttachment
+from backend.app.models import Artist, PostAttachment, attachment_download_url
 
 pytestmark = pytest.mark.integration
 
@@ -33,3 +33,16 @@ async def test_download_streams_with_disposition(client, db, tmp_path):
 async def test_download_404(client):
     resp = await client.get("/api/attachments/999999/download")
     assert resp.status_code == 404
+
+
+@pytest.mark.asyncio
+async def test_attachment_download_url_routes_to_the_download_endpoint(app):
+    """The two serializers no longer hand-format this path (#3072) — but a
+    single definition is only worth having if it still matches the route. Pin
+    it by MATCHING against the real URL map rather than comparing to a literal:
+    a string equality test would pass just as happily after someone renamed the
+    route, which is the exact drift the helper exists to prevent."""
+    built = attachment_download_url(4242)
+    endpoint, args = app.url_map.bind("localhost").match(built)
+    assert endpoint == "attachments.download"
+    assert args == {"attachment_id": 4242}

tests/test_api_extension.py

@@ -129,7 +129,6 @@ async def test_resolve_artist_name_dispatches_per_platform(db, monkeypatch):
     ("https://www.subscribestar.com/foobar",             "subscribestar", "foobar"),
     ("https://subscribestar.adult/foobar",               "subscribestar", "foobar"),
     ("https://www.hentai-foundry.com/user/Foo/profile",  "hentaifoundry", "Foo"),
-    ("https://www.deviantart.com/baz",                   "deviantart",    "baz"),
     ("https://www.pixiv.net/users/12345",                "pixiv",         "12345"),
     ("https://www.pixiv.net/en/users/12345",             "pixiv",         "12345"),
 ])
@@ -160,6 +159,23 @@ async def test_quick_add_source_unknown_url_400(client, ext_key):
     assert "known" in body
 
 
+@pytest.mark.asyncio
+async def test_quick_add_source_rejects_retired_deviantart(client, ext_key):
+    """#3069: a DeviantArt creator URL used to derive cleanly. Now that the
+    platform is retired, the extension's own gate should never offer the
+    button — but a stale content script on an un-updated browser still can,
+    so the backend has to refuse it rather than create an unusable source."""
+    resp = await client.post(
+        "/api/extension/quick-add-source",
+        json={"url": "https://www.deviantart.com/baz"},
+        headers={"X-Extension-Key": ext_key},
+    )
+    assert resp.status_code == 400
+    body = await resp.get_json()
+    assert body["error"] == "unknown_platform"
+    assert "deviantart" not in body["known"]
+
+
 @pytest.mark.asyncio
 async def test_quick_add_source_invalid_url_400(client, ext_key):
     resp = await client.post(
@@ -367,6 +383,53 @@ async def test_extension_manifest_returns_metadata_when_xpi_present(client, monk
     assert body["sha256"] == hashlib.sha256(b"fake-xpi-content").hexdigest()
 
 
+@pytest.mark.asyncio
+async def test_extension_manifest_reports_the_channel_the_image_declares(
+    client, monkeypatch, tmp_path
+):
+    """The channel travels BESIDE the version, never inside it.
+
+    Folding it in as a `1.0.3499884-dev` suffix is the failure this design
+    exists to avoid: the extension's comparator parses each dotted segment as
+    an integer, so a suffixed segment collapses to 0 and every dev build
+    compares equal to every other — "no update available" and "I cannot read
+    this version" stop being distinguishable. Asserting the two are separate
+    keys is what keeps a future edit from merging them.
+    """
+    (tmp_path / "fabledcurator-1.2.3.xpi").write_bytes(b"x")
+    monkeypatch.setattr(extension_module, "XPI_DIR", tmp_path)
+    monkeypatch.setattr(extension_module, "FC_CHANNEL", "dev")
+    resp = await client.get("/api/extension/manifest")
+    assert resp.status_code == 200
+    body = await resp.get_json()
+    assert body["channel"] == "dev"
+    assert body["version"] == "1.2.3"
+
+
+@pytest.mark.asyncio
+async def test_extension_manifest_omits_the_channel_when_the_image_declares_none(
+    client, monkeypatch, tmp_path
+):
+    """A local build, or any image from before the field existed.
+
+    The key must be ABSENT rather than present-and-empty: absence is the state
+    every consumer already handles (an older image conveys it by not having the
+    key at all), so a blank channel reuses that path instead of introducing a
+    second spelling of "unknown" for each reader to special-case.
+    """
+    (tmp_path / "fabledcurator-1.2.3.xpi").write_bytes(b"x")
+    monkeypatch.setattr(extension_module, "XPI_DIR", tmp_path)
+    monkeypatch.setattr(extension_module, "FC_CHANNEL", "")
+    resp = await client.get("/api/extension/manifest")
+    assert resp.status_code == 200
+    body = await resp.get_json()
+    assert "channel" not in body
+    # Everything else still answers — an image with no channel is not a
+    # degraded one, it just cannot say which channel it came from.
+    assert body["installed"] is True
+    assert body["latest_url"] == "/extension/fabledcurator-latest.xpi"
+
+
 # --- /extension/<filename> -----------------------------------------
 
 

tests/test_api_platforms.py

@@ -6,16 +6,17 @@ pytestmark = pytest.mark.integration
 
 
 @pytest.mark.asyncio
-async def test_platforms_returns_gs_six(client):
+async def test_platforms_returns_gs_five(client):
     resp = await client.get("/api/platforms")
     assert resp.status_code == 200
     body = await resp.get_json()
     platforms = body["platforms"]
     assert set(platforms.keys()) == {
         "patreon", "subscribestar", "hentaifoundry",
-        "discord", "pixiv", "deviantart",
+        "discord", "pixiv",
     }
     assert "fanbox" not in platforms
+    assert "deviantart" not in platforms  # retired at #3069
 
 
 @pytest.mark.asyncio

tests/test_artifact_identity.py

@@ -0,0 +1,166 @@
+"""The two values `artifacts.sh` derives, and what each of them promises.
+
+`revision` decides whether a build gets skipped; `version` is what an instance
+reports about itself and what a release tag is named after. Neither has a
+consumer that would notice it going subtly wrong.
+
+## revision
+
+Milestone 318 step 3: each image carries its revision as an `fc.revision`
+label, and build.yml reads that label back off the moving channel tag. Equal
+to the derived revision means the bytes this push would produce are already
+published, so the build is skipped.
+
+That makes the revision load-bearing in a way a version string is not — it is
+compared for equality against a value stamped into a real published artifact.
+Both ways of getting it wrong are silent:
+
+* **it does not identify the content** — a revision that moves when the source
+  did not (a HEAD-derived value, say) never matches, nothing is ever skipped,
+  and the mechanism quietly buys nothing while every lane stays green.
+* **it identifies the wrong content** — a revision that holds still when the
+  source DID change matches a stale label, the build is skipped, and the
+  channel serves bytes that do not correspond to the commit. This is the
+  dangerous direction, and it is what `test_artifact_paths.py` guards from the
+  other side by pinning the path sets.
+
+This module owns the narrower claim: whatever the path sets say, the revision
+is genuinely the commit those paths last changed in.
+
+## version
+
+`YYYY.MM.DD.HHMM`, zero-padded, UTC — one shape across the family (note #3127
+§1, rule 148), so the string this project emits is the same string its siblings
+emit. Two nearly-identical formats are more dangerous than two obviously
+different ones, and the only thing keeping them identical is a test.
+
+The identity-TAG tests this file used to hold are gone with the tag. There is
+no longer a `CHANNELLED` list to drift (the channel is which tag you inspect),
+and no `identity` subcommand to refuse an unqualified call.
+"""
+from __future__ import annotations
+
+import os
+import re
+import subprocess
+from pathlib import Path
+
+import pytest
+
+ROOT = Path(__file__).resolve().parent.parent
+
+ARTIFACTS = ("web", "ml", "agent", "extension")
+
+# 12 hex chars — the prefix build.yml stamps and compares.
+_REVISION = re.compile(r"^[0-9a-f]{12}$")
+
+# YYYY.MM.DD.HHMM, every segment zero-padded to its full width.
+_VERSION = re.compile(r"^\d{4}\.\d{2}\.\d{2}\.\d{4}$")
+
+
+# Everything here goes through artifacts.sh rather than importing a sibling
+# test module. That is the interface build.yml actually calls, so the tests
+# exercise the contract instead of a Python re-implementation of it — and no
+# other test module in this repo imports another, so a cross-test import would
+# be a new convention introduced for no gain.
+def artifacts(*args: str) -> str:
+    return subprocess.run(
+        ["sh", str(ROOT / "scripts" / "artifacts.sh"), *args],
+        capture_output=True, text=True, check=True, cwd=ROOT,
+    ).stdout
+
+
+def revision(artifact: str) -> str:
+    return artifacts("revision", artifact).strip()
+
+
+def newest_by_commit_time(artifact: str) -> str:
+    """The full SHA of the newest commit touching this artifact's shipped set.
+
+    Ordered by committer TIME, matching what artifacts.sh means. Deliberately
+    not `git log -1`: git's default order is reverse-chronological only within
+    topological constraints, so on a merged history it can name a different
+    commit than the newest timestamp does. They agree on this repo today, and
+    a test that silently depends on them continuing to agree would be a flake
+    waiting for the branch shape that separates them.
+    """
+    paths = artifacts("paths", artifact).split()
+    log = subprocess.run(
+        ["git", "log", "--format=%ct %H", "HEAD", "--", *paths],
+        capture_output=True, text=True, check=True, cwd=ROOT,
+    ).stdout.split("\n")
+    commits = [line.split(" ", 1) for line in log if line.strip()]
+    assert commits, (
+        f"no commit in this history touches the {artifact} path set — the "
+        f"derivation has nothing to stand on"
+    )
+    return max(commits, key=lambda c: int(c[0]))[1]
+
+
+@pytest.mark.parametrize("artifact", ARTIFACTS)
+def test_revision_is_the_commit_its_own_shipped_files_last_changed_in(artifact):
+    """The claim the whole skip decision rests on.
+
+    Computed from git rather than asked of the script, so it fails if the
+    derivation ever stops meaning what it says — switching to HEAD, to a build
+    clock, or to a path set it did not actually use. Each of those still
+    produces a plausible 12-hex value, which is why this is worth asserting
+    rather than eyeballing.
+    """
+    expected = newest_by_commit_time(artifact)
+    got = revision(artifact)
+    assert expected.startswith(got), (
+        f"{artifact} derives {got!r}, but the newest commit touching its "
+        f"shipped files is {expected[:12]!r}. The label stamped into the image "
+        f"would not identify its own content."
+    )
+
+
+@pytest.mark.parametrize("artifact", ARTIFACTS)
+def test_revision_is_a_legal_label_value_and_is_stable(artifact):
+    """It is stamped as a docker label and compared for string equality, so a
+    stray newline or a varying value breaks the comparison rather than the
+    build — the mechanism would simply stop hitting, silently."""
+    first = revision(artifact)
+    assert _REVISION.match(first), f"{first!r} is not a 12-char hex revision"
+    assert first == revision(artifact), "revision is not stable across calls"
+
+
+@pytest.mark.parametrize("artifact", ARTIFACTS)
+def test_version_is_zero_padded_calver(artifact):
+    """The family shape, pinned.
+
+    Padding was stripped until 2026-08-28 on the reasoning that each segment
+    should read as a plain integer — which never held, since comparison strips
+    leading zeros on parse anyway. What it did do was make this project emit
+    `2026.8.28.1432` while a sibling emitted `2026.08.28.1432`: two shapes one
+    character apart, which is the hard kind of difference to notice.
+
+    Also catches the midnight case. A `%H%M` of `0322` must survive as `0322`;
+    the old strip-leading-zeros helper turned it into `322`, silently changing
+    a four-digit field into three.
+    """
+    value = artifacts("version", artifact).strip()
+    assert _VERSION.match(value), (
+        f"{artifact} derives {value!r}, which is not zero-padded "
+        f"YYYY.MM.DD.HHMM. Note #3127 §1 and rule 148 both specify the padded "
+        f"form, and a release tag is this string with a `v` in front."
+    )
+
+
+@pytest.mark.parametrize("artifact", ARTIFACTS)
+def test_version_and_revision_describe_the_same_commit(artifact):
+    """They are derived independently and must not be able to disagree.
+
+    A build reports the version and skips on the revision, so a divergence
+    would mean an instance naming one commit while carrying another's bytes —
+    unfalsifiable from outside, since both values look perfectly well-formed.
+    """
+    sha = newest_by_commit_time(artifact)
+    stamped = subprocess.run(
+        ["git", "show", "-s", "--format=%cd", "--date=format-local:%Y.%m.%d.%H%M", sha],
+        capture_output=True, text=True, check=True, cwd=ROOT,
+        env={"TZ": "UTC", "PATH": os.environ.get("PATH", "")},
+    ).stdout.strip()
+    assert artifacts("version", artifact).strip() == stamped
+    assert sha.startswith(revision(artifact))

tests/test_artifact_paths.py

@@ -0,0 +1,199 @@
+"""`scripts/artifacts.sh` path sets must match what the Dockerfiles copy.
+
+Each published artifact's version derives from the newest commit touching its
+own shipped file set (milestone 313). The whole scheme rests on those sets
+being right, and both ways of being wrong are silent:
+
+* **too narrow** — a file ships but is not in the set, so the version does not
+  move when the content does, and a pin serves stale bytes. This is the
+  dangerous direction and the one this module exists for.
+* **too wide** — a file is in the set but never reaches the image, so the
+  artifact re-versions and rebuilds for a change it does not ship.
+
+Nothing else notices either. The version still derives, CI still goes green,
+and the mismatch only surfaces as "I pinned that build and got the wrong
+bytes". So the Dockerfiles are read here and compared against the declaration.
+
+The COPY list is not the whole answer, though. A file that DECIDES what an
+artifact reports belongs in its set even though it is copied into nothing —
+see DERIVERS below, where the same finding is recorded twice (#3156, #3202).
+"""
+from __future__ import annotations
+
+import re
+import subprocess
+from pathlib import Path
+
+import pytest
+
+ROOT = Path(__file__).resolve().parent.parent
+
+# artifact -> (dockerfile, build context relative to the repo root)
+ARTIFACTS = {
+    "web": ("Dockerfile", ""),
+    "ml": ("Dockerfile.ml", ""),
+    "agent": ("agent/Dockerfile", "agent"),
+}
+
+# COPY --from=<stage> copies from an earlier build stage, not from the build
+# context, so its source is not a repo path and cannot be in a path set.
+_COPY = re.compile(r"^\s*COPY\s+(?!--from=)(?P<args>.+)$", re.MULTILINE)
+
+
+def declared_paths(artifact: str) -> list[str]:
+    out = subprocess.run(
+        ["sh", str(ROOT / "scripts" / "artifacts.sh"), "paths", artifact],
+        capture_output=True, text=True, check=True, cwd=ROOT,
+    ).stdout
+    return out.split()
+
+
+def includes(artifact: str) -> list[str]:
+    """The set minus its `:(exclude)…` entries."""
+    return [p for p in declared_paths(artifact) if not p.startswith(":(exclude)")]
+
+
+def copy_sources(dockerfile: str, context: str) -> list[str]:
+    """Repo-relative sources of every context COPY in a Dockerfile."""
+    text = (ROOT / dockerfile).read_text()
+    sources: list[str] = []
+    for m in _COPY.finditer(text):
+        args = m.group("args").split()
+        # Last arg is the destination; everything before it is a source.
+        for src in args[:-1]:
+            # `frontend/package-lock.json*` — the glob is an optional-file
+            # idiom; the directory it sits in is what matters for coverage.
+            src = src.rstrip("*")
+            sources.append(f"{context}/{src}" if context else src)
+    return sources
+
+
+def covered_by(path: str, include: str) -> bool:
+    """`path` ships if an include names it or one of its ancestors."""
+    path = path.rstrip("/").lstrip("./")
+    include = include.rstrip("/")
+    return path == include or path.startswith(include + "/")
+
+
+@pytest.mark.parametrize("artifact", sorted(ARTIFACTS))
+def test_every_copied_path_is_in_the_artifacts_path_set(artifact):
+    """The too-narrow direction — the one that serves stale bytes on a pin."""
+    dockerfile, context = ARTIFACTS[artifact]
+    inc = includes(artifact)
+    for src in copy_sources(dockerfile, context):
+        assert any(covered_by(src, i) for i in inc), (
+            f"{dockerfile} copies {src!r} into the {artifact} image, but no "
+            f"include in scripts/artifacts.sh covers it. The {artifact} "
+            f"version will not move when that file changes, so a pinned build "
+            f"will serve stale bytes. Add it to the path set.\n"
+            f"  declared includes: {inc}"
+        )
+
+
+@pytest.mark.parametrize("artifact", sorted(ARTIFACTS))
+def test_the_dockerfile_itself_is_in_the_path_set(artifact):
+    """Changing a base image or a RUN changes the artifact as surely as
+    changing a source file, so each set must include its own Dockerfile."""
+    dockerfile, _ = ARTIFACTS[artifact]
+    assert any(covered_by(dockerfile, i) for i in includes(artifact)), (
+        f"{dockerfile} is not in the {artifact} path set — a base-image bump "
+        f"would not move the version."
+    )
+
+
+def test_the_web_image_versions_on_an_extension_change():
+    """The web image bundles the signed XPI, so the extension's packaged files
+    are part of what it ships. Miss this and `:latest` serves a NEW extension
+    under an unchanged web version — a pin that quietly disagrees with itself.
+    """
+    inc = includes("web")
+    assert any(covered_by("extension/background/background.js", i) for i in inc), (
+        "the web path set does not cover the extension's packaged files, but "
+        "build.yml downloads the signed XPI into frontend/public/extension/ "
+        "before the docker build"
+    )
+
+
+# A file that DECIDES an artifact's identity is part of what that artifact is
+# built from, even though it is copied into no image. Both entries here are the
+# same finding twice — #3156 for packaging.sh, #3202 for artifacts.sh — and
+# both were latent for the same reason: the version has no backstop.
+#
+# The revision does. Change how a REVISION is computed and the derived value
+# stops matching the label on the published image, which forces a rebuild; the
+# mechanism self-corrects because it compares against a string stamped into a
+# real artifact. Nothing compares a version to anything, so a version-only
+# derivation change is invisible unless the deriver is in the set.
+DERIVERS = [
+    # packaging.sh decides the version build.yml stamps into the packaged
+    # manifest.json, so changing it changes the shipped bytes. Left out,
+    # milestone 313 step 4 turns silent: the new version misses the
+    # ext-<version> cache and gets signed, while web's revision has not moved,
+    # so the reuse path republishes the old image and the fresh signature is
+    # orphaned. Guarded for web too, since web bundles what the extension makes.
+    ("extension/scripts/packaging.sh", ("extension", "web")),
+    # artifacts.sh decides the FC_VERSION baked into the web image (#3202).
+    # Web only, and deliberately: ml and agent ask this script for `revision`
+    # alone, so they are covered by the self-correcting path above, and the
+    # extension takes its version from packaging.sh. Milestone 318 step 5 is
+    # the worked instance — b3989d0 and 5771fd5 share revision fb2c4d5b80be
+    # while the version moved 2026.8.28.1249 -> 2026.08.28.1249. It was
+    # harmless only because FC_VERSION did not exist until one commit later.
+    ("scripts/artifacts.sh", ("web",)),
+]
+
+
+@pytest.mark.parametrize("path, artifacts", DERIVERS, ids=lambda v: str(v))
+def test_a_version_deriver_is_in_the_set_of_what_it_decides(path, artifacts):
+    for artifact in artifacts:
+        inc = includes(artifact)
+        excluded = [
+            p[len(":(exclude)"):] for p in declared_paths(artifact)
+            if p.startswith(":(exclude)")
+        ]
+        assert any(covered_by(path, i) for i in inc), (
+            f"{path} decides the version {artifact} reports, but is not in the "
+            f"{artifact} path set. A change to the derivation would leave the "
+            f"revision untouched, the build skipped, and the published image "
+            f"reporting the old version — with nothing to disagree with it."
+        )
+        assert not any(
+            covered_by(path, e.rstrip("*").rstrip("/")) for e in excluded
+        ), (
+            f"{path} is excluded from the {artifact} path set, so a change to "
+            f"how the version is derived would not move the version — and "
+            f"step 4 would reuse the image that carries the old one"
+        )
+
+
+@pytest.mark.parametrize(
+    "artifact, path",
+    [
+        # Deliberate exclusions — the too-wide direction. Each of these lives
+        # beside shipped code but never reaches an image, and including it
+        # would re-version the artifact for a change it does not carry.
+        #
+        # "Never reaches an image" is the test, not "is not source": DERIVERS
+        # above are also copied into nothing and DO belong in their sets,
+        # because they decide what the image reports. The line between the two
+        # lists is whether the file has a say in the artifact's identity.
+        ("agent", "agent/README.md"),
+        ("agent", "agent/ruff.toml"),
+        ("agent", "agent/docker-compose.yml"),
+        # vite builds from src/, index.html and public/; it never reads test/,
+        # so a frontend test change cannot reach `dist`.
+        ("web", "frontend/test/gallery.spec.js"),
+    ],
+)
+def test_files_that_never_reach_an_image_do_not_version_it(artifact, path):
+    paths = declared_paths(artifact)
+    excluded = [p[len(":(exclude)"):] for p in paths if p.startswith(":(exclude)")]
+    inc = [p for p in paths if not p.startswith(":(exclude)")]
+
+    included = any(covered_by(path, i) for i in inc)
+    exempted = any(covered_by(path, e.rstrip("*").rstrip("/")) for e in excluded)
+    assert not included or exempted, (
+        f"{path} is in the {artifact} path set but is not copied into the "
+        f"image — it would re-version and rebuild {artifact} for a change it "
+        f"does not ship."
+    )

tests/test_artist_directory_service.py

@@ -154,7 +154,7 @@ async def test_list_platform_filter_excludes_no_source(db):
 @pytest.mark.asyncio
 async def test_list_platform_filter_excludes_wrong_platform(db):
     a = await _seed_artist(db, "alice-wplat")
-    await _seed_source(db, a.id, "deviantart", "https://d/alice-wp")
+    await _seed_source(db, a.id, "discord", "https://d/alice-wp")
     await db.commit()
 
     page = await ArtistDirectoryService(db).list_artists(

tests/test_cleanup_service.py

@@ -5,6 +5,7 @@ side effects use tmp_path. Assertions on mutated rows use COLUMN
 SELECTS per reference_async_coredml_test_assertions — never
 re-read ORM attributes after a service mutates and re-fetches.
 """
+import os
 from datetime import UTC, datetime
 
 import pytest
@@ -68,6 +69,8 @@ def test_project_artist_cascade_returns_zeroes_for_empty_artist(db_sync):
     assert result["artist"]["slug"] == "empty"
     assert result["projected"] == {
         "images": 0,
+        "posts": 0,
+        "attachments": 0,
         "sources": 0,
         "thumbs": 0,
         "import_tasks": 0,
@@ -92,6 +95,87 @@ def test_project_artist_cascade_counts_images_and_thumbs_and_bytes(db_sync, tmp_
     assert result["projected"]["bytes_on_disk"] == 3500
 
 
+def test_project_artist_cascade_counts_posts_and_attachments(db_sync, tmp_path):
+    """The body-only artist: zero images, but posts and attachments that the
+    apply destroys. Previewing this as `images: 0` alone is what made a
+    content-only artist read as an empty one (#3067)."""
+    a = _make_artist(db_sync, slug="bodyonly")
+    p1 = Post(artist_id=a.id, external_post_id="bo-1", description="a body")
+    p2 = Post(artist_id=a.id, external_post_id="bo-2", description="another")
+    db_sync.add_all([p1, p2])
+    db_sync.flush()
+    db_sync.add(PostAttachment(
+        post_id=p1.id, artist_id=a.id, sha256="b0d1".ljust(64, "0"),
+        path="/store/b0d1/a.pdf", original_filename="a.pdf",
+        ext=".pdf", size_bytes=5,
+    ))
+    # artist_id NULL, reachable only through its post — the second arm of
+    # _artist_attachments_conditions.
+    db_sync.add(PostAttachment(
+        post_id=p2.id, artist_id=None, sha256="b0d2".ljust(64, "0"),
+        path="/store/b0d2/b.pdf", original_filename="b.pdf",
+        ext=".pdf", size_bytes=5,
+    ))
+    db_sync.commit()
+
+    projected = cleanup_service.project_artist_cascade(
+        db_sync, slug="bodyonly",
+    )["projected"]
+    assert projected["images"] == 0
+    assert projected["posts"] == 2
+    assert projected["attachments"] == 2
+
+
+def test_artist_cascade_preview_matches_apply(db_sync, tmp_path):
+    """Rule 93: the preview's numbers must be what the apply actually does.
+
+    Guards the drift directly rather than trusting that both halves happen to
+    use the same predicate — the preview and apply are asserted against each
+    other on one artist carrying all three row kinds.
+    """
+    a = _make_artist(db_sync, slug="parity")
+    for i in range(3):
+        f = tmp_path / f"par{i}.jpg"
+        f.write_bytes(b"x")
+        _make_image(
+            db_sync, artist=a, path=str(f), sha256=f"{i:064x}", size=10,
+        )
+    posts = [
+        Post(artist_id=a.id, external_post_id=f"par-{i}") for i in range(4)
+    ]
+    db_sync.add_all(posts)
+    db_sync.flush()
+    for i, p in enumerate(posts[:2]):
+        db_sync.add(PostAttachment(
+            post_id=p.id, artist_id=a.id, sha256=f"par{i}".ljust(64, "0"),
+            path=f"/store/par{i}/f.zip", original_filename="f.zip",
+            ext=".zip", size_bytes=9,
+        ))
+    db_sync.commit()
+    artist_id = a.id
+
+    projected = cleanup_service.project_artist_cascade(
+        db_sync, slug="parity",
+    )["projected"]
+    summary = cleanup_service.delete_artist_cascade(
+        db_sync, artist_id=artist_id, images_root=tmp_path,
+    )["summary"]
+
+    assert projected["images"] == summary["images_deleted"] == 3
+    assert projected["posts"] == summary["posts_deleted"] == 4
+    assert projected["attachments"] == summary["attachments_deleted"] == 2
+
+    # And the apply really did remove them — a matching pair of numbers is
+    # worth nothing if neither half touched the DB.
+    assert db_sync.execute(
+        select(func.count(Post.id)).where(Post.artist_id == artist_id)
+    ).scalar_one() == 0
+    assert db_sync.execute(
+        select(func.count(PostAttachment.id))
+        .where(PostAttachment.artist_id == artist_id)
+    ).scalar_one() == 0
+
+
 def test_project_artist_cascade_raises_on_unknown_slug(db_sync):
     with pytest.raises(LookupError):
         cleanup_service.project_artist_cascade(db_sync, slug="nope")
@@ -294,6 +378,88 @@ def test_delete_artist_cascade_idempotent_on_missing(db_sync, tmp_path):
     assert result["summary"]["images_deleted"] == 0
 
 
+def test_delete_artist_cascade_survives_same_sha_on_two_posts(db_sync, tmp_path):
+    """Same file attached to two of the artist's posts must not abort the delete.
+
+    Left to the cascade this raises: artist delete CASCADEs to Post, which SET
+    NULLs post_attachment.post_id, and `uq_post_attachment_null_post_sha`
+    (sha256 alone, WHERE post_id IS NULL) then rejects the second row. That's an
+    ordinary shape — _capture_attachment writes one row per post over one
+    sha-addressed blob by design. Also covers the NULL-artist_id arm of the
+    delete predicate: the second row has no artist_id, only a post that does.
+    """
+    a = _make_artist(db_sync, slug="casatt")
+    p1 = Post(artist_id=a.id, external_post_id="att-p1")
+    p2 = Post(artist_id=a.id, external_post_id="att-p2")
+    db_sync.add_all([p1, p2])
+    db_sync.flush()
+
+    shared_sha = "ca5a".ljust(64, "0")
+    db_sync.add(PostAttachment(
+        post_id=p1.id, artist_id=a.id, sha256=shared_sha,
+        path="/store/ca5a/bundle.zip", original_filename="bundle.zip",
+        ext=".zip", size_bytes=7,
+    ))
+    db_sync.add(PostAttachment(
+        post_id=p2.id, artist_id=None, sha256=shared_sha,
+        path="/store/ca5a/bundle.zip", original_filename="bundle.zip",
+        ext=".zip", size_bytes=7,
+    ))
+    db_sync.commit()
+    artist_id = a.id
+
+    result = cleanup_service.delete_artist_cascade(
+        db_sync, artist_id=artist_id, images_root=tmp_path,
+    )
+
+    assert result["summary"]["attachments_deleted"] == 2
+    assert db_sync.execute(
+        select(func.count(Artist.id)).where(Artist.id == artist_id)
+    ).scalar_one() == 0
+    assert db_sync.execute(
+        select(func.count(PostAttachment.id))
+        .where(PostAttachment.sha256 == shared_sha)
+    ).scalar_one() == 0
+
+
+def test_delete_artist_cascade_keeps_unrelated_null_post_attachment(
+    db_sync, tmp_path,
+):
+    """A filesystem-import row (post_id NULL) sharing the sha is the other way
+    this collides — and it must SURVIVE: it belongs to no artist, so the
+    cascade has no claim on it."""
+    a = _make_artist(db_sync, slug="casorph")
+    p = Post(artist_id=a.id, external_post_id="orph-p1")
+    db_sync.add(p)
+    db_sync.flush()
+
+    sha = "0rfa".ljust(64, "0")
+    standalone = PostAttachment(
+        post_id=None, artist_id=None, sha256=sha,
+        path="/store/0rfa/manual.pdf", original_filename="manual.pdf",
+        ext=".pdf", size_bytes=3,
+    )
+    db_sync.add(standalone)
+    db_sync.add(PostAttachment(
+        post_id=p.id, artist_id=a.id, sha256=sha,
+        path="/store/0rfa/manual.pdf", original_filename="manual.pdf",
+        ext=".pdf", size_bytes=3,
+    ))
+    db_sync.commit()
+    artist_id, standalone_id = a.id, standalone.id
+
+    result = cleanup_service.delete_artist_cascade(
+        db_sync, artist_id=artist_id, images_root=tmp_path,
+    )
+
+    assert result["summary"]["attachments_deleted"] == 1
+    surviving = db_sync.execute(
+        select(PostAttachment.id, PostAttachment.post_id)
+        .where(PostAttachment.sha256 == sha)
+    ).all()
+    assert surviving == [(standalone_id, None)]
+
+
 # --- delete_images --------------------------------------------------
 
 
@@ -916,3 +1082,176 @@ def test_reconcile_preserves_from_attachment_on_provenance_collision(db_sync, tm
         .where(ImageProvenance.image_record_id == img_id)
     ).all()
     assert rows == [(native_id, att_id)]
+
+
+# --- reclaim_orphaned_attachments -----------------------------------
+
+
+def _store_blob(root, sha, *, ext=".pdf", age_hours=48, data=b"blob"):
+    """Write a file into the sha-addressed attachment store, aged past the
+    min-age guard by default."""
+    d = root / "attachments" / sha[:3]
+    d.mkdir(parents=True, exist_ok=True)
+    p = d / f"{sha}{ext}"
+    p.write_bytes(data)
+    old = datetime.now(UTC).timestamp() - age_hours * 3600
+    os.utime(p, (old, old))
+    return p
+
+
+def _attachment(db_sync, *, sha, post=None, artist=None):
+    att = PostAttachment(
+        post_id=post.id if post else None,
+        artist_id=artist.id if artist else None,
+        sha256=sha, path=f"/store/{sha[:3]}/f.pdf",
+        original_filename="f.pdf", ext=".pdf", size_bytes=4,
+    )
+    db_sync.add(att)
+    db_sync.flush()
+    return att
+
+
+def test_reclaim_attachments_dry_run_projects_without_mutating(db_sync, tmp_path):
+    a = _make_artist(db_sync, slug="recl-dry")
+    p = Post(artist_id=a.id, external_post_id="rd-1")
+    db_sync.add(p)
+    db_sync.flush()
+    kept_sha, orphan_sha = "aa11".ljust(64, "0"), "bb22".ljust(64, "0")
+    _attachment(db_sync, sha=kept_sha, post=p, artist=a)
+    _attachment(db_sync, sha=orphan_sha)            # both FKs NULL → orphan
+    db_sync.commit()
+    kept_blob = _store_blob(tmp_path, kept_sha)
+    orphan_blob = _store_blob(tmp_path, orphan_sha)
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=True,
+    )
+    assert result["rows"] == 1
+    assert result["files"] == 1
+    assert result["bytes"] == orphan_blob.stat().st_size
+
+    # Nothing actually happened.
+    assert kept_blob.exists() and orphan_blob.exists()
+    assert db_sync.execute(
+        select(func.count(PostAttachment.id))
+    ).scalar_one() == 2
+
+
+def test_reclaim_attachments_apply_deletes_rows_and_unlinks_blobs(db_sync, tmp_path):
+    a = _make_artist(db_sync, slug="recl-apply")
+    p = Post(artist_id=a.id, external_post_id="ra-1")
+    db_sync.add(p)
+    db_sync.flush()
+    kept_sha, orphan_sha = "cc33".ljust(64, "0"), "dd44".ljust(64, "0")
+    _attachment(db_sync, sha=kept_sha, post=p, artist=a)
+    _attachment(db_sync, sha=orphan_sha)
+    db_sync.commit()
+    kept_blob = _store_blob(tmp_path, kept_sha)
+    orphan_blob = _store_blob(tmp_path, orphan_sha)
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    assert result["rows"] == 1
+    assert result["files"] == 1
+
+    assert kept_blob.exists()          # still referenced
+    assert not orphan_blob.exists()    # nothing points at it any more
+    surviving = db_sync.execute(select(PostAttachment.sha256)).scalars().all()
+    assert surviving == [kept_sha]
+
+
+def test_reclaim_attachments_preview_matches_apply(db_sync, tmp_path):
+    """Rule 93 — the dry-run's numbers are what the apply does. The projection
+    has to negate the orphan predicate to be honest about blobs the delete is
+    about to free, so this is the assertion that catches getting that backwards.
+    """
+    orphan_sha = "ee55".ljust(64, "0")
+    _attachment(db_sync, sha=orphan_sha)
+    db_sync.commit()
+    _store_blob(tmp_path, orphan_sha)
+
+    projected = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=True,
+    )
+    applied = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    for key in ("rows", "files", "bytes"):
+        assert projected[key] == applied[key], key
+    assert applied["rows"] == 1 and applied["files"] == 1
+
+
+def test_reclaim_attachments_keeps_shared_blob_while_any_row_remains(db_sync, tmp_path):
+    """The refcount case this whole sweep exists for: one sha-addressed blob
+    backs several rows, so deleting SOME of them must not free the file."""
+    a = _make_artist(db_sync, slug="recl-shared")
+    p = Post(artist_id=a.id, external_post_id="rs-1")
+    db_sync.add(p)
+    db_sync.flush()
+    sha = "ff66".ljust(64, "0")
+    _attachment(db_sync, sha=sha, post=p, artist=a)   # attributed — survives
+    _attachment(db_sync, sha=sha)                     # orphan — deleted
+    db_sync.commit()
+    blob = _store_blob(tmp_path, sha)
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    assert result["rows"] == 1     # the orphan row went
+    assert result["files"] == 0    # the blob did NOT
+    assert blob.exists()
+
+
+def test_reclaim_attachments_spares_filesystem_import_rows(db_sync, tmp_path):
+    """post_id NULL with an artist_id is the deliberate filesystem-import shape
+    (importer._capture_attachment), not an orphan — it is still attributed."""
+    a = _make_artist(db_sync, slug="recl-fsimport")
+    sha = "1177".ljust(64, "0")
+    _attachment(db_sync, sha=sha, artist=a)    # post NULL, artist set
+    db_sync.commit()
+    blob = _store_blob(tmp_path, sha)
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    assert result["rows"] == 0
+    assert result["files"] == 0
+    assert blob.exists()
+    assert db_sync.execute(
+        select(func.count(PostAttachment.id))
+    ).scalar_one() == 1
+
+
+def test_reclaim_attachments_skips_recent_and_staging_files(db_sync, tmp_path):
+    """A blob is written BEFORE its row commits, so a just-stored file with no
+    row is in-flight, not orphaned. `.partial` staging files belong to
+    cleanup_orphaned_temp_files and must be left alone either way."""
+    fresh_sha, staged_sha = "2288".ljust(64, "0"), "3399".ljust(64, "0")
+    fresh = _store_blob(tmp_path, fresh_sha, age_hours=0)
+    staged = _store_blob(tmp_path, staged_sha, ext=".pdf.partial")
+    db_sync.commit()
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    assert result["files"] == 0
+    assert result["skipped_recent"] == 1
+    assert fresh.exists() and staged.exists()
+
+
+def test_reclaim_attachments_ignores_non_sha_named_files(db_sync, tmp_path):
+    """The walk must only judge files it can identify as store blobs — anything
+    else under the root is none of its business."""
+    d = tmp_path / "attachments" / "zzz"
+    d.mkdir(parents=True)
+    stray = d / "notes.txt"
+    stray.write_text("not a blob")
+    old = datetime.now(UTC).timestamp() - 48 * 3600
+    os.utime(stray, (old, old))
+
+    result = cleanup_service.reclaim_orphaned_attachments(
+        db_sync, images_root=tmp_path, dry_run=False,
+    )
+    assert result["files"] == 0
+    assert stray.exists()

tests/test_download_backends.py

@@ -19,7 +19,7 @@ def test_native_platforms():
 def test_gallery_dl_platforms_are_not_native():
     # The platforms still served by gallery-dl must NOT route to the native
     # ingester — guards an accidental over-broad migration.
-    for platform in ("hentaifoundry", "discord", "deviantart"):
+    for platform in ("hentaifoundry", "discord"):
         assert uses_native_ingester(platform) is False
 
 

tests/test_health.py

@@ -9,3 +9,63 @@ async def test_health_returns_ok(client):
     assert response.status_code == 200
     body = await response.get_json()
     assert body == {"status": "ok"}
+
+
+# --- build identity (milestone 318 step 6) --------------------------------
+#
+# With no version image tags left, /api/health is the only place an instance
+# says which build it is. Both fields are OMITTED when unset rather than sent
+# empty: absence already means "cannot say" — an image predating the field
+# says exactly that by not having the key — and a second spelling would make
+# every reader special-case it (note #3127 §7).
+#
+# The test above is load-bearing for that: it asserts the body is EXACTLY
+# {"status": "ok"} when nothing is stamped, so a well-meaning `or ""` default
+# fails it.
+
+
+@pytest.mark.asyncio
+async def test_health_reports_the_build_it_is(client, monkeypatch):
+    from backend.app.api import health
+
+    monkeypatch.setattr(health, "FC_VERSION", "2026.08.28.1249")
+    monkeypatch.setattr(health, "FC_CHANNEL", "dev")
+
+    body = await (await client.get("/api/health")).get_json()
+    assert body == {
+        "status": "ok",
+        "version": "2026.08.28.1249",
+        "channel": "dev",
+    }
+
+
+@pytest.mark.asyncio
+async def test_health_keeps_the_channel_out_of_the_version(client, monkeypatch):
+    """Rule 149, asserted rather than assumed.
+
+    The tempting shortcut is a `-dev` suffix on the version. The extension's
+    comparator parses each dotted segment with `parseInt`, so a suffixed
+    segment reads as 0 and every dev build compares equal to every other —
+    #2993 exactly. Two separate keys cannot express that mistake.
+    """
+    from backend.app.api import health
+
+    monkeypatch.setattr(health, "FC_VERSION", "2026.08.28.1249")
+    monkeypatch.setattr(health, "FC_CHANNEL", "dev")
+
+    body = await (await client.get("/api/health")).get_json()
+    assert body["version"] == "2026.08.28.1249"
+    assert "dev" not in body["version"]
+
+
+@pytest.mark.asyncio
+async def test_health_omits_a_channel_it_cannot_name(client, monkeypatch):
+    """A locally-built image has a version but no channel. It must not gain an
+    empty one — the key's absence is the answer."""
+    from backend.app.api import health
+
+    monkeypatch.setattr(health, "FC_VERSION", "2026.08.28.1249")
+    monkeypatch.setattr(health, "FC_CHANNEL", "")
+
+    body = await (await client.get("/api/health")).get_json()
+    assert body == {"status": "ok", "version": "2026.08.28.1249"}

tests/test_image_tag_apply.py

@@ -0,0 +1,119 @@
+"""`insert_image_tags` — the shared bulk write behind the WIP-title backfill and
+both auto-apply sweeps (#3072).
+
+The sweeps previously issued one INSERT per applied tag from inside their
+per-image loop; they now hand this helper a chunk's worth of rows. That makes
+this function the single place three writers can be wrong at once, so it is
+tested directly rather than only through its callers.
+"""
+import pytest
+from sqlalchemy import select
+
+from backend.app.models import ImageRecord, Tag, TagKind
+from backend.app.models.tag import image_tag
+from backend.app.services.image_tag_apply import insert_image_tags
+
+pytestmark = pytest.mark.integration
+
+_N = 0
+
+
+def _img(db_sync):
+    global _N
+    _N += 1
+    rec = ImageRecord(
+        path=f"/images/ita/{_N}.jpg", sha256=f"a{_N:063d}",
+        size_bytes=1, mime="image/jpeg", width=1, height=1,
+        origin="imported_filesystem", integrity_status="unknown",
+    )
+    db_sync.add(rec)
+    db_sync.flush()
+    return rec
+
+
+def _tag(db_sync, name):
+    t = Tag(name=name, kind=TagKind.general)
+    db_sync.add(t)
+    db_sync.flush()
+    return t
+
+
+def _rows(db_sync, tag_id):
+    """(image_record_id, source) pairs currently carrying `tag_id`."""
+    return dict(db_sync.execute(
+        select(image_tag.c.image_record_id, image_tag.c.source)
+        .where(image_tag.c.tag_id == tag_id)
+    ).all())
+
+
+def _row(image_record_id, tag_id, source):
+    return {
+        "image_record_id": image_record_id, "tag_id": tag_id, "source": source,
+    }
+
+
+def test_inserts_every_row_in_one_call(db_sync):
+    t = _tag(db_sync, "ita-basic")
+    imgs = [_img(db_sync) for _ in range(3)]
+    insert_image_tags(
+        db_sync, [_row(i.id, t.id, "head_auto") for i in imgs]
+    )
+    assert _rows(db_sync, t.id) == {i.id: "head_auto" for i in imgs}
+
+
+def test_spans_several_tags_in_a_single_call(db_sync):
+    """The sweeps accumulate across ALL heads before flushing, so one call
+    carries rows for different tags. A per-tag implementation would drop all
+    but the first."""
+    t1, t2 = _tag(db_sync, "ita-multi-1"), _tag(db_sync, "ita-multi-2")
+    a, b = _img(db_sync), _img(db_sync)
+    insert_image_tags(db_sync, [
+        _row(a.id, t1.id, "head_auto"), _row(b.id, t1.id, "head_auto"),
+        _row(a.id, t2.id, "head_auto"),
+    ])
+    assert _rows(db_sync, t1.id) == {a.id: "head_auto", b.id: "head_auto"}
+    assert _rows(db_sync, t2.id) == {a.id: "head_auto"}
+
+
+def test_an_existing_tag_keeps_its_original_source(db_sync):
+    """THE assertion this helper exists for. A sweep re-running over an image
+    the operator tagged by hand must not restamp it as machine-applied — that
+    would silently poison the head's own training data, which excludes the
+    auto sources. ON CONFLICT DO NOTHING, never DO UPDATE."""
+    t = _tag(db_sync, "ita-manual")
+    rec = _img(db_sync)
+    insert_image_tags(db_sync, [_row(rec.id, t.id, "manual")])
+
+    insert_image_tags(db_sync, [_row(rec.id, t.id, "head_auto")])
+
+    assert _rows(db_sync, t.id) == {rec.id: "manual"}
+
+
+def test_a_repeat_within_one_call_does_not_raise(db_sync):
+    """Two heads can both fire on the same (image, tag) inside one chunk. The
+    conflict is resolved by the statement, not by the caller de-duplicating."""
+    t = _tag(db_sync, "ita-dupe")
+    rec = _img(db_sync)
+    insert_image_tags(db_sync, [
+        _row(rec.id, t.id, "head_auto"), _row(rec.id, t.id, "head_auto"),
+    ])
+    assert _rows(db_sync, t.id) == {rec.id: "head_auto"}
+
+
+def test_more_rows_than_the_chunk_size_all_land(db_sync):
+    """The chunk exists to stay under Postgres' 65535 bound-parameter ceiling.
+    Driven with a tiny chunk so the split is real rather than theoretical — at
+    the 5000 default no test would ever reach a second statement."""
+    t = _tag(db_sync, "ita-chunked")
+    imgs = [_img(db_sync) for _ in range(7)]
+    insert_image_tags(
+        db_sync, [_row(i.id, t.id, "head_auto") for i in imgs], chunk=2
+    )
+    assert _rows(db_sync, t.id) == {i.id: "head_auto" for i in imgs}
+
+
+def test_no_rows_is_a_no_op(db_sync):
+    """A dry-run chunk, or a chunk where every candidate was already skipped,
+    hands over an empty list. `.values([])` is a SQL error, so the empty case
+    must never reach the statement."""
+    insert_image_tags(db_sync, [])