Merge pull request 'Release: Feature 24 (Automated deploys)' (#211) from release/2026-10-12b into main

Reviewed-on: #211
Reviewed-by: The Reviewer <18+review_agent@code.cubecraftcreations.com>

AGENTS.md

@@ -84,7 +84,7 @@ behaviour, exercise it (hit the endpoint, render the page).
 | `internal/*` (planned) | `auth`, `db`, `store`, `secretbox`, `household`, `dinners`, `recipes`, `ai/llm`, `ai/harness`, `ai/skills`, `suggest`, `plan`, `shopping`, `ideas`, `gcal`, `notify`, `push`, `chat`, `jobs`, `scheduler`. |
 | `migrations/` | SQL migrations (planned). |
 | `web/` | React PWA source. |
-| `deploy/` | Deployment files (planned). |
+| `deploy/` | Deployment files: compose stack, systemd units, [deploy runbook](deploy/RUNBOOK.md) and the automated deploy (`auto-deploy.sh` with `systemd/mealquest-auto-deploy.*`). |
 | `docs/` | Project documentation, including [ARCHITECTURE.md](docs/ARCHITECTURE.md). |
 | `design/` | Requirements and UI handoff. **Read-only reference** — do not edit. |
 | `Dockerfile` | Multi-stage image build, at the repository root. |

deploy/RUNBOOK.md

@@ -0,0 +1,357 @@
+# MealQuest deploy runbook
+
+The one way to deploy MealQuest to production: the docker compose project
+`mealquest` on office-srvr `10.60.1.14` (see
+[ARCHITECTURE.md](../docs/ARCHITECTURE.md#topology)). Production runs `main`
+or a release tag; only `overseer` merges `dev` → `main`. Every deploy builds
+on the host, from the checked-out commit, with `make docker`.
+
+Host setup (clone, firewall, the systemd unit, Traefik, DNS, Litestream) and
+the database restore are in CTXD project `mealquest`, `RUNBOOKS.MD`. This
+file is the build, restart, verification and rollback procedure; where it
+uses a CTXD section, it names it.
+
+All commands run on `10.60.1.14` as `overseer`, in bash, from the checkout
+`/home/overseer/mealquest`. A block that ends with `FAILED` or a non-zero
+`$?` stops the deploy: do not go on to the next step.
+
+## How the image carries its version
+
+`make docker` runs `docker build --build-arg VERSION=<version>` and tags the
+image `mealquest:<version>`, where `<version>` is
+`git describe --tags --always` of the checkout (a release tag such as
+`v1.0.0`, otherwise the short SHA), with `-dirty` added when a tracked file
+is modified. The [Dockerfile](../Dockerfile) links it into the binary, and
+`/health` reports it:
+
+```json
+{"status":"ok","version":"v1.0.0"}
+```
+
+Never pass `--build-arg VERSION` by hand, and never build with
+`docker compose build` or `docker compose up --build`: compose passes no
+version, so that image reports `dev`.
+
+The compose service `mealquest` runs the image named `mealquest-mealquest`
+(compose's name for the image it would build; check with
+`docker compose -f deploy/compose.yaml config --images mealquest`). It
+already exists, so `docker compose up` uses it and never builds. A deploy
+points that name at the `make docker` image with `docker tag`. Every
+`mealquest:<version>` image is kept for rollback; do not prune them.
+
+## Prerequisites
+
+- `make`, `git`, `curl`, `python3` and Docker Engine with Compose on the
+  host (`make` is not in the base install: `sudo apt-get install make`).
+- `deploy/.env` (git-ignored, owner `overseer`, mode 0600) sets every
+  variable [`deploy/.env.example`](.env.example) sets, with
+  `MEALQUEST_BIND_ADDRESS=10.60.1.14`. It must not set the variables
+  `deploy/compose.yaml` fixes (`MEALQUEST_HTTP_ADDR`, `MEALQUEST_DATA_DIR`,
+  `MEALQUEST_SEARXNG_URL` and the `*_FILE` variables). Creating it: CTXD
+  `RUNBOOKS.MD` "Stack and first deploy" > "3. Environment file".
+- `deploy/secrets/` (git-ignored, mode 0700) holds `secret_key`,
+  `oidc_client_secret`, `google_client_secret`, `vapid_private_key` and
+  `litestream_key`, each owned by `10001:10001` with mode 0600, and
+  `searxng.env`, owned by `overseer` with mode 0600. Creating them: CTXD
+  `RUNBOOKS.MD` "Stack and first deploy" > "4. Secrets".
+- The compose file is the checkout's [`deploy/compose.yaml`](compose.yaml);
+  the systemd unit [`mealquest.service`](systemd/mealquest.service) starts
+  and stops it and is the only way the stack starts.
+- DNS and Traefik already route `https://mealquest.cubecraftlabs.com` to
+  `10.60.1.14:8090` (CTXD `RUNBOOKS.MD` "Reverse proxy (Traefik)", "DNS").
+
+Check `deploy/.env` and the secrets before every deploy (a new release may
+add a variable; the service refuses to boot without it). Both commands must
+print nothing:
+
+```sh
+cd /home/overseer/mealquest
+comm -23 <(grep -oE '^MEALQUEST_[A-Z_]+' deploy/.env.example | sort -u) \
+         <(grep -oE '^MEALQUEST_[A-Z_]+' deploy/.env | sort -u)   # variables missing from deploy/.env
+sudo find deploy/secrets -type f ! -perm 600                       # secrets not mode 0600
+```
+
+## Deploy
+
+### 1. Build
+
+Record the release production is running now, check out the new release
+and build its image. Nothing restarts.
+
+The rollback target is the version the running service reports on
+`/health`, read before the checkout changes, never the checkout's `HEAD`:
+after a failed or deferred deploy the checkout can be ahead of production.
+The block resolves that version to a commit and prints it as the rollback
+SHA. A running version of `dev` or `*-dirty` names no commit: the block
+fails, and the rollback target must be established by hand before going on.
+
+The checkout must be clean (the same test as the Makefile's `GIT_DIRTY`:
+no modified tracked file outside the generated `internal/frontend/dist`).
+The block computes the image tag exactly as `make docker` does, `-dirty`
+included, and checks that image exists after the build.
+
+```sh
+cd /home/overseer/mealquest && (
+set -euo pipefail
+fail() { echo "$*" >&2; exit 1; }
+built=
+trap 'st=$?
+if [ "$st" -eq 0 ] && [ -n "$built" ]; then echo "built: $built"
+else [ "$st" -ne 0 ] || st=1; echo "build FAILED (status $st)" >&2; fi
+exit "$st"' EXIT
+bind=$(grep -E '^MEALQUEST_BIND_ADDRESS=' deploy/.env | cut -d = -f 2-)
+[ -n "$bind" ] || fail "deploy/.env sets no MEALQUEST_BIND_ADDRESS"
+running=$(curl -fsS "http://$bind:8090/health" | python3 -c 'import json, sys; print(json.load(sys.stdin)["version"])') \
+  || fail "http://$bind:8090/health did not return a version: no running release to roll back to"
+case "$running" in
+  dev|*-dirty) fail "the running release reports $running, which names no commit" ;;
+esac
+git fetch --tags origin
+rollback=$(git rev-parse --verify --quiet "$running^{commit}") \
+  || fail "running version $running is not a commit in this repository"
+echo "rollback SHA: $rollback (running version $running)"   # RECORD it for rollback
+git checkout --detach "refs/remotes/origin/main^{commit}"   # or the release tag, e.g. v1.0.0
+dirty=$(git status --porcelain --untracked-files=no -- . ':(exclude)internal/frontend/dist')
+[ -z "$dirty" ] || fail "the checkout has modified tracked files; clean it first:
+$dirty"
+v=$(git describe --tags --always)${dirty:+-dirty}   # the Makefile's VERSION
+make docker
+docker image inspect "mealquest:$v" > /dev/null 2>&1 || fail "no image mealquest:$v: the build failed"
+built="mealquest:$v"
+)
+```
+
+### 2. Ship and restart
+
+1. **Pre-restart gates.** Run CTXD `RUNBOOKS.MD` "Routine deploy and
+   rollback" > "Deploy" step 2 (a verified pre-deploy restore point; record
+   it with the rollback SHA) and step 3 (boot-required configuration). The
+   restart applies migrations the previous release refuses, so without that
+   restore point there is no rollback. If either fails, do not restart:
+   production stays on the previous release.
+2. **Point the stack at the new image and restart.** `docker compose pull`
+   fetches the pinned `litestream` and `searxng` images before anything
+   stops; `mealquest` is never pulled. The restart applies the firewall,
+   then runs `down` and `up -d`, and returns once `mealquest` passes its
+   healthcheck (`/health` inside the container).
+
+```sh
+cd /home/overseer/mealquest && (
+set -euo pipefail
+fail() { echo "$*" >&2; exit 1; }
+up=
+trap 'st=$?
+if [ "$st" -eq 0 ] && [ -n "$up" ]; then echo "restarted on: $up"
+else [ "$st" -ne 0 ] || st=1; echo "restart FAILED (status $st)" >&2; fi
+exit "$st"' EXIT
+v=$(git describe --tags --always)
+docker compose -f deploy/compose.yaml pull litestream searxng
+docker tag "mealquest:$v" mealquest-mealquest
+sudo systemctl restart mealquest
+for s in mealquest litestream searxng; do
+  id=$(docker compose -f deploy/compose.yaml ps --status running --quiet "$s")
+  [ -n "$id" ] || fail "$s is not running"
+done
+id=$(docker compose -f deploy/compose.yaml ps --quiet mealquest)
+health=$(docker inspect -f '{{.State.Health.Status}}' "$id")
+[ "$health" = healthy ] || fail "mealquest is $health, not healthy"
+up="mealquest:$v"
+)
+```
+
+On failure read `journalctl -u mealquest` and
+`docker compose -f deploy/compose.yaml logs mealquest`: a configuration
+error names the variable. Fix it or roll back.
+
+### 3. Verify the deployed version and a real screen
+
+A `/health` 200 alone proves only that some build is up. This step checks,
+on the LAN address and through Traefik, that:
+
+- `/health` reports exactly the checkout's `git describe --tags --always`;
+- `/` and the Plan screen's route `/plan` serve the built PWA shell (the
+  React root and a hashed `/assets/` script, not the placeholder page an
+  unbuilt embed serves);
+- that script is served as JavaScript.
+
+```sh
+cd /home/overseer/mealquest && (
+set -euo pipefail
+fail() { echo "$*" >&2; exit 1; }
+want=
+trap 'st=$?
+if [ "$st" -eq 0 ] && [ -n "$want" ]; then echo "deployed and verified: $want ($(git rev-parse HEAD))"
+else [ "$st" -ne 0 ] || st=1; echo "verification FAILED (status $st): the deploy is not verified" >&2; fi
+exit "$st"' EXIT
+v=$(git describe --tags --always)
+for base in http://10.60.1.14:8090 https://mealquest.cubecraftlabs.com; do
+  got=$(curl -fsS "$base/health" | python3 -c 'import json, sys; print(json.load(sys.stdin)["version"])') \
+    || fail "$base/health did not return a version"
+  [ "$got" = "$v" ] || fail "$base/health reports version $got, the checkout is $v"
+  for path in / /plan; do
+    page=$(curl -fsS "$base$path") || fail "$base$path did not return 200"
+    grep -q '<div id="root"></div>' <<< "$page" || fail "$base$path is not the PWA shell"
+    js=$(grep -oE 'src="/assets/[^"]+\.js"' <<< "$page" | head -n 1 | cut -d '"' -f 2) || true
+    [ -n "$js" ] || fail "$base$path references no built /assets/ script"
+    type=$(curl -fsS -o /dev/null -w '%{content_type}' "$base$js") || fail "$base$js did not return 200"
+    case "$type" in
+      text/javascript*) ;;
+      *) fail "$base$js is $type, not JavaScript" ;;
+    esac
+  done
+  echo "$base: version $got, PWA shell and $js served"
+done
+want=$v
+)
+```
+
+- `deployed and verified: <version> (<SHA>)` and `$?` = `0` → open
+  `https://mealquest.cubecraftlabs.com` in a browser: sign in and the Home
+  screen renders with the household's data. Then record the deployed
+  version and SHA, the rollback SHA, the pre-deploy restore point and the
+  date in the release's Maestro task.
+- `verification FAILED`, any error or a non-zero `$?` → the deploy is not
+  done. A version mismatch means the stack runs a different image than the
+  checkout: repeat step 2, or roll back.
+
+## Rollback to a previous release
+
+Rolling back points `mealquest-mealquest` at the previous release's
+`mealquest:<version>` image again; that release's checkout keeps the
+compose file and the version check consistent with it.
+
+1. **Check whether the rollback crosses a migration** with CTXD
+   `RUNBOOKS.MD` "Routine deploy and rollback" > "Rollback to a previous
+   SHA" (its first block), run against the rollback SHA that deploy step 1
+   recorded. If it prints versions unknown to the rollback SHA, restore the
+   database to the pre-deploy restore point first (CTXD `RUNBOOKS.MD`
+   "Restore the database", with `-timestamp <restore point>`); changes
+   since then are lost, so tell the household.
+2. **Check out the rollback SHA and retag its image.** An image built before
+   this runbook was not tagged `mealquest:<version>`; `make docker` builds
+   it from the same commit.
+
+```sh
+cd /home/overseer/mealquest && (
+set -euo pipefail
+fail() { echo "$*" >&2; exit 1; }
+P=<rollback SHA>   # recorded by step 1, from the running release's /health
+back=
+trap 'st=$?
+if [ "$st" -eq 0 ] && [ -n "$back" ]; then echo "rolled back to: $back"
+else [ "$st" -ne 0 ] || st=1; echo "rollback FAILED (status $st)" >&2; fi
+exit "$st"' EXIT
+git checkout --detach "$P"
+v=$(git describe --tags --always)
+docker image inspect "mealquest:$v" > /dev/null 2>&1 || make docker
+docker image inspect "mealquest:$v" > /dev/null 2>&1 || fail "no image mealquest:$v"
+docker tag "mealquest:$v" mealquest-mealquest
+sudo systemctl restart mealquest
+back="mealquest:$v"
+)
+```
+
+3. **Run step 3 above** unchanged. The rollback is done only when it prints
+   `deployed and verified:` with the previous release's version.
+
+## Automated deploys
+
+[`auto-deploy.sh`](auto-deploy.sh), installed as
+`/usr/local/sbin/mealquest-auto-deploy`, runs this runbook's deploy unattended:
+step 1 (build), the prerequisite `deploy/.env` and secrets check, CTXD
+`RUNBOOKS.MD` "Deploy" steps 2 (pre-deploy restore point) and 3
+(boot-required configuration), step 2 (ship and restart) and step 3
+(verify). No check is skipped or weakened, and it stops at the first
+failure. Every failure before the restart leaves production on the previous
+release.
+
+### What runs, and when
+
+[`mealquest-auto-deploy.timer`](systemd/mealquest-auto-deploy.timer) starts
+[`mealquest-auto-deploy.service`](systemd/mealquest-auto-deploy.service)
+at every quarter hour (`OnCalendar=*-*-* *:0/15:00`), plus up to 5 minutes'
+random delay; a check missed while the host was down runs as soon as it is
+back (`Persistent=true`). The service runs the script as `overseer` from the
+checkout, and only while `mealquest.service` is active: a stack stopped on
+purpose (a database or media restore) is skipped, not redeployed. Output goes
+to the journal (`journalctl -u mealquest-auto-deploy`).
+
+Each run fetches `origin` and compares `origin/main` with the last
+`deployed=` SHA in the deploy record `/home/overseer/mealquest-auto-deploy.log`:
+
+| `origin/main` | Result |
+| --- | --- |
+| is the last deployed SHA | No-op: exit 0, no output. |
+| CI (Gitea commit status) `pending` | Exit 0, no output; the next run checks again. |
+| CI `failure`, `error`, `warning`, or no status yet | Error in the journal, exit 1, nothing deployed. |
+| the last record line is `FAILED` for this SHA | Error in the journal, exit 1, nothing deployed: a failed deploy is never retried by the timer. |
+| CI `success` | Deploy. |
+
+The CI status is read from
+`https://code.cubecraftlabs.com/api/v1/repos/Maniacs_Mansion/MealQuest/commits/<SHA>/status`
+(public; no credential). A commit merged moments before a run can briefly
+report no status, which fails that run; the next run reads it again.
+
+A deploy appends one line to the record:
+
+```text
+2026-10-12T15:04:05Z previous=<rollback SHA> deployed=<SHA> restore_point=2026-10-12T15:01:59Z
+```
+
+`previous` is the rollback SHA (from the running release's `/health`,
+never the checkout) and `restore_point` the verified pre-deploy restore
+point: record both in the release's Maestro task. A failed deploy appends
+`<date> FAILED target=<SHA> status=<exit status>` and logs
+`auto-deploy FAILED (status <n>)` to the journal, adding whether it failed
+before the restart (production untouched) or after it (read
+`journalctl -u mealquest`, then fix or roll back). Either way the checkout is
+left at the target commit; the running release on `/health` stays the
+rollback target.
+
+### Install on the host
+
+The script and units are installed copies: repeat this whenever a release
+changes `deploy/auto-deploy.sh` or the units. From the checkout, at a release
+that contains them:
+
+```sh
+cd /home/overseer/mealquest
+sudo install -m 0755 deploy/auto-deploy.sh /usr/local/sbin/mealquest-auto-deploy
+sudo install -m 0644 deploy/systemd/mealquest-auto-deploy.service \
+  deploy/systemd/mealquest-auto-deploy.timer /etc/systemd/system/
+sudo systemctl daemon-reload
+```
+
+The timer refuses to deploy until the record holds a deploy, so seed it with
+one manual deploy of the release production should run (below), then start
+the timer:
+
+```sh
+sudo /usr/local/sbin/mealquest-auto-deploy --ref refs/remotes/origin/main
+sudo systemctl enable --now mealquest-auto-deploy.timer
+systemctl list-timers mealquest-auto-deploy.timer
+```
+
+To stop automated deploys: `sudo systemctl disable --now mealquest-auto-deploy.timer`.
+
+### Manual override
+
+```sh
+sudo /usr/local/sbin/mealquest-auto-deploy --ref <ref>
+```
+
+deploys an explicit ref (a SHA, tag or `origin/<branch>`) through the same
+blocks, bypassing the CI and no-op gates: for a roll-forward fix, a re-run
+after a `FAILED` record, or seeding the record. Run as root, the script
+re-executes itself as `overseer`; it holds a lock, so it never runs alongside
+the timer's run. It prints `deployed and verified:` and exits 0 only when
+step 3 passed.
+
+### Rollback
+
+Rollback is not automated: follow "Rollback to a previous release" above
+and CTXD `RUNBOOKS.MD` "Routine deploy and rollback" > "Rollback to a
+previous SHA", using the `previous` SHA and `restore_point` from the record.
+The timer does not undo a manual rollback: `origin/main` is still the last
+recorded deploy, so its runs are no-ops until `main` moves, and a new `main`
+then deploys as usual. Disable the timer if the rollback must hold.

deploy/auto-deploy.sh

@@ -0,0 +1,251 @@
+#!/usr/bin/env bash
+# MealQuest automated deploy: deploy/RUNBOOK.md, run unattended.
+#
+#   mealquest-auto-deploy              deploy origin/main if it is new and CI passed
+#   mealquest-auto-deploy --ref <ref>  deploy <ref> now, bypassing the CI and no-op gates
+#
+# Installed as /usr/local/sbin/mealquest-auto-deploy and run as overseer by
+# mealquest-auto-deploy.service (deploy/systemd); install, semantics and the
+# manual override: deploy/RUNBOOK.md "Automated deploys". Run as root (sudo)
+# it re-executes itself as overseer.
+#
+# Without --ref: exit 0 with no output when origin/main is the last deployed
+# SHA in the deploy record or its CI is pending; exit 1, deploying nothing,
+# when its CI failed or reports no status, or when the last attempt to deploy
+# that SHA failed. Every deploy runs the runbook's blocks in order and stops
+# at the first failure; nothing before the restart touches production. A
+# verified deploy appends one line to the deploy record:
+#   <UTC date> previous=<SHA> deployed=<SHA> restore_point=<UTC date>
+# a failed one appends:
+#   <UTC date> FAILED target=<SHA> status=<exit status>
+set -euo pipefail
+
+CHECKOUT=/home/overseer/mealquest
+RECORD=/home/overseer/mealquest-auto-deploy.log
+LOCK=/home/overseer/mealquest-auto-deploy.lock
+STATUS_API=https://code.cubecraftlabs.com/api/v1/repos/Maniacs_Mansion/MealQuest/commits
+PUBLIC=https://mealquest.cubecraftlabs.com
+D=/var/lib/docker/volumes/mealquest-data/_data
+COMPOSE=(docker compose -f deploy/compose.yaml)
+
+fail() { echo "$*" >&2; exit 1; }
+
+if [ "$(id -u)" -eq 0 ]; then
+  exec sudo -u overseer -H -- "$0" "$@"
+fi
+[ "$(id -un)" = overseer ] || fail "run as overseer (or root), not $(id -un)"
+
+ref=
+case "$#" in
+  0) ;;
+  2) [ "$1" = --ref ] && [ -n "$2" ] || fail "usage: $0 [--ref <ref>]"; ref=$2 ;;
+  *) fail "usage: $0 [--ref <ref>]" ;;
+esac
+
+exec 9>> "$LOCK"
+flock -n 9 || fail "another mealquest-auto-deploy run holds $LOCK"
+
+cd "$CHECKOUT"
+git fetch --quiet --tags origin || fail "git fetch origin failed"
+
+if [ -n "$ref" ]; then
+  target=$(git rev-parse --verify --quiet "$ref^{commit}") || fail "$ref is not a commit in $CHECKOUT"
+else
+  # The full remote-tracking ref: a bare origin/main resolves a tag or local
+  # branch of that name first, which could deploy a commit not on main.
+  target=$(git rev-parse --verify --quiet "refs/remotes/origin/main^{commit}") \
+    || fail "refs/remotes/origin/main is not a commit in $CHECKOUT"
+  [ -f "$RECORD" ] || fail "no deploy record at $RECORD: seed it with one manual deploy (deploy/RUNBOOK.md \"Automated deploys\")"
+  last=$(grep -oE 'deployed=[0-9a-f]{40}' "$RECORD" | tail -n 1 | cut -d = -f 2) || true
+  [ -n "$last" ] || fail "$RECORD records no deploy: seed it with one manual deploy (deploy/RUNBOOK.md \"Automated deploys\")"
+  [ "$target" != "$last" ] || exit 0
+  if tail -n 1 "$RECORD" | grep -q " FAILED target=$target status="; then
+    fail "auto-deploy of $target already failed (see $RECORD and the journal): fix it, then deploy with --ref"
+  fi
+  state=$(curl -fsS --max-time 30 "$STATUS_API/$target/status" \
+    | python3 -c 'import json, sys; d = json.load(sys.stdin); print(d["state"] if d["total_count"] else "absent")') \
+    || fail "could not read the CI status of $target from $STATUS_API"
+  case "$state" in
+    success) ;;
+    pending) exit 0 ;;
+    *) fail "CI for origin/main $target is $state, not success: not deploying" ;;
+  esac
+fi
+
+echo "auto-deploy: deploying $target${ref:+ (--ref $ref)}"
+restarted=
+W=
+on_exit() {
+  local st=$?
+  [ -z "$W" ] || rm -rf "$W" || st=1
+  if [ "$st" -ne 0 ]; then
+    echo "$(date -u +%Y-%m-%dT%H:%M:%SZ) FAILED target=$target status=$st" >> "$RECORD"
+    if [ -n "$restarted" ]; then
+      echo "auto-deploy FAILED (status $st) after the restart: production may be down or unverified; read journalctl -u mealquest, then fix or roll back (deploy/RUNBOOK.md)" >&2
+    else
+      echo "auto-deploy FAILED (status $st): production untouched" >&2
+    fi
+  fi
+  exit "$st"
+}
+trap on_exit EXIT
+
+# deploy/RUNBOOK.md "Deploy" > "1. Build": record the rollback SHA from the
+# RUNNING release, check out the target, reject a dirty checkout, build.
+bind=$(grep -E '^MEALQUEST_BIND_ADDRESS=' deploy/.env | cut -d = -f 2-) || fail "deploy/.env sets no MEALQUEST_BIND_ADDRESS"
+[ -n "$bind" ] || fail "deploy/.env sets no MEALQUEST_BIND_ADDRESS"
+running=$(curl -fsS "http://$bind:8090/health" | python3 -c 'import json, sys; print(json.load(sys.stdin)["version"])') \
+  || fail "http://$bind:8090/health did not return a version: no running release to roll back to"
+case "$running" in
+  dev|*-dirty) fail "the running release reports $running, which names no commit" ;;
+esac
+rollback=$(git rev-parse --verify --quiet "$running^{commit}") \
+  || fail "running version $running is not a commit in this repository"
+echo "rollback SHA: $rollback (running version $running)"
+git checkout --quiet --detach "$target"
+dirty=$(git status --porcelain --untracked-files=no -- . ':(exclude)internal/frontend/dist')
+[ -z "$dirty" ] || fail "the checkout has modified tracked files; clean it first:
+$dirty"
+v=$(git describe --tags --always)${dirty:+-dirty}   # the Makefile's VERSION
+make docker
+docker image inspect "mealquest:$v" > /dev/null 2>&1 || fail "no image mealquest:$v: the build failed"
+echo "built: mealquest:$v"
+
+# deploy/RUNBOOK.md "Prerequisites": deploy/.env sets every variable of the
+# target's .env.example, and every secret is mode 0600.
+missing=$(comm -23 <(grep -oE '^MEALQUEST_[A-Z_]+' deploy/.env.example | sort -u) \
+                   <(grep -oE '^MEALQUEST_[A-Z_]+' deploy/.env | sort -u))
+[ -z "$missing" ] || fail "deploy/.env is missing: $missing"
+loose=$(sudo find deploy/secrets -type f ! -perm 600)
+[ -z "$loose" ] || fail "secrets not mode 0600: $loose"
+
+# CTXD RUNBOOKS.MD "Routine deploy and rollback" > "Deploy" step 2: a verified
+# pre-deploy restore point. The scripts are CTXD "Count check". They and the
+# check files live in a private mktemp directory (mode 0700), not at shared
+# /tmp paths, and python runs with -I: the snapshot runs as root, and a
+# shared directory on its import path would let any local user run code as
+# root.
+W=$(mktemp -d)
+cat > "$W/mqcount.py" <<'EOF'
+import sqlite3, sys, hashlib
+c = sqlite3.connect(sys.argv[1])
+tabs = [r[0] for r in c.execute("select name from sqlite_master where type=\"table\" and name not like \"sqlite_%\" and name not like \"_litestream%\" order by name")]
+for t in ("dinners", "week_plans", "schema_migrations"):
+    print(t, c.execute(f"select count(*) from {t}").fetchone()[0])
+h = hashlib.sha256()
+for t in tabs:
+    for row in c.execute(f"select * from \"{t}\" order by rowid"):
+        h.update(repr(row).encode())
+print("tables", len(tabs), "content-sha256", h.hexdigest()[:16])
+print("integrity", c.execute("pragma integrity_check").fetchone()[0])
+EOF
+cat > "$W/mqsnap.py" <<'EOF'
+import os, sqlite3, sys
+src, dst = sys.argv[1], sys.argv[2]
+if os.path.exists(dst):
+    sys.exit(f"{dst} already exists")
+s = sqlite3.connect(f"file:{src}?mode=ro", uri=True)
+d = sqlite3.connect(dst)
+s.backup(d)
+s.close()
+result = d.execute("pragma integrity_check").fetchall()
+d.close()
+if result != [("ok",)]:
+    sys.exit(f"snapshot integrity_check failed: {result}")
+print("snapshot", dst, "integrity ok")
+EOF
+point=$(
+set -eu
+point=
+trap 'st=$?
+rm -rf "$W" || st=1
+sudo sh -c "rm -f $D/predeploy-check.db*" || st=1
+if [ "$st" -eq 0 ] && [ -n "$point" ]; then echo "pre-deploy restore point: $point" >&2; echo "$point"
+else [ "$st" -ne 0 ] || st=1; echo "verification FAILED (status $st): do not restart" >&2; fi
+exit "$st"' EXIT
+exec 3>&1 1>&2   # block output to the journal; only the restore point to $point
+mkdir "$W/mqcheck"
+# a. Replication running, no log errors (grep: 0 match, 1 none, 2+ error), newest LTX listed
+id=$("${COMPOSE[@]}" ps --status running --quiet litestream) || fail "docker compose ps failed"
+[ -n "$id" ] || fail "litestream is not running"
+"${COMPOSE[@]}" logs --since 24h litestream > "$W/mqcheck/litestream.log" || fail "docker compose logs failed"
+gs=0
+grep -i error "$W/mqcheck/litestream.log" > "$W/mqcheck/litestream.errors" || gs=$?
+case "$gs" in
+  0) cat "$W/mqcheck/litestream.errors"; fail "litestream logged errors in the last 24h";;
+  1) ;;
+  *) fail "the litestream log scan failed (status $gs)";;
+esac
+"${COMPOSE[@]}" exec -T litestream \
+  litestream ltx -config /etc/litestream.yml -level all /data/mealquest.db
+# b. Restore replica to a scratch file (live files untouched), move off the volume, snapshot live
+"${COMPOSE[@]}" run --rm --no-deps litestream \
+  restore -config /etc/litestream.yml -o /data/predeploy-check.db /data/mealquest.db
+sudo sh -c "mv $D/predeploy-check.db* $W/mqcheck/"
+sudo python3 -I "$W/mqsnap.py" "$D/mealquest.db" "$W/mqcheck/live.db"
+sudo chown -R "$(id -un)" "$W/mqcheck"
+sudo ls -ln "$D"
+# mealquest.db-wal and -shm owned by 10001, never 0 (CTXD "Count check")
+root=$(sudo find "$D" -maxdepth 1 \( -name mealquest.db-wal -o -name mealquest.db-shm \) -uid 0)
+[ -z "$root" ] || fail "owned by root, not 10001 (CTXD RUNBOOKS.MD \"Count check\"): $root"
+# c. Restored replica: integrity ok and identical counts/content-sha256 to the live snapshot
+python3 -I "$W/mqcount.py" "$W/mqcheck/predeploy-check.db" > "$W/mqcheck/restored.txt"
+python3 -I "$W/mqcount.py" "$W/mqcheck/live.db" > "$W/mqcheck/live.txt"
+cat "$W/mqcheck/restored.txt"
+grep -qx 'integrity ok' "$W/mqcheck/restored.txt" || fail "the restored replica failed integrity_check"
+diff "$W/mqcheck/restored.txt" "$W/mqcheck/live.txt" || fail "the restored replica differs from the live snapshot"
+# d. Restore point: after every check, before the restart
+point=$(date -u +%Y-%m-%dT%H:%M:%SZ)
+exec 1>&3
+)
+
+# CTXD RUNBOOKS.MD "Routine deploy and rollback" > "Deploy" step 3:
+# boot-required configuration, secrets read the way the service reads them.
+[ -f deploy/.env ] || fail "missing deploy/.env"
+cid=$(grep -m 1 -E '^MEALQUEST_GOOGLE_CLIENT_ID=' deploy/.env | cut -d = -f 2- | tr -d '"' | tr -d "'") \
+  || fail "missing MEALQUEST_GOOGLE_CLIENT_ID in deploy/.env"
+[ -n "$cid" ] || fail "missing MEALQUEST_GOOGLE_CLIENT_ID in deploy/.env"
+[ -e deploy/secrets/google_client_secret ] || fail "missing deploy/secrets/google_client_secret"
+[ -s deploy/secrets/google_client_secret ] || fail "deploy/secrets/google_client_secret is empty"
+id=$("${COMPOSE[@]}" ps --status running --quiet mealquest) || fail "docker compose ps failed"
+[ -n "$id" ] || fail "the mealquest container is not running: start it (the check reads secrets the way the service does)"
+docker exec "$id" sh -c 'test -s /run/secrets/google_client_secret && test -r /run/secrets/google_client_secret' \
+  || fail "google_client_secret is not readable by the service user inside the container"
+echo "boot-required configuration present"
+
+# deploy/RUNBOOK.md "2. Ship and restart".
+"${COMPOSE[@]}" pull litestream searxng
+docker tag "mealquest:$v" mealquest-mealquest
+restarted=yes
+sudo systemctl restart mealquest
+for s in mealquest litestream searxng; do
+  id=$("${COMPOSE[@]}" ps --status running --quiet "$s")
+  [ -n "$id" ] || fail "$s is not running"
+done
+id=$("${COMPOSE[@]}" ps --quiet mealquest)
+health=$(docker inspect -f '{{.State.Health.Status}}' "$id")
+[ "$health" = healthy ] || fail "mealquest is $health, not healthy"
+echo "restarted on: mealquest:$v"
+
+# deploy/RUNBOOK.md "3. Verify the deployed version and a real screen".
+for base in "http://$bind:8090" "$PUBLIC"; do
+  got=$(curl -fsS "$base/health" | python3 -c 'import json, sys; print(json.load(sys.stdin)["version"])') \
+    || fail "$base/health did not return a version"
+  [ "$got" = "$v" ] || fail "$base/health reports version $got, the checkout is $v"
+  for path in / /plan; do
+    page=$(curl -fsS "$base$path") || fail "$base$path did not return 200"
+    grep -q '<div id="root"></div>' <<< "$page" || fail "$base$path is not the PWA shell"
+    js=$(grep -oE 'src="/assets/[^"]+\.js"' <<< "$page" | head -n 1 | cut -d '"' -f 2) || true
+    [ -n "$js" ] || fail "$base$path references no built /assets/ script"
+    type=$(curl -fsS -o /dev/null -w '%{content_type}' "$base$js") || fail "$base$js did not return 200"
+    case "$type" in
+      text/javascript*) ;;
+      *) fail "$base$js is $type, not JavaScript" ;;
+    esac
+  done
+  echo "$base: version $got, PWA shell and $js served"
+done
+
+deployed=$(git rev-parse HEAD)
+echo "$(date -u +%Y-%m-%dT%H:%M:%SZ) previous=$rollback deployed=$deployed restore_point=$point" >> "$RECORD"
+echo "deployed and verified: $v ($deployed), previous $rollback, restore point $point"

deploy/systemd/mealquest-auto-deploy.service

@@ -0,0 +1,23 @@
+# Deploys origin/main to production when it is new and its CI passed
+# (/usr/local/sbin/mealquest-auto-deploy, from deploy/auto-deploy.sh).
+# Triggered by mealquest-auto-deploy.timer; output goes to the journal.
+# Installing it and the manual override: deploy/RUNBOOK.md "Automated deploys".
+#
+# ExecCondition= skips the run while mealquest.service is not active, so a
+# stack stopped on purpose (a database or media restore) is never redeployed
+# under the operator. It is deliberately not Requires=/Requisite= on
+# mealquest.service: those would start the stack, or stop this run when the
+# deploy itself restarts mealquest.service. Runs as overseer, who owns the
+# checkout and uses sudo for the runbook's privileged steps.
+[Unit]
+Description=MealQuest automated deploy of main
+Wants=network-online.target
+After=network-online.target docker.service mealquest.service
+
+[Service]
+Type=oneshot
+User=overseer
+Group=overseer
+WorkingDirectory=/home/overseer/mealquest
+ExecCondition=/usr/bin/systemctl is-active --quiet mealquest.service
+ExecStart=/usr/local/sbin/mealquest-auto-deploy

deploy/systemd/mealquest-auto-deploy.timer

@@ -0,0 +1,13 @@
+# Checks for a new main at every quarter hour (plus up to 5 minutes' random
+# delay). Persistent= runs a check at once when the host comes back if a
+# scheduled one was missed while it was down.
+[Unit]
+Description=Periodic MealQuest automated deploy
+
+[Timer]
+OnCalendar=*-*-* *:0/15:00
+RandomizedDelaySec=5min
+Persistent=true
+
+[Install]
+WantedBy=timers.target

docs/ARCHITECTURE.md

@@ -259,6 +259,29 @@ automatic** (D-021).
   unarchived dinner still `pending` with no `ground_dinner` job for its revision
   queued or running is set `failed` with `interrupted`, so no dinner stays
   pending with nothing left to ground it.
+  Photo backfill. `POST /api/v1/settings/photo-backfill` (CSRF-checked)
+  enqueues a `backfill_dinner_photos` job and answers 202 `{id}` without
+  waiting for it; while one is queued or running it enqueues nothing and
+  answers with that job's id, so two passes never overlap, across restarts
+  too. The kind is refused by `POST /api/v1/jobs`. The job
+  grounds, one at a time with `Ground` on its name, every unarchived dinner
+  that is `failed` with no photo and no pasted `recipe_url` — grounded,
+  `manual` and `pending` dinners are never listed, and a pasted link is
+  never replaced by a search. Success sets grounding_status `grounded` as
+  the `ground_dinner` job does (the recipe keeps its source URL and site);
+  failure leaves the dinner `failed` with the new `grounding_error`, for
+  the fix-recipe flow. One dinner's failure never stops the others; a
+  dinner renamed, archived or grounded by hand meanwhile is left as it is.
+  Its result is `{dinners: [{dinner_id, outcome, reason}]}`, outcome
+  `grounded`, `failed` (with its reason) or `skipped`. A dinner whose write
+  fails is `failed` with `grounding_error` and left as it was; the others
+  still run, then the job fails with `internal_error`, keeping that result.
+  The responses returning a `dinners.Dinner` representation (the list; the
+  create, update, archive, favorite and grounding responses; the idea
+  responses) carry `photo_backfill_eligible`, and
+  `GET /api/v1/dinners/{id}` does not; it says whether the backfill grounds
+  that dinner, computed by the same condition it lists dinners by, so a
+  client offers a search only where one runs.
   `GET /api/v1/dinners/{id}/recipe-candidates`
   searches for the dinner's name and returns `{candidates: [{candidate_id,
   title, site_name, photo_url, minutes}]}`, at most 3, ranked as `Ground`

docs/STATES.md

@@ -1,50 +0,0 @@
-# Empty, loading and error states
-
-Every route the PWA serves — the handoff's routes plus the added Settings
-screens — with its empty, loading and error state. The handoff does not mock
-these states; it gives one pattern for each
-([handoff README](../design/MealQuest%20UI%20Designs/design_handoff_mealquest/README.md),
-"Empty / loading / error states"), and every screen uses the shared component
-for it:
-
-| State | Component | Pattern |
-| --- | --- | --- |
-| Loading | `Skeleton` ([web/src/ui/Skeleton.tsx](../web/src/ui/Skeleton.tsx)) | Pill-shaped skeletons in `--color-surface` with a 1.2 s opacity pulse (no pulse under reduced motion). |
-| Error | `ErrorBanner` ([web/src/ui/ErrorBanner.tsx](../web/src/ui/ErrorBanner.tsx)) | `--color-neutral-100` card, grape-700 (`--color-accent-700`) text, and a **Retry** button that reads or asks again. |
-| Empty | `EmptyState` ([web/src/ui/EmptyState.tsx](../web/src/ui/EmptyState.tsx)) | Bagel title, body text and an optional primary action — the Discover deck's "All caught up / New side quests arrive Sunday." with "Get more now" is the model. |
-
-[web/src/states.test.tsx](../web/src/states.test.tsx) renders the loading,
-empty and error variant of each route below against a stubbed API, asserting
-the copy and pressing Retry. Each screen's own test file covers its other
-states.
-
-The loading and error rows name the screen's main read. Smaller reads on a
-screen (a Quest log count, a calendar list, a webhook card) use the same
-`Skeleton` and `ErrorBanner` in place.
-
-## Handoff routes
-
-| Route | Screen | Empty | Loading | Error |
-| --- | --- | --- | --- | --- |
-| `/` | Home | `EmptyState` "No quest planned tonight" with **Plan this week**; rows that do not apply are hidden. | Skeletons for the date, Tonight card and status rows. | "Couldn't load tonight's quest. …" + Retry. |
-| `/onboarding` | Onboarding — Welcome step | None: the step reads nothing; it shows the welcome and the Regulars, Side quests and Quest log terms with **Let's start**. | None: nothing to read. | None: nothing to read. |
-| `/onboarding/regulars` | Onboarding — Regulars step | "Your regulars · 0" with the Common picks to add; **Next: notifications** stays disabled until enough Regulars are saved. | Skeleton under "Your regulars". | "Your Regulars could not load: …" + Retry. |
-| `/onboarding/notifications` | Onboarding — Notifications step | None: the step reads nothing. | While finishing, **Turn on notifications** and **Maybe later** are disabled. | "Could not finish setting up: …"; the buttons come back to try again, with **Add Regulars** when fewer than three Regulars are saved. |
-| `/plan` | Weekly plan (Draft board) | `EmptyState` "Nothing planned yet" with **Draft this week**. | Seven skeleton tiles ("Plan loading"; "Drafting this week" while the AI drafts). | "Couldn't load this week's plan. …" + Retry. |
-| `/plan/shopping` | Supply run | `EmptyState` "No shopping list yet" with **Make the shopping list**; an archived week with none shows "No shopping list". A list with no ingredients shows "Nothing to buy". | Skeleton title and dinner cards. | "Couldn't load the shopping list. …" + Retry. |
-| `/plan/shopping/archive` | Supply archive | `EmptyState` "Nothing archived yet" — "Each week's Supply run becomes a post here once it's made." | Skeleton posts. | "Couldn't load the Supply archive. …" + Retry. |
-| `/discover` | Side quests deck | `EmptyState` **"All caught up"** — "New side quests arrive Sunday." with **Get more now**. | Skeleton card stack (also while a batch finds more). | "Couldn't load your side quests. …" + Retry. |
-| `/quick` | Quick dinners | No ideas found: "Couldn't find Quick dinners. No ideas came back with a recipe to match." + Retry (an `ErrorBanner`, as no idea was found to show). | Six skeleton rows while the tab's ideas are found. | "Couldn't start looking for Quick dinners. …" + Retry. |
-| `/cook-from-pantry` | Cook from what I have | Before a search: no ingredient chips and **Find ideas** disabled. No ideas found: "Couldn't find ideas. No ideas came back with a recipe to match." + Retry. | Five skeleton cards ("Ideas loading"). | "Couldn't start looking for ideas. …" + Retry. |
-| `/dinner/:id` | Dinner detail | `EmptyState` "Dinner not found" — "This dinner isn't in our kitchen — it may have been removed." | Skeleton hero, title and lines. | "Couldn't load this dinner. …" + Retry. |
-| `/settings` | Settings home | Status lines "None on rotation yet", "Not connected", "Push off · Google Chat not connected". | A skeleton per status line. | "Settings could not load: …" + Retry, which reads every failed status again. |
-| `/settings/notifications` | Notifications & Google Chat | No webhook stored: the webhook form with **Save webhook**. | Skeletons for the channel matrix and the webhook card. | "Notification settings could not load: …" + Retry ("The Google Chat webhook could not load: …" for the webhook card). A failed webhook test says "Test message not sent: …" under the form. |
-| `/settings/ai` | Admin · AI assistant | No key stored: "No key stored" with **Add key**. | Skeleton. | "AI settings could not load: …" + Retry. |
-
-## Added Settings screens
-
-| Route | Screen | Empty | Loading | Error |
-| --- | --- | --- | --- | --- |
-| `/settings/household` | Household | None: the household's settings always exist, seeded on first start; the screen shows the seeded values. | Skeleton. | "Household settings could not load: …" + Retry. |
-| `/settings/regulars` | Regular dinners | "On rotation · 0" — "No Regulars yet. Add the dinners your household makes on repeat." | Skeleton. | "Your Regulars could not load: …" + Retry. |
-| `/settings/calendar` | Google Calendar | "Not connected" with **Connect Google Calendar**. | Skeleton (and one for the calendar picker). | "Google Calendar status could not load: …" + Retry ("Your calendars could not load: …" + Retry for the picker). |

internal/db/dinners.sql.go

@@ -214,6 +214,26 @@ func (q *Queries) FailPendingDinnerGrounding(ctx context.Context, arg FailPendin
 	return result.RowsAffected()
 }
 
+const getActivePhotoBackfillJob = `-- name: GetActivePhotoBackfillJob :one
+SELECT id
+FROM jobs
+WHERE kind = ?1
+	AND status IN ('queued', 'running')
+	AND json_extract(payload, '$.dinner_id') IS NULL
+ORDER BY id
+LIMIT 1
+`
+
+// The id of the full photo backfill job (no dinner_id) of kind queued or
+// running. StartPhotoBackfill enqueues one only when there is none, so
+// there is at most one. Returns sql.ErrNoRows when there is none.
+func (q *Queries) GetActivePhotoBackfillJob(ctx context.Context, kind string) (string, error) {
+	row := q.db.QueryRowContext(ctx, getActivePhotoBackfillJob, kind)
+	var id string
+	err := row.Scan(&id)
+	return id, err
+}
+
 const getDinner = `-- name: GetDinner :one
 SELECT id, name, normalized_name, status, origin, method, minutes_total, difficulty, cuisine, is_favorite, recipe_id, photo_id, grounding_status, updated_by, created_at, updated_at, recipe_url, grounding_error, grounding_revision FROM dinners
 WHERE id = ?
@@ -344,6 +364,58 @@ func (q *Queries) GroundDinner(ctx context.Context, arg GroundDinnerParams) (Din
 	return i, err
 }
 
+const groundFailedDinner = `-- name: GroundFailedDinner :execrows
+UPDATE dinners
+SET recipe_id = ?1,
+	photo_id = ?2,
+	method = COALESCE(method, ?3),
+	minutes_total = COALESCE(minutes_total, ?4),
+	difficulty = COALESCE(difficulty, ?5),
+	grounding_status = 'grounded',
+	grounding_error = NULL,
+	updated_at = ?6
+WHERE id = ?7
+	AND status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND grounding_revision = ?8
+`
+
+type GroundFailedDinnerParams struct {
+	RecipeID          sql.NullString
+	PhotoID           sql.NullString
+	Method            sql.NullString
+	MinutesTotal      sql.NullInt64
+	Difficulty        sql.NullString
+	UpdatedAt         string
+	ID                string
+	GroundingRevision int64
+}
+
+// Records a photo backfill's grounding of a dinner it listed, while the
+// dinner is still unarchived, failed, photoless, without a recipe_url and at
+// the grounding_revision it was listed at; a dinner grounded by hand,
+// renamed or archived meanwhile is left as it is (0 rows). Like
+// GroundPendingDinner it keeps a method, minutes_total or difficulty a
+// member already set.
+func (q *Queries) GroundFailedDinner(ctx context.Context, arg GroundFailedDinnerParams) (int64, error) {
+	result, err := q.db.ExecContext(ctx, groundFailedDinner,
+		arg.RecipeID,
+		arg.PhotoID,
+		arg.Method,
+		arg.MinutesTotal,
+		arg.Difficulty,
+		arg.UpdatedAt,
+		arg.ID,
+		arg.GroundingRevision,
+	)
+	if err != nil {
+		return 0, err
+	}
+	return result.RowsAffected()
+}
+
 const groundPendingDinner = `-- name: GroundPendingDinner :execrows
 UPDATE dinners
 SET recipe_id = ?1,
@@ -444,6 +516,64 @@ func (q *Queries) ListDinnersByStatus(ctx context.Context, status string) ([]Din
 	return items, nil
 }
 
+const listPhotoBackfillDinners = `-- name: ListPhotoBackfillDinners :many
+SELECT id, name, normalized_name, status, origin, method, minutes_total, difficulty, cuisine, is_favorite, recipe_id, photo_id, grounding_status, updated_by, created_at, updated_at, recipe_url, grounding_error, grounding_revision FROM dinners
+WHERE status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND id = COALESCE(?1, id)
+ORDER BY normalized_name, id
+`
+
+// The dinners a photo backfill grounds by name: unarchived, failed, with no
+// photo and no pasted recipe_url (the household's own link is never
+// replaced by a search), ordered by normalized name. A non-NULL id narrows
+// the list to that dinner. Grounded, manual and pending dinners are never
+// listed: they have a recipe, or a ground_dinner job grounding them.
+func (q *Queries) ListPhotoBackfillDinners(ctx context.Context, id sql.NullString) ([]Dinner, error) {
+	rows, err := q.db.QueryContext(ctx, listPhotoBackfillDinners, id)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var items []Dinner
+	for rows.Next() {
+		var i Dinner
+		if err := rows.Scan(
+			&i.ID,
+			&i.Name,
+			&i.NormalizedName,
+			&i.Status,
+			&i.Origin,
+			&i.Method,
+			&i.MinutesTotal,
+			&i.Difficulty,
+			&i.Cuisine,
+			&i.IsFavorite,
+			&i.RecipeID,
+			&i.PhotoID,
+			&i.GroundingStatus,
+			&i.UpdatedBy,
+			&i.CreatedAt,
+			&i.UpdatedAt,
+			&i.RecipeUrl,
+			&i.GroundingError,
+			&i.GroundingRevision,
+		); err != nil {
+			return nil, err
+		}
+		items = append(items, i)
+	}
+	if err := rows.Close(); err != nil {
+		return nil, err
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return items, nil
+}
+
 const listUnarchivedDinners = `-- name: ListUnarchivedDinners :many
 SELECT dinners.id, dinners.name, dinners.normalized_name, dinners.status, dinners.origin, dinners.method, dinners.minutes_total, dinners.difficulty, dinners.cuisine, dinners.is_favorite, dinners.recipe_id, dinners.photo_id, dinners.grounding_status, dinners.updated_by, dinners.created_at, dinners.updated_at, dinners.recipe_url, dinners.grounding_error, dinners.grounding_revision, recipes.source_url AS recipe_source_url
 FROM dinners
@@ -560,6 +690,40 @@ func (q *Queries) ListUnplannedQuestLogDinners(ctx context.Context, fromWeekStar
 	return items, nil
 }
 
+const refailDinnerGrounding = `-- name: RefailDinnerGrounding :execrows
+UPDATE dinners
+SET grounding_error = ?1,
+	updated_at = ?2
+WHERE id = ?3
+	AND status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND grounding_revision = ?4
+`
+
+type RefailDinnerGroundingParams struct {
+	GroundingError    sql.NullString
+	UpdatedAt         string
+	ID                string
+	GroundingRevision int64
+}
+
+// Records why a photo backfill could not ground a dinner it listed, under
+// the same conditions as GroundFailedDinner (0 rows otherwise).
+func (q *Queries) RefailDinnerGrounding(ctx context.Context, arg RefailDinnerGroundingParams) (int64, error) {
+	result, err := q.db.ExecContext(ctx, refailDinnerGrounding,
+		arg.GroundingError,
+		arg.UpdatedAt,
+		arg.ID,
+		arg.GroundingRevision,
+	)
+	if err != nil {
+		return 0, err
+	}
+	return result.RowsAffected()
+}
+
 const setDinnerFavorite = `-- name: SetDinnerFavorite :one
 UPDATE dinners
 SET is_favorite = ?,

internal/db/querier.go

@@ -117,6 +117,10 @@ type Querier interface {
 	// when there is none, so there is at most one. Returns sql.ErrNoRows when
 	// there is none.
 	GetActiveDraftJob(ctx context.Context, arg GetActiveDraftJobParams) (GetActiveDraftJobRow, error)
+	// The id of the full photo backfill job (no dinner_id) of kind queued or
+	// running. StartPhotoBackfill enqueues one only when there is none, so
+	// there is at most one. Returns sql.ErrNoRows when there is none.
+	GetActivePhotoBackfillJob(ctx context.Context, kind string) (string, error)
 	// The batch still pending or running, oldest first. Returns sql.ErrNoRows
 	// when there is none.
 	GetActiveSuggestionBatch(ctx context.Context) (SuggestionBatch, error)
@@ -175,6 +179,13 @@ type Querier interface {
 	// unknown id, returns sql.ErrNoRows. A method, minutes_total or difficulty
 	// no longer as observed is kept: a member edited it while it was grounding.
 	GroundDinner(ctx context.Context, arg GroundDinnerParams) (Dinner, error)
+	// Records a photo backfill's grounding of a dinner it listed, while the
+	// dinner is still unarchived, failed, photoless, without a recipe_url and at
+	// the grounding_revision it was listed at; a dinner grounded by hand,
+	// renamed or archived meanwhile is left as it is (0 rows). Like
+	// GroundPendingDinner it keeps a method, minutes_total or difficulty a
+	// member already set.
+	GroundFailedDinner(ctx context.Context, arg GroundFailedDinnerParams) (int64, error)
 	// Records the background grounding of an unarchived dinner still pending at
 	// the grounding_revision it was grounded for; a dinner grounded some other
 	// way, renamed or archived meanwhile is left as it is (0 rows). A method,
@@ -206,6 +217,12 @@ type Querier interface {
 	// The pending Rotation checks whose dinner is still tried, oldest plan
 	// date first, with their dinners.
 	ListPendingRotationReviews(ctx context.Context) ([]ListPendingRotationReviewsRow, error)
+	// The dinners a photo backfill grounds by name: unarchived, failed, with no
+	// photo and no pasted recipe_url (the household's own link is never
+	// replaced by a search), ordered by normalized name. A non-NULL id narrows
+	// the list to that dinner. Grounded, manual and pending dinners are never
+	// listed: they have a recipe, or a ground_dinner job grounding them.
+	ListPhotoBackfillDinners(ctx context.Context, id sql.NullString) ([]Dinner, error)
 	// The days of plan plan_id with their dinners, Monday first. A day with
 	// no dinner is left out.
 	ListPlanDays(ctx context.Context, planID string) ([]ListPlanDaysRow, error)
@@ -307,6 +324,9 @@ type Querier interface {
 	// Records a new draft over plan id: its status and note, raising its
 	// revision.
 	RedraftWeekPlan(ctx context.Context, arg RedraftWeekPlanParams) error
+	// Records why a photo backfill could not ground a dinner it listed, under
+	// the same conditions as GroundFailedDinner (0 rows otherwise).
+	RefailDinnerGrounding(ctx context.Context, arg RefailDinnerGroundingParams) (int64, error)
 	// Records a generation of list id from plan plan_id.
 	RegenerateShoppingList(ctx context.Context, arg RegenerateShoppingListParams) error
 	// Records a failed delivery and when to try again. The lease must still be

internal/dinners/backfill.go

@@ -0,0 +1,199 @@
+package dinners
+
+import (
+	"context"
+	"database/sql"
+	"encoding/json"
+	"errors"
+	"fmt"
+
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/db"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/store"
+)
+
+// KindBackfillPhotos is the job that grounds, by name, the dinners left
+// without a photo: those whose grounding failed and that have no pasted
+// recipe_url (ListPhotoBackfillDinners).
+const KindBackfillPhotos = "backfill_dinner_photos"
+
+// Photo backfill outcomes for one dinner.
+const (
+	BackfillGrounded = "grounded"
+	BackfillFailed   = "failed"
+	// BackfillSkipped: the dinner was grounded by hand, renamed or archived
+	// while the backfill grounded it, so the backfill left it as it is.
+	BackfillSkipped = "skipped"
+)
+
+// PhotoBackfillEligible reports whether a photo backfill grounds r: the
+// condition ListPhotoBackfillDinners lists dinners by, and GroundFailedDinner
+// and RefailDinnerGrounding write under, for one dinner. The dinner is
+// unarchived, failed, with no photo and no pasted recipe_url.
+func PhotoBackfillEligible(r db.Dinner) bool {
+	return r.Status != StatusArchived &&
+		r.GroundingStatus == GroundingFailed &&
+		!r.PhotoID.Valid &&
+		!r.RecipeUrl.Valid
+}
+
+// BackfillJob is the KindBackfillPhotos payload. DinnerID narrows the
+// backfill to that dinner; empty backfills every dinner without a photo.
+type BackfillJob struct {
+	DinnerID string `json:"dinner_id,omitempty"`
+}
+
+// BackfillOutcome is what a photo backfill did for one dinner. Reason is
+// set only for BackfillFailed: the Reason recorded as its grounding_error.
+type BackfillOutcome struct {
+	DinnerID string  `json:"dinner_id"`
+	Outcome  string  `json:"outcome"`
+	Reason   *string `json:"reason"`
+}
+
+// backfillResult is the KindBackfillPhotos result: the outcome of each
+// dinner the backfill tried, in the order it tried them.
+type backfillResult struct {
+	Dinners []BackfillOutcome `json:"dinners"`
+}
+
+func (r backfillResult) PublicJSON() ([]byte, error) { return json.Marshal(r) }
+
+// StartPhotoBackfill enqueues a KindBackfillPhotos job over every dinner
+// without a photo and returns its id, without waiting for it. While a full
+// backfill is queued or running it enqueues nothing and returns that job's
+// id instead, so two passes never ground the same dinner at once; the jobs
+// table holds the claim, so it holds across restarts too.
+func (s *Service) StartPhotoBackfill(ctx context.Context) (string, error) {
+	var id string
+	created := false
+	err := s.st.WithTx(ctx, func(tx *sql.Tx) error {
+		active, err := db.New(tx).GetActivePhotoBackfillJob(ctx, KindBackfillPhotos)
+		switch {
+		case errors.Is(err, sql.ErrNoRows):
+		case err != nil:
+			return fmt.Errorf("dinners: read active photo backfill: %w", err)
+		default:
+			id = active
+			return nil
+		}
+		id, err = s.backfillJob.EnqueueTx(ctx, tx, BackfillJob{})
+		if err != nil {
+			return fmt.Errorf("dinners: enqueue photo backfill: %w", err)
+		}
+		created = true
+		return nil
+	})
+	if err != nil {
+		return "", err
+	}
+	if created {
+		s.backfillJob.Wake()
+	}
+	return id, nil
+}
+
+// errBackfillNotRecorded fails a photo backfill that could not record the
+// outcome of one or more of its dinners.
+var errBackfillNotRecorded = errors.New("dinners: photo backfill: dinner outcomes not recorded")
+
+// runBackfillJob grounds each dinner ListPhotoBackfillDinners lists for p,
+// one at a time, with recipes.Grounder.Ground on its name: the first ranked
+// candidate whose photo caches and whose ingredients normalize grounds it,
+// with the recipe's source URL and site kept as any grounding keeps them.
+// Success records the recipe and photo with grounding_status grounded;
+// failure leaves the dinner failed with the new Reason as its
+// grounding_error, for the household to fix its recipe by hand. One
+// dinner's failure, its search's, model's or database write's, never stops
+// the others. A failure to list the dinners, or the job's context ending,
+// fails the job at once. A write that failed fails the job once every
+// dinner has been tried, as a grounding job's failed write does, with the
+// outcomes published as its result, so the job never reports done for a
+// dinner whose outcome is not stored.
+func (s *Service) runBackfillJob(ctx context.Context, p BackfillJob, publish func(backfillResult) error) (backfillResult, error) {
+	q := db.New(s.st.Reader)
+	rows, err := q.ListPhotoBackfillDinners(ctx, sql.NullString{String: p.DinnerID, Valid: p.DinnerID != ""})
+	if err != nil {
+		return backfillResult{}, fmt.Errorf("dinners: photo backfill: list: %w", err)
+	}
+	res := backfillResult{Dinners: make([]BackfillOutcome, 0, len(rows))}
+	grounded, unrecorded := false, false
+	for _, row := range rows {
+		out, recorded, err := s.backfillDinner(ctx, row)
+		if err != nil {
+			return backfillResult{}, err
+		}
+		grounded = grounded || out.Outcome == BackfillGrounded
+		unrecorded = unrecorded || !recorded
+		res.Dinners = append(res.Dinners, out)
+	}
+	if grounded {
+		s.wake()
+	}
+	if unrecorded {
+		if err := publish(res); err != nil {
+			return backfillResult{}, errors.Join(errBackfillNotRecorded, err)
+		}
+		return backfillResult{}, errBackfillNotRecorded
+	}
+	return res, nil
+}
+
+// backfillDinner grounds row, a dinner ListPhotoBackfillDinners listed, as
+// runBackfillJob describes, and reports whether its outcome was recorded.
+// Its error is only the job's context ending.
+func (s *Service) backfillDinner(ctx context.Context, row db.Dinner) (BackfillOutcome, bool, error) {
+	out := BackfillOutcome{DinnerID: row.ID}
+	w := db.New(s.st.Writer)
+	gr, err := s.grounder.Ground(ctx, row.Name, row.Name)
+	if ctxErr := ctx.Err(); ctxErr != nil {
+		return BackfillOutcome{}, false, ctxErr
+	}
+	if err != nil {
+		reason := reasonFor(err, false)
+		// Only the reason is logged: the error may quote a dependency's
+		// response, which jobs never store, return or log.
+		s.logger.InfoContext(ctx, "dinner photo backfill failed", "dinner_id", row.ID, "reason", reason)
+		n, err := w.RefailDinnerGrounding(ctx, db.RefailDinnerGroundingParams{
+			GroundingError:    sql.NullString{String: reason, Valid: true},
+			UpdatedAt:         store.Now().String(),
+			ID:                row.ID,
+			GroundingRevision: row.GroundingRevision,
+		})
+		out, recorded := s.backfillRecorded(ctx, out, BackfillFailed, reason, n, err)
+		return out, recorded, nil
+	}
+	n, err := w.GroundFailedDinner(ctx, db.GroundFailedDinnerParams{
+		RecipeID:          sql.NullString{String: gr.RecipeID, Valid: true},
+		PhotoID:           sql.NullString{String: gr.PhotoID, Valid: true},
+		Method:            sql.NullString{String: gr.Method, Valid: true},
+		MinutesTotal:      nullInt(gr.MinutesTotal),
+		Difficulty:        sql.NullString{String: gr.Difficulty, Valid: true},
+		UpdatedAt:         store.Now().String(),
+		ID:                row.ID,
+		GroundingRevision: row.GroundingRevision,
+	})
+	out, recorded := s.backfillRecorded(ctx, out, BackfillGrounded, "", n, err)
+	return out, recorded, nil
+}
+
+// backfillRecorded is out once the backfill recorded outcome (with reason,
+// for BackfillFailed) on the dinner, writing n rows with err: a dinner that
+// changed meanwhile (no row) is BackfillSkipped, and one whose write failed
+// is BackfillFailed with ReasonGroundingError, the dinner left as it was,
+// and not recorded.
+func (s *Service) backfillRecorded(ctx context.Context, out BackfillOutcome, outcome, reason string, n int64, err error) (BackfillOutcome, bool) {
+	switch {
+	case err != nil:
+		// The error is a database error, never a dependency's text.
+		s.logger.ErrorContext(ctx, "dinner photo backfill not recorded", "dinner_id", out.DinnerID, "error", err)
+		out.Outcome, reason = BackfillFailed, ReasonGroundingError
+	case n == 0:
+		out.Outcome, reason = BackfillSkipped, ""
+	default:
+		out.Outcome = outcome
+	}
+	if reason != "" {
+		out.Reason = &reason
+	}
+	return out, err == nil
+}

internal/dinners/backfill_test.go

@@ -0,0 +1,425 @@
+package dinners_test
+
+import (
+	"database/sql"
+	"encoding/json"
+	"errors"
+	"slices"
+	"sync"
+	"testing"
+	"time"
+
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/ai/harness"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/db"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/dinners"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/jobs"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/recipes"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/recipes/recipestest"
+	"code.cubecraftlabs.com/Maniacs_Mansion/Meal-Quest/internal/store"
+)
+
+// failedDinner stores a Regular named name whose grounding failed with
+// no_recipe_found, without a photo or recipe_url, and returns its id.
+func (e *env) failedDinner(t *testing.T, name string) string {
+	t.Helper()
+	q := db.New(e.st.Writer)
+	now := store.Now().String()
+	row, err := q.CreateDinner(t.Context(), db.CreateDinnerParams{
+		ID:              store.NewID(),
+		Name:            name,
+		NormalizedName:  harness.Normalize(name),
+		Status:          dinners.StatusRegular,
+		Origin:          "regular",
+		GroundingStatus: dinners.GroundingPending,
+		UpdatedBy:       sql.NullString{String: member, Valid: true},
+		CreatedAt:       now,
+		UpdatedAt:       now,
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if _, err := q.FailPendingDinnerGrounding(t.Context(), db.FailPendingDinnerGroundingParams{
+		GroundingError:    sql.NullString{String: dinners.ReasonNoRecipe, Valid: true},
+		UpdatedAt:         now,
+		ID:                row.ID,
+		GroundingRevision: row.GroundingRevision,
+	}); err != nil {
+		t.Fatal(err)
+	}
+	return row.ID
+}
+
+func outcome(id, out string, reason *string) dinners.BackfillOutcome {
+	return dinners.BackfillOutcome{DinnerID: id, Outcome: out, Reason: reason}
+}
+
+func checkOutcomes(t *testing.T, got, want []dinners.BackfillOutcome) {
+	t.Helper()
+	gj, _ := json.Marshal(got)
+	wj, _ := json.Marshal(want)
+	if string(gj) != string(wj) {
+		t.Fatalf("outcomes = %s, want %s", gj, wj)
+	}
+}
+
+func TestBackfillGroundsFirstCandidateWithAPhoto(t *testing.T) {
+	t.Parallel()
+	e := newUnstartedEnv(t)
+	// curryA ranks first but its photo is gone; curryB's photo caches.
+	delete(e.images.JPEGs, "https://a.example/curry.jpg")
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 25))
+	id := e.failedDinner(t, "Chicken curry")
+
+	got, err := e.svc.RunBackfillJob(t.Context(), dinners.BackfillJob{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	checkOutcomes(t, got, []dinners.BackfillOutcome{outcome(id, dinners.BackfillGrounded, nil)})
+
+	row := e.dinner(t, id)
+	if row.GroundingStatus != dinners.GroundingGrounded || !row.PhotoID.Valid || !row.RecipeID.Valid || row.GroundingError.Valid {
+		t.Fatalf("dinner = %+v, want grounded with a photo and no error", row)
+	}
+	rec, err := db.New(e.st.Reader).GetRecipe(t.Context(), row.RecipeID.String)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if rec.SourceUrl != curryB || rec.SiteName.String != "B Kitchen" || rec.PhotoID != row.PhotoID {
+		t.Errorf("recipe = %s from %q photo %s, want %s from B Kitchen photo %s", rec.SourceUrl, rec.SiteName.String, rec.PhotoID.String, curryB, row.PhotoID.String)
+	}
+	if got := e.search.Queries(); !slices.Equal(got, []string{"Chicken curry"}) {
+		t.Errorf("searched %q, want the dinner's name", got)
+	}
+}
+
+func TestBackfillReusesCachedRecipeAndPhoto(t *testing.T) {
+	t.Parallel()
+	e := newUnstartedEnv(t)
+	e.search.Results["Curry chicken"] = []recipes.Candidate{{URL: curryA}}
+	// One reply: the second dinner must reuse the recipe the first stored.
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	first := e.failedDinner(t, "Chicken curry")
+	second := e.failedDinner(t, "Curry chicken")
+
+	got, err := e.svc.RunBackfillJob(t.Context(), dinners.BackfillJob{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	checkOutcomes(t, got, []dinners.BackfillOutcome{
+		outcome(first, dinners.BackfillGrounded, nil),
+		outcome(second, dinners.BackfillGrounded, nil),
+	})
+	a, b := e.dinner(t, first), e.dinner(t, second)
+	if a.RecipeID != b.RecipeID || a.PhotoID != b.PhotoID || !a.PhotoID.Valid {
+		t.Errorf("dinners grounded in recipe %s photo %s and recipe %s photo %s, want the same cached ones",
+			a.RecipeID.String, a.PhotoID.String, b.RecipeID.String, b.PhotoID.String)
+	}
+	if n := len(e.model.Requests()); n != 1 {
+		t.Errorf("normalize_recipe ran %d times, want 1", n)
+	}
+	if n := slices.Index(e.pages.Fetched(), curryA); n < 0 || slices.Contains(e.pages.Fetched()[n+1:], curryA) {
+		t.Errorf("fetched %q, want %s fetched once", e.pages.Fetched(), curryA)
+	}
+}
+
+func TestBackfillIsolatesEachDinnersFailure(t *testing.T) {
+	t.Parallel()
+	e := newUnstartedEnv(t)
+	e.search.Fail = map[string]error{"Broken search": errors.New("search provider down")}
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	// Listed by normalized name: both failures come before the curry.
+	broken := e.failedDinner(t, "Broken search")
+	curry := e.failedDinner(t, "Chicken curry")
+	absent := e.failedDinner(t, "Abalone surprise")
+
+	got, err := e.svc.RunBackfillJob(t.Context(), dinners.BackfillJob{})
+	if err != nil {
+		t.Fatal(err)
+	}
+	checkOutcomes(t, got, []dinners.BackfillOutcome{
+		outcome(absent, dinners.BackfillFailed, ptr(dinners.ReasonNoRecipe)),
+		outcome(broken, dinners.BackfillFailed, ptr(dinners.ReasonSearchFailed)),
+		outcome(curry, dinners.BackfillGrounded, nil),
+	})
+	for id, reason := range map[string]string{absent: dinners.ReasonNoRecipe, broken: dinners.ReasonSearchFailed} {
+		row := e.dinner(t, id)
+		if row.GroundingStatus != dinners.GroundingFailed || row.GroundingError.String != reason || row.PhotoID.Valid {
+			t.Errorf("dinner %s = %+v, want failed with %s and no photo", row.Name, row, reason)
+		}
+	}
+	if row := e.dinner(t, curry); row.GroundingStatus != dinners.GroundingGrounded || !row.PhotoID.Valid {
+		t.Errorf("curry = %+v, want grounded with a photo", row)
+	}
+}
+
+func TestBackfillSkipsDinnersNotMissingAPhoto(t *testing.T) {
+	t.Parallel()
+	e := newEnv(t)
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	c, err := e.svc.CreateRegular(t.Context(), dinners.NewRegular{Name: ptr("Chicken curry")}, member)
+	if err != nil {
+		t.Fatal(err)
+	}
+	e.waitJob(t, *c.GroundingJobID)
+	if row := e.dinner(t, c.ID); row.GroundingStatus != dinners.GroundingGrounded {
+		t.Fatalf("dinner = %+v, want grounded", row)
+	}
+	// A failed dinner with a pasted link keeps that link for a fix by hand.
+	linked, err := e.svc.CreateRegular(t.Context(), dinners.NewRegular{Name: ptr("Mystery pie"), RecipeURL: ptr("https://d.example/pie")}, member)
+	if err != nil {
+		t.Fatal(err)
+	}
+	e.waitJob(t, *linked.GroundingJobID)
+	if row := e.dinner(t, linked.ID); row.GroundingStatus != dinners.GroundingFailed {
+		t.Fatalf("linked dinner = %+v, want failed", row)
+	}
+	archived := e.failedDinner(t, "Old stew")
+	if _, err := e.svc.Archive(t.Context(), archived, member); err != nil {
+		t.Fatal(err)
+	}
+	searched := len(e.search.Queries())
+
+	id, err := e.svc.StartPhotoBackfill(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	job := e.waitJob(t, id)
+	if job.Kind != dinners.KindBackfillPhotos || string(job.Result) != `{"dinners":[]}` {
+		t.Fatalf("job = %+v (result %s), want a backfill with no dinners", job, job.Result)
+	}
+	if got := e.search.Queries()[searched:]; len(got) != 0 {
+		t.Errorf("backfill searched %q, want nothing", got)
+	}
+}
+
+// TestPhotoBackfillEligibleIsWhatTheBackfillLists holds the API's
+// photo_backfill_eligible to ListPhotoBackfillDinners for a dinner the
+// backfill grounds and for each kind it never does.
+func TestPhotoBackfillEligibleIsWhatTheBackfillLists(t *testing.T) {
+	t.Parallel()
+	e := newEnv(t)
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	grounded, err := e.svc.CreateRegular(t.Context(), dinners.NewRegular{Name: ptr("Chicken curry")}, member)
+	if err != nil {
+		t.Fatal(err)
+	}
+	e.waitJob(t, *grounded.GroundingJobID)
+	linked, err := e.svc.CreateRegular(t.Context(), dinners.NewRegular{Name: ptr("Mystery pie"), RecipeURL: ptr("https://d.example/pie")}, member)
+	if err != nil {
+		t.Fatal(err)
+	}
+	e.waitJob(t, *linked.GroundingJobID)
+	failed := e.failedDinner(t, "Abalone surprise")
+	archived, err := e.svc.Archive(t.Context(), e.failedDinner(t, "Old stew"), member)
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	shown := map[string]dinners.Dinner{archived.ID: archived}
+	regulars, err := e.svc.List(t.Context(), dinners.StatusRegular)
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, d := range regulars {
+		shown[d.ID] = d
+	}
+	q := db.New(e.st.Reader)
+	for id, want := range map[string]bool{grounded.ID: false, linked.ID: false, failed: true, archived.ID: false} {
+		listed, err := q.ListPhotoBackfillDinners(t.Context(), sql.NullString{String: id, Valid: true})
+		if err != nil {
+			t.Fatal(err)
+		}
+		d, ok := shown[id]
+		if !ok {
+			t.Fatalf("dinner %s not shown", id)
+		}
+		if got := len(listed) == 1; got != want || d.PhotoBackfillEligible != want {
+			t.Errorf("%s: listed %t, photo_backfill_eligible %t, want both %t", d.Name, got, d.PhotoBackfillEligible, want)
+		}
+	}
+}
+
+func TestBackfillOfOneDinnerGroundsOnlyIt(t *testing.T) {
+	t.Parallel()
+	e := newUnstartedEnv(t)
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	curry := e.failedDinner(t, "Chicken curry")
+	other := e.failedDinner(t, "Abalone surprise")
+
+	got, err := e.svc.RunBackfillJob(t.Context(), dinners.BackfillJob{DinnerID: curry})
+	if err != nil {
+		t.Fatal(err)
+	}
+	checkOutcomes(t, got, []dinners.BackfillOutcome{outcome(curry, dinners.BackfillGrounded, nil)})
+	if row := e.dinner(t, other); row.GroundingStatus != dinners.GroundingFailed || row.PhotoID.Valid {
+		t.Errorf("other dinner = %+v, want left failed", row)
+	}
+	if got := e.search.Queries(); !slices.Equal(got, []string{"Chicken curry"}) {
+		t.Errorf("searched %q, want only the dinner's name", got)
+	}
+}
+
+func TestBackfillLeavesDinnerChangedWhileGrounding(t *testing.T) {
+	t.Parallel()
+	e := newUnstartedEnv(t)
+	id := e.failedDinner(t, "Chicken curry")
+	gate := make(chan struct{})
+	e.search.Gate = gate
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+
+	done := make(chan []dinners.BackfillOutcome)
+	go func() {
+		got, err := e.svc.RunBackfillJob(t.Context(), dinners.BackfillJob{DinnerID: id})
+		if err != nil {
+			t.Error(err)
+		}
+		done <- got
+	}()
+	// The backfill listed the dinner and waits at its search.
+	for len(e.search.Queries()) == 0 {
+		time.Sleep(time.Millisecond)
+	}
+	if _, err := e.svc.Update(t.Context(), id, dinners.Patch{Name: ptr("Chicken korma")}, member); err != nil {
+		t.Fatal(err)
+	}
+	close(gate)
+	checkOutcomes(t, <-done, []dinners.BackfillOutcome{outcome(id, dinners.BackfillSkipped, nil)})
+	if row := e.dinner(t, id); row.GroundingStatus != dinners.GroundingFailed || row.PhotoID.Valid {
+		t.Errorf("dinner = %+v, want left failed without a photo", row)
+	}
+}
+
+// triggerAtOnce calls svc.StartPhotoBackfill from n goroutines released
+// together and returns the job ids they got.
+func triggerAtOnce(t *testing.T, svc *dinners.Service, n int) []string {
+	t.Helper()
+	ids := make([]string, n)
+	errs := make([]error, n)
+	release := make(chan struct{})
+	var wg sync.WaitGroup
+	for i := range n {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			<-release
+			ids[i], errs[i] = svc.StartPhotoBackfill(t.Context())
+		}()
+	}
+	close(release)
+	wg.Wait()
+	if err := errors.Join(errs...); err != nil {
+		t.Fatal(err)
+	}
+	return ids
+}
+
+// backfillJobs counts the KindBackfillPhotos jobs in st.
+func backfillJobs(t *testing.T, st *store.Store) int {
+	t.Helper()
+	var n int
+	if err := st.Reader.QueryRowContext(t.Context(), `SELECT count(*) FROM jobs WHERE kind = ?`, dinners.KindBackfillPhotos).Scan(&n); err != nil {
+		t.Fatal(err)
+	}
+	return n
+}
+
+func TestStartPhotoBackfillIsSingleFlight(t *testing.T) {
+	t.Parallel()
+	const triggers = 8
+	dataDir := t.TempDir()
+	st := openStore(t, dataDir)
+	e := newUnstartedEnvOn(t, st, dataDir)
+	e.failedDinner(t, "Chicken curry")
+	gate := make(chan struct{})
+	e.search.Gate = gate
+
+	// Queued: simultaneous triggers all get the one pass they enqueued.
+	ids := triggerAtOnce(t, e.svc, triggers)
+	first := ids[0]
+	for _, id := range ids {
+		if id != first {
+			t.Fatalf("simultaneous triggers = %q, want one job id", ids)
+		}
+	}
+	if n := backfillJobs(t, st); n != 1 {
+		t.Fatalf("%d backfill jobs after simultaneous triggers, want 1", n)
+	}
+
+	// After a restart the queued pass is still the one: the claim is the
+	// stored job, not the process's memory.
+	if err := st.Close(); err != nil {
+		t.Fatal(err)
+	}
+	st = openStore(t, dataDir)
+	restarted := newUnstartedEnvOn(t, st, dataDir)
+	restarted.search.Gate = gate
+	restarted.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	if again, err := restarted.svc.StartPhotoBackfill(t.Context()); err != nil || again != first {
+		t.Fatalf("trigger after restart = %q, %v; want the queued %q", again, err, first)
+	}
+
+	// Running: the pass waits at its search, and simultaneous triggers all
+	// return it.
+	restarted.start(t)
+	for len(restarted.search.Queries()) == 0 {
+		time.Sleep(time.Millisecond)
+	}
+	for _, id := range triggerAtOnce(t, restarted.svc, triggers) {
+		if id != first {
+			t.Fatalf("trigger while running = %q, want the running %q", id, first)
+		}
+	}
+	if n := backfillJobs(t, st); n != 1 {
+		t.Fatalf("%d backfill jobs, want 1", n)
+	}
+
+	// Finished: the next trigger starts a new pass.
+	close(gate)
+	if job := restarted.waitJob(t, first); job.Status != jobs.StatusSucceeded {
+		t.Fatalf("job = %+v (result %s), want succeeded", job, job.Result)
+	}
+	next, err := restarted.svc.StartPhotoBackfill(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	if next == first {
+		t.Fatalf("trigger after the pass finished = %q, want a new job", next)
+	}
+	restarted.waitJob(t, next)
+}
+
+func TestBackfillFailsJobWhenAWriteFails(t *testing.T) {
+	t.Parallel()
+	e := newEnv(t)
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	absent := e.failedDinner(t, "Abalone surprise")
+	curry := e.failedDinner(t, "Chicken curry")
+	if _, err := e.st.Writer.ExecContext(t.Context(), `CREATE TRIGGER refuse_grounded BEFORE UPDATE OF grounding_status ON dinners
+		WHEN NEW.id = '`+curry+`' AND NEW.grounding_status = 'grounded'
+		BEGIN SELECT RAISE(ABORT, 'grounded refused'); END`); err != nil {
+		t.Fatal(err)
+	}
+
+	id, err := e.svc.StartPhotoBackfill(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	job := e.waitJob(t, id)
+	if job.Status != jobs.StatusFailed || job.Error == nil || *job.Error != jobs.CodeInternal {
+		t.Fatalf("job = %+v, want failed with %s", job, jobs.CodeInternal)
+	}
+	var got struct {
+		Dinners []dinners.BackfillOutcome `json:"dinners"`
+	}
+	if err := json.Unmarshal(job.Result, &got); err != nil {
+		t.Fatalf("result %s: %v", job.Result, err)
+	}
+	// The other dinner was still tried and recorded.
+	checkOutcomes(t, got.Dinners, []dinners.BackfillOutcome{
+		outcome(absent, dinners.BackfillFailed, ptr(dinners.ReasonNoRecipe)),
+		outcome(curry, dinners.BackfillFailed, ptr(dinners.ReasonGroundingError)),
+	})
+	if row := e.dinner(t, curry); row.GroundingStatus != dinners.GroundingFailed || row.PhotoID.Valid {
+		t.Errorf("curry = %+v, want left failed without a photo", row)
+	}
+}

internal/dinners/dinners.go

@@ -79,6 +79,11 @@ type Dinner struct {
 	UpdatedBy      *string `json:"updated_by"`
 	CreatedAt      string  `json:"created_at"`
 	UpdatedAt      string  `json:"updated_at"`
+
+	// PhotoBackfillEligible is whether a photo backfill grounds the dinner
+	// (PhotoBackfillEligible): false for one with a pasted recipe link,
+	// which a search never replaces.
+	PhotoBackfillEligible bool `json:"photo_backfill_eligible"`
 }
 
 // NewRegular is a Regular to add: its Name, which the household always
@@ -138,6 +143,8 @@ type Service struct {
 	grounder  *recipes.Grounder
 	logger    *slog.Logger
 	groundJob *jobs.Kind[GroundJob]
+	// backfillJob enqueues KindBackfillPhotos jobs.
+	backfillJob *jobs.Kind[BackfillJob]
 	// wake is called after every committed change to a dinner's method or
 	// minutes, which can move its crock-pot start reminder.
 	wake func()
@@ -162,8 +169,8 @@ type Service struct {
 
 // NewService returns a service over st that grounds dinners with grounder,
 // calls wake (the scheduler's Wake) after every committed change to a
-// dinner's method or minutes, and registers the KindGroundDinner job on
-// runner, which must not have started yet.
+// dinner's method or minutes, and registers the KindGroundDinner and
+// KindBackfillPhotos jobs on runner, which must not have started yet.
 func NewService(st *store.Store, grounder *recipes.Grounder, runner *jobs.Runner, wake func(), logger *slog.Logger) (*Service, error) {
 	switch {
 	case st == nil:
@@ -184,6 +191,14 @@ func NewService(st *store.Store, grounder *recipes.Grounder, runner *jobs.Runner
 		return nil, err
 	}
 	s.groundJob = kind
+	backfill, err := jobs.Register(runner, KindBackfillPhotos, s.runBackfillJob)
+	if err != nil {
+		return nil, err
+	}
+	if err := backfill.ServiceOnly(); err != nil {
+		return nil, err
+	}
+	s.backfillJob = backfill
 	return s, nil
 }
 
@@ -702,6 +717,8 @@ func fromRow(r db.Dinner) Dinner {
 		UpdatedBy:       nullString(r.UpdatedBy),
 		CreatedAt:       r.CreatedAt,
 		UpdatedAt:       r.UpdatedAt,
+
+		PhotoBackfillEligible: PhotoBackfillEligible(r),
 	}
 	if r.MinutesTotal.Valid {
 		d.MinutesTotal = &r.MinutesTotal.Int64

internal/dinners/export_test.go

@@ -74,3 +74,18 @@ func (s *Service) RunGroundJob(ctx context.Context, p GroundJob) (string, error)
 
 // SafeSourceURL is safeSourceURL.
 var SafeSourceURL = safeSourceURL
+
+// RunBackfillJob runs the KindBackfillPhotos handler for p in the caller's
+// goroutine, as the job runner would, and returns its outcomes: on an
+// error, those it last published.
+func (s *Service) RunBackfillJob(ctx context.Context, p BackfillJob) ([]BackfillOutcome, error) {
+	var published backfillResult
+	r, err := s.runBackfillJob(ctx, p, func(partial backfillResult) error {
+		published = partial
+		return nil
+	})
+	if err != nil {
+		return published.Dinners, err
+	}
+	return r.Dinners, nil
+}

internal/dinners/grounding_test.go

@@ -39,6 +39,7 @@ type env struct {
 	runner *jobs.Runner
 	search *recipestest.Search
 	pages  *recipestest.Pages
+	images *recipestest.Images
 	model  *llm.Fake
 }
 
@@ -108,7 +109,7 @@ func newUnstartedEnvOn(t *testing.T, st *store.Store, dataDir string) *env {
 	if err != nil {
 		t.Fatal(err)
 	}
-	return &env{svc: svc, st: st, runner: runner, search: search, pages: pages, model: model}
+	return &env{svc: svc, st: st, runner: runner, search: search, pages: pages, images: images, model: model}
 }
 
 // start starts the env's job runner, stopping it when the test ends.

internal/server/dinners.go

@@ -112,6 +112,20 @@ func postDinner(logger *slog.Logger, svc *dinners.Service) http.HandlerFunc {
 	}
 }
 
+// postPhotoBackfill starts a photo backfill over every dinner without a
+// photo and answers 202 with its job id, which GET /jobs/{id} reports on,
+// without waiting for the job.
+func postPhotoBackfill(logger *slog.Logger, svc *dinners.Service) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		id, err := svc.StartPhotoBackfill(r.Context())
+		if err != nil {
+			writeDinnerError(w, r, logger, "start photo backfill", err)
+			return
+		}
+		writeJSON(w, r, logger, http.StatusAccepted, jobCreated{ID: id})
+	}
+}
+
 // patchDinner edits a dinner, recording the session's actor as updated_by.
 func patchDinner(logger *slog.Logger, svc *dinners.Service) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {

internal/server/grounding_test.go

@@ -556,3 +556,43 @@ func TestPostDinnerEnqueueFailureAnswersFailedDinner(t *testing.T) {
 		t.Errorf("listed = %v, want failed", got)
 	}
 }
+
+func TestPostPhotoBackfillGroundsFailedDinners(t *testing.T) {
+	t.Parallel()
+	e := newDinnersEnv(t).withRecipes(t)
+	e.start(t)
+
+	// No model reply is scripted, so grounding on creation fails.
+	rec, created := callDinners(t, e.h, http.MethodPost, dinnersPath, `{"name":"Chicken curry"}`)
+	if rec.Code != http.StatusCreated {
+		t.Fatalf("POST = %d: %s", rec.Code, rec.Body)
+	}
+	id := created["id"].(string)
+	waitJob(t, e.h, created["grounding_job_id"].(string))
+	if d := listedDinner(t, e.h, id); d["grounding_status"] != "failed" || d["photo_url"] != nil {
+		t.Fatalf("dinner = %v, want failed without a photo", d)
+	}
+
+	e.model.QueueGenerate(recipestest.NormalizedReply("stovetop", 40))
+	rec, started := callDinners(t, e.h, http.MethodPost, "/api/v1/settings/photo-backfill", "")
+	jobID, _ := started["id"].(string)
+	if rec.Code != http.StatusAccepted || jobID == "" {
+		t.Fatalf("POST photo-backfill = %d: %s, want 202 with a job id", rec.Code, rec.Body)
+	}
+	job := waitJob(t, e.h, jobID)
+	want := fmt.Sprintf(`map[dinners:[map[dinner_id:%s outcome:grounded reason:<nil>]]]`, id)
+	if job["status"] != "succeeded" || job["kind"] != "backfill_dinner_photos" || fmt.Sprint(job["result"]) != want {
+		t.Fatalf("job = %v, want backfill_dinner_photos succeeded with %s", job, want)
+	}
+	d := listedDinner(t, e.h, id)
+	photo, _ := d["photo_url"].(string)
+	if d["grounding_status"] != "grounded" || d["grounding_error"] != nil || !strings.HasSuffix(photo, "-400.jpg") {
+		t.Errorf("dinner = %v, want grounded with a photo", d)
+	}
+
+	// The backfill is started only by its own endpoint.
+	rec, _ = callDinners(t, e.h, http.MethodPost, "/api/v1/jobs", `{"kind":"backfill_dinner_photos","payload":{}}`)
+	if rec.Code != http.StatusBadRequest || !strings.Contains(rec.Body.String(), "service_only_kind") {
+		t.Errorf("POST /jobs backfill = %d: %s, want 400 service_only_kind", rec.Code, rec.Body)
+	}
+}

internal/server/server.go

@@ -162,6 +162,7 @@ func NewRouter(logger *slog.Logger, version string, trusted forwarded.Trusted, a
 	apiV1.Get("/settings/chat-webhook", getChatWebhook(logger, chatSvc))
 	apiV1.Put("/settings/chat-webhook", putChatWebhook(logger, chatSvc))
 	apiV1.Post("/settings/chat-webhook/test", postChatWebhookTest(logger, chatSvc))
+	apiV1.Post("/settings/photo-backfill", postPhotoBackfill(logger, dinnerSvc))
 	apiV1.Get("/dinners", getDinners(logger, dinnerSvc))
 	apiV1.Post("/dinners", postDinner(logger, dinnerSvc))
 	apiV1.Get("/dinners/{id}", getDinner(logger, dinnerSvc))

migrations/queries/dinners.sql

@@ -181,3 +181,65 @@ WHERE grounding_status = 'pending'
 			AND json_extract(jobs.payload, '$.dinner_id') = dinners.id
 			AND json_extract(jobs.payload, '$.revision') = dinners.grounding_revision
 	);
+
+-- name: ListPhotoBackfillDinners :many
+-- The dinners a photo backfill grounds by name: unarchived, failed, with no
+-- photo and no pasted recipe_url (the household's own link is never
+-- replaced by a search), ordered by normalized name. A non-NULL id narrows
+-- the list to that dinner. Grounded, manual and pending dinners are never
+-- listed: they have a recipe, or a ground_dinner job grounding them.
+SELECT * FROM dinners
+WHERE status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND id = COALESCE(sqlc.narg(id), id)
+ORDER BY normalized_name, id;
+
+-- name: GetActivePhotoBackfillJob :one
+-- The id of the full photo backfill job (no dinner_id) of kind queued or
+-- running. StartPhotoBackfill enqueues one only when there is none, so
+-- there is at most one. Returns sql.ErrNoRows when there is none.
+SELECT id
+FROM jobs
+WHERE kind = sqlc.arg(kind)
+	AND status IN ('queued', 'running')
+	AND json_extract(payload, '$.dinner_id') IS NULL
+ORDER BY id
+LIMIT 1;
+
+-- name: GroundFailedDinner :execrows
+-- Records a photo backfill's grounding of a dinner it listed, while the
+-- dinner is still unarchived, failed, photoless, without a recipe_url and at
+-- the grounding_revision it was listed at; a dinner grounded by hand,
+-- renamed or archived meanwhile is left as it is (0 rows). Like
+-- GroundPendingDinner it keeps a method, minutes_total or difficulty a
+-- member already set.
+UPDATE dinners
+SET recipe_id = sqlc.arg(recipe_id),
+	photo_id = sqlc.arg(photo_id),
+	method = COALESCE(method, sqlc.arg(method)),
+	minutes_total = COALESCE(minutes_total, sqlc.arg(minutes_total)),
+	difficulty = COALESCE(difficulty, sqlc.arg(difficulty)),
+	grounding_status = 'grounded',
+	grounding_error = NULL,
+	updated_at = sqlc.arg(updated_at)
+WHERE id = sqlc.arg(id)
+	AND status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND grounding_revision = sqlc.arg(grounding_revision);
+
+-- name: RefailDinnerGrounding :execrows
+-- Records why a photo backfill could not ground a dinner it listed, under
+-- the same conditions as GroundFailedDinner (0 rows otherwise).
+UPDATE dinners
+SET grounding_error = sqlc.arg(grounding_error),
+	updated_at = sqlc.arg(updated_at)
+WHERE id = sqlc.arg(id)
+	AND status <> 'archived'
+	AND grounding_status = 'failed'
+	AND photo_id IS NULL
+	AND recipe_url IS NULL
+	AND grounding_revision = sqlc.arg(grounding_revision);

web/src/dinners/DinnerDetailActions.test.tsx

@@ -52,6 +52,8 @@ function listed(is_favorite: boolean) {
     is_favorite,
     photo_url: null,
     grounding_status: 'grounded',
+    grounding_error: null,
+    photo_backfill_eligible: false,
   }
 }
 

web/src/dinners/api.ts

@@ -20,8 +20,12 @@ export type Difficulty = (typeof DIFFICULTIES)[number]
 export type GroundingStatus = (typeof GROUNDING_STATUSES)[number]
 
 // Regular is the part of a dinner the Regulars screens read. photo_id is
-// the stored photo's id, from photo_url. pending is true while an
-// optimistic add waits for the server; its id is not the server's.
+// the stored photo's id, from photo_url. grounding_error is why its
+// grounding failed, a reason code, set only while grounding_status is
+// failed. photo_backfill_eligible is whether a photo backfill (Search
+// again) searches for it: never for a dinner with a pasted recipe link.
+// pending is true while an optimistic add waits for the server;
+// its id is not the server's.
 export type Regular = {
   id: string
   name: string
@@ -31,6 +35,8 @@ export type Regular = {
   is_favorite: boolean
   photo_id: string | undefined
   grounding_status: GroundingStatus
+  grounding_error: string | null
+  photo_backfill_eligible: boolean
   pending?: true
 }
 
@@ -113,6 +119,9 @@ export function parseDinner(where: string, body: unknown): Regular {
     throw new Error(`${where}: response is not a dinner`)
   }
   if (typeof body.is_favorite !== 'boolean') throw new Error(`${where}: is_favorite is not a boolean`)
+  if (typeof body.photo_backfill_eligible !== 'boolean') {
+    throw new Error(`${where}: photo_backfill_eligible is not a boolean`)
+  }
   return {
     id: body.id,
     name: body.name,
@@ -122,6 +131,11 @@ export function parseDinner(where: string, body: unknown): Regular {
     is_favorite: body.is_favorite,
     photo_id: photoId(where, body.photo_url),
     grounding_status: oneOf(where, 'grounding_status', GROUNDING_STATUSES, body.grounding_status),
+    grounding_error: nullable(body.grounding_error, (v) => {
+      if (typeof v !== 'string') throw new Error(`${where}: grounding_error is not a string`)
+      return v
+    }),
+    photo_backfill_eligible: body.photo_backfill_eligible,
   }
 }
 
@@ -225,6 +239,8 @@ export function useAddRegular() {
         is_favorite: false,
         photo_id: undefined,
         grounding_status: 'pending',
+        grounding_error: null,
+        photo_backfill_eligible: false,
         pending: true,
       }
       await changeOptimistically(queryClient, (list) => [...list, optimistic])

web/src/plan/PlanPage.test.tsx

@@ -114,6 +114,8 @@ function dinner(id: string, name: string, difficulty: string | null, minutes: nu
     is_favorite: false,
     photo_url: null,
     grounding_status: 'grounded',
+    grounding_error: null,
+    photo_backfill_eligible: false,
   }
 }
 

web/src/settings/RegularsSettingsPage.module.css

@@ -300,3 +300,27 @@
 .lock {
   display: contents;
 }
+
+/* Find missing photos: its hint, button and progress, padded like a sheet. */
+.card.backfill {
+  display: flex;
+  flex-direction: column;
+  gap: 12px;
+  padding: 16px;
+}
+
+/* Why a Regular's recipe was not found. */
+.reason {
+  margin: 0;
+  padding: 12px 14px;
+  border-radius: 18px;
+  background: var(--color-neutral-100);
+  color: var(--color-neutral-900);
+  font-family: var(--font-body);
+  font-size: 14px;
+  line-height: 1.4;
+}
+
+.reasonLabel {
+  font-weight: 700;
+}

web/src/settings/RegularsSettingsPage.test.tsx

@@ -85,7 +85,11 @@ const TACOS = dinnerBody('d1', 'Beef tacos', {
   photo_url: '/media/photos/p1-400.jpg',
 })
 const CHILI = dinnerBody('d2', 'Chili', { method: 'crockpot', minutes_total: 480, grounding_status: 'pending' })
-const PIZZA = dinnerBody('d3', 'Pizza night', { grounding_status: 'failed', grounding_error: 'no_recipe_found' })
+const PIZZA = dinnerBody('d3', 'Pizza night', {
+  grounding_status: 'failed',
+  grounding_error: 'no_recipe_found',
+  photo_backfill_eligible: true,
+})
 const ROAST = dinnerBody('d4', 'Pot roast', { method: 'oven', grounding_status: 'manual' })
 
 function rows() {
@@ -496,6 +500,204 @@ describe('RegularsSettingsPage', () => {
     })
   })
 
+  describe('Photo backfill', () => {
+    const BACKFILL = 'POST /api/v1/settings/photo-backfill'
+    const JOB = 'GET /api/v1/jobs/j9'
+
+    // backfillJob is a GET /api/v1/jobs/j9 answer for the photo backfill.
+    function backfillJob(status: string, result: unknown = null, error: string | null = null) {
+      return jsonResponse(200, { id: 'j9', kind: 'backfill_dinner_photos', status, result, error })
+    }
+
+    // polls answers each poll of the job with the next answer, then the last
+    // one again; an answer may change the server first.
+    function polls(...answers: (() => Response)[]) {
+      let next = 0
+      return () => answers[Math.min(next++, answers.length - 1)]!()
+    }
+
+    function fixSheet() {
+      return within(screen.getByRole('dialog', { name: 'Fix recipe' }))
+    }
+
+    const NO_CANDIDATES = { 'GET /api/v1/dinners/d3/recipe-candidates': () => jsonResponse(200, { candidates: [] }) }
+
+    it('shows why the recipe was not found in the fix-recipe sheet', async () => {
+      stubServer([PIZZA], NO_CANDIDATES)
+      await renderPage()
+
+      openRow('Pizza night')
+      expect(fixSheet().getByText('No recipe page the search found had a recipe we could use.')).toBeInTheDocument()
+      expect(fixSheet().getByRole('button', { name: 'Search again' })).toBeEnabled()
+    })
+
+    it('offers no search again for a pasted link that had no usable recipe', async () => {
+      const linked = { ...PIZZA, grounding_error: 'recipe_unusable', photo_backfill_eligible: false }
+      stubServer([linked], NO_CANDIDATES)
+      await renderPage()
+
+      openRow('Pizza night')
+      expect(fixSheet().getByText('The pasted recipe link has no recipe we could use.')).toBeInTheDocument()
+      expect(fixSheet().queryByRole('button', { name: 'Search again' })).not.toBeInTheDocument()
+    })
+
+    it('offers no search again for a dinner with a pasted link whatever its reason', async () => {
+      const linked = { ...PIZZA, grounding_error: 'ai_failed', photo_backfill_eligible: false }
+      stubServer([linked], NO_CANDIDATES)
+      await renderPage()
+
+      openRow('Pizza night')
+      expect(fixSheet().getByText("The AI assistant couldn't read the recipe.")).toBeInTheDocument()
+      expect(fixSheet().queryByRole('button', { name: 'Search again' })).not.toBeInTheDocument()
+    })
+
+    it('searches again from the fix-recipe sheet and shows the new reason when it still fails', async () => {
+      const server = stubServer([PIZZA], {
+        ...NO_CANDIDATES,
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: polls(
+          () => backfillJob('running'),
+          () => {
+            server.regulars = [{ ...PIZZA, grounding_error: 'search_failed' }]
+            return backfillJob('succeeded', { dinners: [{ dinner_id: 'd3', outcome: 'failed', reason: 'search_failed' }] })
+          },
+        ),
+      })
+      await renderPage()
+
+      openRow('Pizza night')
+      fireEvent.click(fixSheet().getByRole('button', { name: 'Search again' }))
+      expect(await fixSheet().findByRole('button', { name: 'Searching again…' })).toBeDisabled()
+      expect(fixSheet().getByText('Searching again… this can take a few minutes.')).toBeInTheDocument()
+      expect(sent(server.fetchMock)).toEqual([{ request: BACKFILL, body: undefined }])
+
+      expect(await fixSheet().findByText('Searched again: still no recipe found.', {}, { timeout: 3000 })).toBeInTheDocument()
+      expect(await fixSheet().findByText("The recipe search couldn't be reached.")).toBeInTheDocument()
+      expect(fixSheet().getByRole('button', { name: 'Search again' })).toBeEnabled()
+    })
+
+    it('closes the fix-recipe sheet once searching again finds the recipe and its photo', async () => {
+      const server = stubServer([PIZZA], {
+        ...NO_CANDIDATES,
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: () => {
+          server.regulars = [{ ...PIZZA, grounding_status: 'grounded', grounding_error: null, photo_url: '/media/photos/p8-400.jpg' }]
+          return backfillJob('succeeded', { dinners: [{ dinner_id: 'd3', outcome: 'grounded', reason: null }] })
+        },
+      })
+      await renderPage()
+
+      openRow('Pizza night')
+      fireEvent.click(fixSheet().getByRole('button', { name: 'Search again' }))
+      await waitFor(() => expect(screen.queryByRole('dialog')).not.toBeInTheDocument())
+      expect(within(row('Pizza night')).getByText('Recipe found')).toBeInTheDocument()
+      expect(within(row('Pizza night')).getByRole('img', { name: 'Pizza night' })).toHaveAttribute('src', '/media/photos/p8-400.jpg')
+    })
+
+    it('says when a dinner was not included in the search', async () => {
+      stubServer([PIZZA], {
+        ...NO_CANDIDATES,
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: () => backfillJob('succeeded', { dinners: [] }),
+      })
+      await renderPage()
+
+      openRow('Pizza night')
+      fireEvent.click(fixSheet().getByRole('button', { name: 'Search again' }))
+      expect(await fixSheet().findByText('This dinner was not included in that search.')).toBeInTheDocument()
+    })
+
+    it('finds missing photos: the placeholder shows until the backfill finds a photo', async () => {
+      const found = { ...PIZZA, grounding_status: 'grounded', grounding_error: null, photo_url: '/media/photos/p8-400.jpg' }
+      const server = stubServer([TACOS, PIZZA], {
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: polls(
+          () => backfillJob('queued'),
+          () => {
+            server.regulars = [TACOS, found]
+            return backfillJob('succeeded', { dinners: [{ dinner_id: 'd3', outcome: 'grounded', reason: null }] })
+          },
+        ),
+      })
+      await renderPage()
+
+      // The photoless dinner shows the striped placeholder, not an image.
+      const placeholder = within(row('Pizza night')).getByRole('img', { name: 'Pizza night' })
+      expect(placeholder).not.toHaveAttribute('src')
+      expect(placeholder).toHaveTextContent('dish photo')
+
+      const trigger = screen.getByRole('button', { name: 'Find missing photos' })
+      expectMinHitArea(trigger)
+      fireEvent.click(trigger)
+      expect(await screen.findByRole('button', { name: 'Finding photos…' })).toBeDisabled()
+      expect(await screen.findByText('Finding missing photos… this can take a few minutes.')).toBeInTheDocument()
+      expect(within(row('Pizza night')).getByRole('img', { name: 'Pizza night' })).not.toHaveAttribute('src')
+      expect(sent(server.fetchMock)).toEqual([{ request: BACKFILL, body: undefined }])
+
+      expect(await screen.findByText('Photo search done: 1 photo found.', {}, { timeout: 3000 })).toBeInTheDocument()
+      await waitFor(() =>
+        expect(within(row('Pizza night')).getByRole('img', { name: 'Pizza night' })).toHaveAttribute('src', '/media/photos/p8-400.jpg'),
+      )
+      expect(screen.getByRole('button', { name: 'Find missing photos' })).toBeEnabled()
+      expect(reads(server.fetchMock)).toBe(2)
+    })
+
+    it('sums up what the backfill did', async () => {
+      stubServer([PIZZA], {
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: () =>
+          backfillJob('succeeded', {
+            dinners: [
+              { dinner_id: 'd3', outcome: 'grounded', reason: null },
+              { dinner_id: 'd5', outcome: 'grounded', reason: null },
+              { dinner_id: 'd6', outcome: 'failed', reason: 'no_recipe_found' },
+              { dinner_id: 'd7', outcome: 'skipped', reason: null },
+            ],
+          }),
+      })
+      await renderPage()
+
+      fireEvent.click(screen.getByRole('button', { name: 'Find missing photos' }))
+      expect(
+        await screen.findByText(
+          'Photo search done: 2 photos found · 1 dinner still needs a recipe · 1 dinner changed meanwhile, left as it was.',
+        ),
+      ).toBeInTheDocument()
+    })
+
+    it('says when no dinner was missing a photo', async () => {
+      stubServer([TACOS], {
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: () => backfillJob('succeeded', { dinners: [] }),
+      })
+      await renderPage()
+
+      fireEvent.click(screen.getByRole('button', { name: 'Find missing photos' }))
+      expect(await screen.findByText('No dinners were missing a photo.')).toBeInTheDocument()
+    })
+
+    it('says why the backfill stopped', async () => {
+      stubServer([PIZZA], {
+        [BACKFILL]: () => jsonResponse(202, { id: 'j9' }),
+        [JOB]: () => backfillJob('failed', { dinners: [] }, 'internal_error'),
+      })
+      await renderPage()
+
+      fireEvent.click(screen.getByRole('button', { name: 'Find missing photos' }))
+      expect(await screen.findByRole('alert')).toHaveTextContent('Photo search stopped: Something went wrong on the server.')
+      expect(screen.getByRole('button', { name: 'Find missing photos' })).toBeEnabled()
+    })
+
+    it('says why the backfill could not start', async () => {
+      stubServer([PIZZA], { [BACKFILL]: () => errorResponse(403, 'csrf_failed') })
+      await renderPage()
+
+      fireEvent.click(screen.getByRole('button', { name: 'Find missing photos' }))
+      expect(await screen.findByRole('alert')).toHaveTextContent('Photo search not started: csrf_failed message')
+      expect(screen.getByRole('button', { name: 'Find missing photos' })).toBeEnabled()
+    })
+  })
+
   it('refreshes grounding status while any Regular is finding its recipe, then stops', async () => {
     // Only intervals are faked: TanStack Query polls with setInterval, and
     // Testing Library's waits keep their real timers.

web/src/settings/RegularsSettingsPage.tsx

@@ -34,6 +34,7 @@ import { Sheet } from '../ui/Sheet.tsx'
 import { Skeleton } from '../ui/Skeleton.tsx'
 import { Tag } from '../ui/Tag.tsx'
 import { TextInput } from '../ui/TextInput.tsx'
+import { usePhotoBackfill, type BackfillResult, type PhotoBackfill } from './photoBackfill.ts'
 import styles from './RegularsSettingsPage.module.css'
 
 // The most minutes a dinner may take, as the dinners API allows: a day.
@@ -57,7 +58,8 @@ const LINK = /^https?:\/\//i
 // method and minutes, favorite heart and whether its recipe was found. The
 // field at the top adds one by name, or by pasted recipe link with a name
 // the household gives it; tapping one
-// edits or removes it, or fixes its recipe when none was found.
+// edits or removes it, or fixes its recipe when none was found. Find
+// missing photos searches again for every dinner left without one.
 export function RegularsSettingsPage() {
   const navigate = useNavigate()
   return (
@@ -86,6 +88,7 @@ function RegularsManager() {
   const remove = useRemoveRegular()
   const favorite = useToggleFavorite()
   const { refusals, track } = useRefusals()
+  const backfill = usePhotoBackfill()
   const [draft, setDraft] = useState('')
   const [linkName, setLinkName] = useState('')
   const [open, setOpen] = useState<Open>(null)
@@ -94,8 +97,10 @@ function RegularsManager() {
   const isLink = LINK.test(draft.trim())
   const ready = draft.trim() !== '' && (!isLink || linkName.trim() !== '')
   const list = regulars.data ?? []
-  // A Regular removed meanwhile closes its sheet.
+  // A Regular removed meanwhile closes its sheet, and one whose recipe was
+  // found meanwhile closes its fix-recipe sheet.
   const opened = open && list.find((regular) => regular.id === open.id && !regular.pending)
+  const fixing = opened && open.sheet === 'fix' && opened.grounding_status === 'failed' ? opened : undefined
 
   function onSubmit(event: FormEvent<HTMLFormElement>) {
     event.preventDefault()
@@ -188,6 +193,7 @@ function RegularsManager() {
           )}
         </section>
       )}
+      {regulars.isSuccess && <FindMissingPhotos backfill={backfill} />}
 
       {opened && open.sheet === 'edit' && (
         <EditRegularSheet
@@ -197,12 +203,13 @@ function RegularsManager() {
           onRemove={() => onRemove(opened)}
         />
       )}
-      {opened && open.sheet === 'fix' && (
+      {fixing && (
         <FixRecipeSheet
-          key={opened.id}
-          regular={opened}
+          key={fixing.id}
+          regular={fixing}
+          backfill={backfill}
           onClose={() => setOpen(null)}
-          onEdit={() => setOpen({ sheet: 'edit', id: opened.id })}
+          onEdit={() => setOpen({ sheet: 'edit', id: fixing.id })}
         />
       )}
     </>
@@ -231,6 +238,88 @@ function GroundingBadge({ status }: { status: GroundingStatus }) {
   return <Tag className={`${styles.badge} ${badge.className}`}>{badge.text}</Tag>
 }
 
+// groundingReasons says why a dinner's recipe was not found, by its
+// grounding_error reason code.
+const groundingReasons: Record<string, string> = {
+  no_recipe_found: 'No recipe page the search found had a recipe we could use.',
+  recipe_unusable: 'The pasted recipe link has no recipe we could use.',
+  search_not_configured: 'Recipe search is not set up. An admin can set it up in Settings.',
+  search_failed: "The recipe search couldn't be reached.",
+  ai_failed: "The AI assistant couldn't read the recipe.",
+  interrupted: 'The server stopped while finding the recipe.',
+  grounding_error: 'Something went wrong on the server.',
+}
+
+// groundingReason says why a dinner's recipe was not found; an unknown
+// reason code is shown as it is.
+export function groundingReason(code: string): string {
+  return groundingReasons[code] ?? code
+}
+
+// backfillJobFailures says why a photo backfill job failed, by its failure
+// code; an unknown code is shown as it is.
+const backfillJobFailures: Record<string, string> = {
+  internal_error: 'Something went wrong on the server.',
+}
+
+// backfillProblem says what went wrong with a photo backfill, or null.
+function backfillProblem(backfill: PhotoBackfill): string | null {
+  if (backfill.startError) return `Photo search not started: ${backfill.startError.message}`
+  if (backfill.status === 'failed' && backfill.error) {
+    return `Photo search stopped: ${backfillJobFailures[backfill.error.message] ?? backfill.error.message}`
+  }
+  if (backfill.error) return `Couldn't check on the photo search: ${backfill.error.message}`
+  return null
+}
+
+function count(n: number, one: string, many: string): string {
+  return `${n} ${n === 1 ? one : many}`
+}
+
+// backfillSummary says what a finished photo backfill did.
+function backfillSummary(result: BackfillResult): string {
+  const tally = (outcome: string) => result.dinners.filter((dinner) => dinner.outcome === outcome).length
+  if (result.dinners.length === 0) return 'No dinners were missing a photo.'
+  const parts = [count(tally('grounded'), 'photo found', 'photos found')]
+  const failed = tally('failed')
+  if (failed > 0) parts.push(`${count(failed, 'dinner still needs', 'dinners still need')} a recipe`)
+  const skipped = tally('skipped')
+  if (skipped > 0) parts.push(`${count(skipped, 'dinner', 'dinners')} changed meanwhile, left as it was`)
+  return `Photo search done: ${parts.join(' · ')}.`
+}
+
+// FindMissingPhotos searches again, by name, for a recipe and photo for
+// every dinner left without a photo because its recipe was not found.
+function FindMissingPhotos({ backfill }: { backfill: PhotoBackfill }) {
+  const headingId = useId()
+  const problem = backfillProblem(backfill)
+  return (
+    <section aria-labelledby={headingId} className={styles.section}>
+      <h2 id={headingId} className={styles.groupLabel}>
+        Missing photos
+      </h2>
+      <Card className={`${styles.card} ${styles.backfill}`}>
+        <p className={styles.hint}>
+          Search again for a recipe and photo for every dinner whose recipe wasn&apos;t found. Dinners with a pasted
+          recipe link keep it.
+        </p>
+        <Button disabled={backfill.active} onClick={backfill.start}>
+          {backfill.active ? 'Finding photos…' : 'Find missing photos'}
+        </Button>
+        <p role="status" className={styles.status}>
+          {backfill.active && 'Finding missing photos… this can take a few minutes.'}
+          {backfill.status === 'succeeded' && backfill.result && backfillSummary(backfill.result)}
+        </p>
+        {problem && (
+          <p role="alert" className={styles.problem}>
+            {problem}
+          </p>
+        )}
+      </Card>
+    </section>
+  )
+}
+
 type RegularRowProps = {
   regular: Regular
   onOpen: () => void
@@ -447,6 +536,7 @@ const fixPaths = [
 
 type FixRecipeSheetProps = {
   regular: Regular
+  backfill: PhotoBackfill
   onClose: () => void
   onEdit: () => void
 }
@@ -459,10 +549,20 @@ function lines(text: string): string[] {
     .filter((line) => line !== '')
 }
 
-// FixRecipeSheet grounds a Regular whose recipe was not found: in one of
-// up to three recipe pages found for its name, in a pasted recipe link, or
-// in ingredients and steps typed in.
-function FixRecipeSheet({ regular, onClose, onEdit }: FixRecipeSheetProps) {
+// retryOutcomes says what searching again did for this dinner.
+const retryOutcomes = {
+  grounded: 'Searched again: a recipe and photo were found.',
+  failed: 'Searched again: still no recipe found.',
+  skipped: 'It changed while it was searched again, so it was left as it is.',
+} as const
+
+// FixRecipeSheet shows why a Regular's recipe was not found and grounds it:
+// by searching its name again (a photo backfill), in one of up to three
+// recipe pages found for its name, in a pasted recipe link, or in
+// ingredients and steps typed in. Searching again is offered only for a
+// dinner the backfill searches (photo_backfill_eligible): it never replaces
+// a pasted link.
+function FixRecipeSheet({ regular, backfill, onClose, onEdit }: FixRecipeSheetProps) {
   const [path, setPath] = useState<FixPath>('pick')
   const ground = useGroundRegular()
 
@@ -475,6 +575,15 @@ function FixRecipeSheet({ regular, onClose, onEdit }: FixRecipeSheetProps) {
       <p className={styles.hint}>
         We couldn&apos;t find a recipe for {regular.name}. Pick a match below, paste a recipe link or type it in.
       </p>
+      {regular.grounding_error !== null && (
+        <p className={styles.reason}>
+          <span className={styles.reasonLabel}>Why: </span>
+          {groundingReason(regular.grounding_error)}
+        </p>
+      )}
+      {regular.photo_backfill_eligible && (
+        <RetrySearch id={regular.id} backfill={backfill} disabled={ground.isPending} />
+      )}
       <Segmented
         label="How to fix the recipe"
         options={fixPaths}
@@ -504,6 +613,34 @@ function FixRecipeSheet({ regular, onClose, onEdit }: FixRecipeSheetProps) {
   )
 }
 
+// RetrySearch searches the dinner's name again with a photo backfill, and
+// says what it did for the dinner once it ends. The backfill searches
+// every dinner without a photo, this one with them.
+function RetrySearch({ id, backfill, disabled }: { id: string; backfill: PhotoBackfill; disabled: boolean }) {
+  const problem = backfillProblem(backfill)
+  const tried = backfill.result?.dinners.find((dinner) => dinner.dinner_id === id)
+  let status: string | null = null
+  if (backfill.active) status = 'Searching again… this can take a few minutes.'
+  else if (backfill.status === 'succeeded') {
+    status = tried ? retryOutcomes[tried.outcome] : 'This dinner was not included in that search.'
+  }
+  return (
+    <>
+      <Button variant="outline" disabled={disabled || backfill.active} onClick={backfill.start}>
+        {backfill.active ? 'Searching again…' : 'Search again'}
+      </Button>
+      <p role="status" className={styles.status}>
+        {status}
+      </p>
+      {problem && (
+        <p role="alert" className={styles.problem}>
+          {problem}
+        </p>
+      )}
+    </>
+  )
+}
+
 function PickCandidate({ id, onPick }: { id: string; onPick: (candidateId: string) => void }) {
   const candidates = useRecipeCandidates(id)
   if (candidates.isPending) return <Skeleton className={styles.candidatesLoading} />

web/src/settings/photoBackfill.test.tsx

@@ -0,0 +1,63 @@
+import { QueryClientProvider } from '@tanstack/react-query'
+import { act, renderHook, waitFor } from '@testing-library/react'
+import type { ReactNode } from 'react'
+import { afterEach, describe, expect, it, vi } from 'vitest'
+import { ApiProvider, createApiClient, createQueryClient, createVerifiedSession } from '../api/index.ts'
+import { HOME_KEY } from '../home/api.ts'
+import { planKey } from '../plan/api.ts'
+import { createNetworkStatus } from '../pwa/networkStatus.ts'
+import { fakeLocation, jsonResponse, ME, stubFetch } from '../test/fetchStub.ts'
+import { parseBackfillResult, usePhotoBackfill } from './photoBackfill.ts'
+
+afterEach(() => {
+  vi.unstubAllGlobals()
+})
+
+const RESULT = { dinners: [{ dinner_id: 'd3', outcome: 'grounded', reason: null }] }
+
+function backfillJob(status: string, result: unknown) {
+  return jsonResponse(200, { id: 'j9', kind: 'backfill_dinner_photos', status, result, error: null })
+}
+
+describe('usePhotoBackfill', () => {
+  it('reads the plans, home and dinners again once the backfill ends, so a found photo shows everywhere', async () => {
+    const fetchMock = stubFetch(
+      jsonResponse(200, ME),
+      jsonResponse(202, { id: 'j9' }),
+      backfillJob('running', null),
+      backfillJob('succeeded', RESULT),
+    )
+    const api = createApiClient(fakeLocation(), createNetworkStatus(window), createVerifiedSession(localStorage))
+    await api.loadMe()
+    const queryClient = createQueryClient()
+    const plan = planKey('2026-10-05')
+    for (const key of [plan, HOME_KEY, ['dinners', 'quest_log']]) queryClient.setQueryData(key, {})
+    const wrapper = ({ children }: { children: ReactNode }) => (
+      <ApiProvider client={api}>
+        <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+      </ApiProvider>
+    )
+    const { result } = renderHook(() => usePhotoBackfill(), { wrapper })
+
+    act(() => result.current.start())
+    await waitFor(() => expect(result.current.status).toBe('running'))
+    expect(result.current.active).toBe(true)
+    expect(queryClient.getQueryState(plan)?.isInvalidated).toBe(false)
+
+    await waitFor(() => expect(result.current.status).toBe('succeeded'), { timeout: 3000 })
+    expect(result.current).toMatchObject({ active: false, error: null, result: RESULT })
+    expect(fetchMock).toHaveBeenCalledTimes(4)
+    for (const key of [plan, HOME_KEY, ['dinners', 'quest_log']]) {
+      expect(queryClient.getQueryState(key)?.isInvalidated).toBe(true)
+    }
+  })
+})
+
+describe('parseBackfillResult', () => {
+  it('refuses a result that is not a photo backfill', () => {
+    expect(() => parseBackfillResult('job', { dinners: [{ dinner_id: 'd3', outcome: 'lost', reason: null }] })).toThrow(
+      'job: outcome "lost" is not one of grounded, failed, skipped',
+    )
+    expect(() => parseBackfillResult('job', { stage: 'x' })).toThrow('job: result is not a photo backfill')
+  })
+})

web/src/settings/photoBackfill.ts

@@ -0,0 +1,110 @@
+import { useMutation, useQueryClient } from '@tanstack/react-query'
+import { useEffect, useMemo, useState } from 'react'
+import { useApi, useJob, type JobStatus } from '../api/index.ts'
+import { isRecord } from '../api/errors.ts'
+import { oneOf } from '../dinners/api.ts'
+import { HOME_KEY } from '../home/api.ts'
+
+export const PHOTO_BACKFILL_PATH = '/api/v1/settings/photo-backfill'
+
+// BACKFILL_JOB_KIND is the kind of the job a photo backfill runs as.
+export const BACKFILL_JOB_KIND = 'backfill_dinner_photos'
+
+export const BACKFILL_OUTCOMES = ['grounded', 'failed', 'skipped'] as const
+export type BackfillOutcome = (typeof BACKFILL_OUTCOMES)[number]
+
+// BackfillDinner is what a photo backfill did for one dinner. reason is
+// set only when it failed: the reason code now its grounding_error.
+export type BackfillDinner = { dinner_id: string; outcome: BackfillOutcome; reason: string | null }
+
+// BackfillResult is the photo backfill job's result: each dinner it tried,
+// in the order it tried them.
+export type BackfillResult = { dinners: BackfillDinner[] }
+
+// Every query showing a dinner's photo, read again once a backfill ends:
+// the dinners lists and details, the plans and the home dashboard.
+const PHOTO_KEYS = [['dinners'], ['plans'], HOME_KEY] as const
+
+// parseBackfillResult checks a photo backfill job's result, which useJob
+// passes on unread.
+export function parseBackfillResult(where: string, body: unknown): BackfillResult {
+  if (!isRecord(body) || !Array.isArray(body.dinners)) throw new Error(`${where}: result is not a photo backfill`)
+  return {
+    dinners: body.dinners.map((dinner): BackfillDinner => {
+      if (!isRecord(dinner) || typeof dinner.dinner_id !== 'string' || (dinner.reason !== null && typeof dinner.reason !== 'string')) {
+        throw new Error(`${where}: result has a dinner that is not a backfill outcome`)
+      }
+      return {
+        dinner_id: dinner.dinner_id,
+        outcome: oneOf(where, 'outcome', BACKFILL_OUTCOMES, dinner.outcome),
+        reason: dinner.reason,
+      }
+    }),
+  }
+}
+
+// PhotoBackfill is a photo backfill started from this screen, as it goes.
+export type PhotoBackfill = {
+  // start asks the server for a backfill over every dinner without a
+  // photo. While one is already queued or running the server answers with
+  // that job, so starting never runs two at once.
+  start: () => void
+  // active is true from start until the job ends, except while a poll's
+  // failure leaves its progress unknown: starting again then asks the
+  // server for the job it is running and polls it anew.
+  active: boolean
+  // Why the backfill could not be started.
+  startError: Error | null
+  // The job's status, undefined until it is started and first polled.
+  status: JobStatus | undefined
+  // The job's result, once it has one.
+  result: BackfillResult | null
+  // The job's failure, a poll's failure, or a result that is not a backfill.
+  error: Error | null
+}
+
+// usePhotoBackfill starts a photo backfill (POST
+// /api/v1/settings/photo-backfill, 202 {id}) and follows its job with
+// useJob. When the job ends, every query showing a dinner's photo is read
+// again, so dinners it grounded show their photo everywhere.
+export function usePhotoBackfill(): PhotoBackfill {
+  const { apiFetch } = useApi()
+  const queryClient = useQueryClient()
+  const [jobId, setJobId] = useState<string | null>(null)
+  const job = useJob<unknown>(BACKFILL_JOB_KIND, jobId)
+  const begin = useMutation({
+    mutationFn: async (): Promise<string> => {
+      const body = await apiFetch<unknown>(PHOTO_BACKFILL_PATH, { method: 'POST' })
+      if (!isRecord(body) || typeof body.id !== 'string') throw new Error(`POST ${PHOTO_BACKFILL_PATH}: response has no job id`)
+      return body.id
+    },
+    onSuccess: (id) => {
+      setJobId(id)
+      job.poll()
+    },
+  })
+
+  const ended = job.status === 'succeeded' || job.status === 'failed'
+  useEffect(() => {
+    if (!ended) return
+    for (const queryKey of PHOTO_KEYS) void queryClient.invalidateQueries({ queryKey })
+  }, [ended, jobId, queryClient])
+
+  const parsed = useMemo(() => {
+    if (job.result === null) return { result: null, error: null }
+    try {
+      return { result: parseBackfillResult(`GET /api/v1/jobs/${jobId}`, job.result), error: null }
+    } catch (error) {
+      return { result: null, error: error as Error }
+    }
+  }, [job.result, jobId])
+
+  return {
+    start: () => begin.mutate(),
+    active: begin.isPending || (jobId !== null && !ended && job.error === null),
+    startError: begin.error,
+    status: job.status,
+    result: parsed.result,
+    error: job.error ?? parsed.error,
+  }
+}

web/src/shopping/offlineQueue.test.tsx

@@ -1,5 +1,5 @@
-import { defaultScheduler, notifyManager, QueryClientProvider } from '@tanstack/react-query'
-import { act, fireEvent, render, screen, within } from '@testing-library/react'
+import { QueryClientProvider } from '@tanstack/react-query'
+import { act, fireEvent, render, screen, waitFor, within } from '@testing-library/react'
 import { MemoryRouter } from 'react-router'
 import { afterEach, beforeEach, describe, expect, it, vi, type Mock, type MockInstance } from 'vitest'
 import { ApiProvider, createApiClient, createQueryClient, createVerifiedSession } from '../api/index.ts'
@@ -213,46 +213,8 @@ function reload(page: { unmount: () => void }) {
   vi.setSystemTime(now)
 }
 
-// advance jumps the clock past a deadline or a wait, firing the timers due
-// by then; what they set off runs as later turns (until, settle) let it.
-function advance(ms: number) {
-  act(() => vi.advanceTimersByTime(ms))
-}
-
-// The host's setImmediate, which these tests do not fake: it runs its
-// callback on the next turn of the event loop, once every promise already
-// resolved has run, whatever the clock says.
-const nextTurn = (globalThis as unknown as { setImmediate: (callback: () => void) => void }).setImmediate
-
-// turn lets the page take one turn of the event loop, inside act: the
-// promises it has resolved run, and React renders and runs the effects they
-// cause. No time passes, real or fake.
-async function turn() {
-  await act(() => new Promise<void>((resolve) => nextTurn(resolve)))
-}
-
-// TURNS bounds how many turns until waits and settle takes: far more than
-// any change here needs to reach the page.
-const TURNS = 50
-
-// until lets the page take turns until check passes, and returns what check
-// returns; it fails with check's error if check still fails after TURNS.
-async function until<T>(check: () => T): Promise<T> {
-  for (let turns = 0; ; turns++) {
-    try {
-      return check()
-    } catch (error) {
-      if (turns === TURNS) throw error
-    }
-    await turn()
-  }
-}
-
-// settle lets the page take TURNS turns, the clock standing still: whatever
-// it would do without a deadline or a wait passing, such as send a PATCH too
-// soon, it has done.
-async function settle() {
-  for (let turns = 0; turns < TURNS; turns++) await turn()
+async function advance(ms: number) {
+  await act(() => vi.advanceTimersByTimeAsync(ms))
 }
 
 // The page's setTimeout, watched: retrySecondTry waits for a timer to be set.
@@ -262,8 +224,8 @@ let timers: MockInstance<typeof setTimeout>
 // doubled by its failed first try — once that wait's timer is set, so the jump
 // never lands before the timer it is meant to fire.
 async function retrySecondTry() {
-  await until(() => expect(timers).toHaveBeenCalledWith(expect.any(Function), 2 * RETRY_FIRST))
-  advance(2 * RETRY_FIRST)
+  await waitFor(() => expect(timers).toHaveBeenCalledWith(expect.any(Function), 2 * RETRY_FIRST))
+  await advance(2 * RETRY_FIRST)
 }
 
 function checkbox(name: string) {
@@ -278,21 +240,16 @@ beforeEach(() => {
   onLine = true
   reachable = true
   mockViewport(390)
-  // The clock and setTimeout move only when a test jumps past a deadline or
-  // a wait (advance), never with real time: under load, real time would fire
-  // a deadline or a wait the test never reached. React Query sends its
-  // batched notices a turn later, as it does in the app on a setTimeout of 0,
-  // so the page renders its changes without the clock moving.
-  vi.useFakeTimers({ toFake: ['Date', 'setTimeout', 'clearTimeout'] })
+  // setTimeout moves with real time, and a test can jump past a deadline or
+  // a wait (advance).
+  vi.useFakeTimers({ toFake: ['Date', 'setTimeout', 'clearTimeout'], shouldAdvanceTime: true })
   vi.setSystemTime(new Date('2026-10-07T16:00:00Z'))
-  notifyManager.setScheduler(nextTurn)
   timers = vi.spyOn(globalThis, 'setTimeout')
 })
 
 afterEach(() => {
   vi.restoreAllMocks()
   vi.useRealTimers()
-  notifyManager.setScheduler(defaultScheduler)
   vi.unstubAllGlobals()
   Reflect.deleteProperty(window, 'matchMedia')
 })
@@ -301,7 +258,7 @@ describe('offline toggle queue', () => {
   it('applies toggles made offline at once, saves them in order and sends them in that order once online', async () => {
     const { fetchMock } = stubServer()
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     goOffline()
     fireEvent.click(checkbox('Taco seasoning'))
@@ -319,7 +276,7 @@ describe('offline toggle queue', () => {
     expect(patches(fetchMock)).toEqual([])
 
     goOnline()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(fetchMock)).toEqual([
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i1')} {"needed":false}`,
@@ -335,14 +292,14 @@ describe('offline toggle queue', () => {
   it('queues a toggle whose PATCH fails with a network error and sends it again once back online', async () => {
     const { fetchMock } = stubServer()
     await openPage()
-    fireEvent.click(await until(() => screen.getByRole('checkbox', { name: 'Taco seasoning' })))
-    await until(() => expect(queue()).toEqual([]))
+    fireEvent.click(await screen.findByRole('checkbox', { name: 'Taco seasoning' }))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(screen.getByText('3 / 3')).toBeInTheDocument()
 
     // The browser still says online, but the server cannot be reached.
     reachable = false
     fireEvent.click(checkbox('Corn tortillas'))
-    expect(await until(() => within(row('Corn tortillas')).getByText('will sync'))).toBeInTheDocument()
+    expect(await within(row('Corn tortillas')).findByText('will sync')).toBeInTheDocument()
     expect(patches(fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`, `${ITEM('i2')} {"needed":false}`])
     expect(checkbox('Corn tortillas')).not.toBeChecked()
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
@@ -350,7 +307,7 @@ describe('offline toggle queue', () => {
     expect(screen.queryByRole('alert')).not.toBeInTheDocument()
 
     goOnline()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`, `${ITEM('i2')} {"needed":false}`, `${ITEM('i2')} {"needed":false}`])
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(screen.queryByText('will sync')).not.toBeInTheDocument()
@@ -359,7 +316,7 @@ describe('offline toggle queue', () => {
   it('opens the saved list after a reload offline and replays the saved toggles in order once online', async () => {
     const { fetchMock } = stubServer()
     const first = await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
     goOffline()
     fireEvent.click(checkbox('Taco seasoning'))
     fireEvent.click(checkbox('Ground beef'))
@@ -368,7 +325,7 @@ describe('offline toggle queue', () => {
     // Reload while offline: the settings, /me and the list cannot be read.
     reload(first)
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
     expect(checkbox('Taco seasoning')).toBeChecked()
     expect(checkbox('Ground beef')).not.toBeChecked()
     expect(within(row('Taco seasoning')).getByText('will sync')).toBeInTheDocument()
@@ -376,9 +333,9 @@ describe('offline toggle queue', () => {
     expect(patches(fetchMock)).toEqual([])
 
     goOnline()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`, `${ITEM('i1')} {"needed":false}`])
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(checkbox('Taco seasoning')).toBeChecked()
     expect(checkbox('Ground beef')).not.toBeChecked()
@@ -387,7 +344,7 @@ describe('offline toggle queue', () => {
   it('discards a queued toggle the server refuses because the list was archived, puts the count back and says so', async () => {
     const server = stubServer(() => errorResponse(409, 'archived'))
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     goOffline()
     fireEvent.click(checkbox('Taco seasoning'))
@@ -395,25 +352,25 @@ describe('offline toggle queue', () => {
 
     server.archive()
     goOnline()
-    const alert = await until(() => screen.getByRole('alert'))
+    const alert = await screen.findByRole('alert')
     expect(alert).toHaveTextContent("Couldn't mark Taco seasoning to buy. archived message")
-    expect(await until(() => screen.getByText('Archived'))).toBeInTheDocument()
+    expect(await screen.findByText('Archived')).toBeInTheDocument()
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(within(row('Taco seasoning')).getByText('have it')).toBeInTheDocument()
     expect(screen.queryByText('will sync')).not.toBeInTheDocument()
     expect(queue()).toEqual([])
     // Sent once, never again.
     goOnline()
-    await until(() => expect(patches(server.fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`]))
+    await waitFor(() => expect(patches(server.fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`]))
   })
 
   it('does not queue a toggle the server refuses: it is put back and says why', async () => {
     const server = stubServer(() => errorResponse(400, 'invalid'))
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     fireEvent.click(checkbox('Ground beef'))
-    expect(await until(() => screen.getByRole('alert'))).toHaveTextContent("Couldn't mark Ground beef “have it”. invalid message")
+    expect(await screen.findByRole('alert')).toHaveTextContent("Couldn't mark Ground beef “have it”. invalid message")
     expect(checkbox('Ground beef')).toBeChecked()
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(screen.queryByText('will sync')).not.toBeInTheDocument()
@@ -421,7 +378,7 @@ describe('offline toggle queue', () => {
 
     goOffline()
     goOnline()
-    await until(() => expect(patches(server.fetchMock)).toEqual([`${ITEM('i1')} {"needed":false}`]))
+    await waitFor(() => expect(patches(server.fetchMock)).toEqual([`${ITEM('i1')} {"needed":false}`]))
   })
 })
 
@@ -438,16 +395,16 @@ describe('offline toggle queue, replay', () => {
     localStorage.setItem(SAVED_KEY, JSON.stringify({ schedule: null, runs, queue: queued }))
 
     await openPage()
-    await until(() => expect(answers.waiting()).toBe(1))
-    fireEvent.click(await until(() => screen.getByRole('checkbox', { name: 'Corn tortillas' })))
+    await waitFor(() => expect(answers.waiting()).toBe(1))
+    fireEvent.click(await screen.findByRole('checkbox', { name: 'Corn tortillas' }))
     for (let sent = 1; sent <= 4; sent++) {
-      await until(() => expect(patches(server.fetchMock)).toHaveLength(sent))
+      await waitFor(() => expect(patches(server.fetchMock)).toHaveLength(sent))
       // The next one is not sent before this one is answered.
-      await settle()
+      await new Promise((resolve) => setTimeout(resolve, 20))
       expect(patches(server.fetchMock)).toHaveLength(sent)
       act(() => answers.release())
     }
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(server.fetchMock)).toEqual([
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('n1')} {"needed":false}`,
@@ -455,27 +412,27 @@ describe('offline toggle queue, replay', () => {
       `${ITEM('i2')} {"needed":false}`,
     ])
     expect(server.mostInFlight()).toBe(1)
-    await until(() => expect(screen.getByText('1 / 3')).toBeInTheDocument())
+    await waitFor(() => expect(screen.getByText('1 / 3')).toBeInTheDocument())
   })
 
   it('replays at least once: a toggle the server applied before a reload is sent again and leaves the item and count the same', async () => {
     const answers = held()
     const server = stubServer(undefined, answers.hold)
     const first = await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
     fireEvent.click(checkbox('Taco seasoning'))
     // The server applied it; the page reloads before its answer arrives.
-    await until(() => expect(answers.waiting()).toBe(1))
+    await waitFor(() => expect(answers.waiting()).toBe(1))
     expect(queue()).toEqual(['i3 true'])
     reload(first)
     act(() => answers.release())
 
     await openPage()
-    await until(() => expect(answers.waiting()).toBe(1))
+    await waitFor(() => expect(answers.waiting()).toBe(1))
     act(() => answers.release())
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(server.fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`, `${ITEM('i3')} {"needed":true}`])
-    await until(() => expect(screen.getByText('3 / 3')).toBeInTheDocument())
+    await waitFor(() => expect(screen.getByText('3 / 3')).toBeInTheDocument())
     expect(checkbox('Taco seasoning')).toBeChecked()
     expect(server.run('2026-10-05')).toMatchObject({ needed_count: 3, total_count: 3 })
     expect(screen.queryByRole('alert')).not.toBeInTheDocument()
@@ -485,18 +442,18 @@ describe('offline toggle queue, replay', () => {
     const answers = held()
     stubServer(undefined, answers.hold)
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
     goOffline()
     fireEvent.click(checkbox('Ground beef'))
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
 
     goOnline()
-    await until(() => expect(answers.waiting()).toBe(1))
+    await waitFor(() => expect(answers.waiting()).toBe(1))
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
     expect(queue()).toEqual(['i1 false'])
 
     act(() => answers.release())
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(queue()).toEqual([])
     expect(screen.getByText('1 / 3')).toBeInTheDocument()
   })
@@ -506,29 +463,29 @@ describe('offline toggle queue, replay', () => {
     // The first PATCH never answers; every later one answers at once.
     const server = stubServer(undefined, () => (++tries === 1 ? new Promise<void>(() => {}) : Promise.resolve()))
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     fireEvent.click(checkbox('Taco seasoning'))
-    await until(() => expect(patches(server.fetchMock)).toHaveLength(1))
+    await waitFor(() => expect(patches(server.fetchMock)).toHaveLength(1))
     fireEvent.click(checkbox('Ground beef'))
     expect(queue()).toEqual(['i3 true', 'i1 false'])
 
     // Past the deadline the hung PATCH is aborted; both still wait, in order.
-    advance(TOGGLE_DEADLINE)
-    expect(await until(() => within(row('Taco seasoning')).getByText('will sync'))).toBeInTheDocument()
+    await advance(TOGGLE_DEADLINE)
+    expect(await within(row('Taco seasoning')).findByText('will sync')).toBeInTheDocument()
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
     expect(patches(server.fetchMock)).toHaveLength(1)
     expect(queue()).toEqual(['i3 true', 'i1 false'])
 
     await retrySecondTry()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(server.fetchMock)).toEqual([
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i1')} {"needed":false}`,
     ])
     expect(server.mostInFlight()).toBe(1)
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(screen.queryByRole('alert')).not.toBeInTheDocument()
   })
@@ -539,26 +496,26 @@ describe('offline toggle queue, replay', () => {
     // answers. Every later /me answers at once.
     const server = stubServer(undefined, undefined, () => (++tries === 2 ? new Promise<void>(() => {}) : Promise.resolve()))
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     fireEvent.click(checkbox('Taco seasoning'))
-    await until(() => expect(tries).toBe(2))
+    await waitFor(() => expect(tries).toBe(2))
     fireEvent.click(checkbox('Ground beef'))
     expect(queue()).toEqual(['i3 true', 'i1 false'])
 
     // Past the deadline the hung /me is abandoned; both still wait, in order.
-    advance(TOGGLE_DEADLINE)
-    expect(await until(() => within(row('Taco seasoning')).getByText('will sync'))).toBeInTheDocument()
+    await advance(TOGGLE_DEADLINE)
+    expect(await within(row('Taco seasoning')).findByText('will sync')).toBeInTheDocument()
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
     expect(patches(server.fetchMock)).toEqual([])
 
     // The next try sends a fresh /me, which the server answers.
     await retrySecondTry()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(mes(server.fetchMock)).toBe(3)
     expect(patches(server.fetchMock)).toEqual([`${ITEM('i3')} {"needed":true}`, `${ITEM('i1')} {"needed":false}`])
     expect(server.mostInFlight()).toBe(1)
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(screen.queryByRole('alert')).not.toBeInTheDocument()
   })
@@ -579,29 +536,29 @@ describe('offline toggle queue, replay', () => {
       () => (refusedAt !== null && ++tries === 1 ? new Promise<void>(() => {}) : Promise.resolve()),
     )
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     fireEvent.click(checkbox('Taco seasoning'))
-    await until(() => expect(tries).toBe(1))
+    await waitFor(() => expect(tries).toBe(1))
     fireEvent.click(checkbox('Ground beef'))
     expect(patches(server.fetchMock)).toHaveLength(1)
 
     // Past the deadline the hung refresh is abandoned; both still wait, in order.
-    advance(TOGGLE_DEADLINE)
-    expect(await until(() => within(row('Taco seasoning')).getByText('will sync'))).toBeInTheDocument()
+    await advance(TOGGLE_DEADLINE)
+    expect(await within(row('Taco seasoning')).findByText('will sync')).toBeInTheDocument()
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
     expect(queue()).toEqual(['i3 true', 'i1 false'])
 
     // The next try is refused again and sends a fresh refresh, which answers.
     await retrySecondTry()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     expect(patches(server.fetchMock)).toEqual([
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i3')} {"needed":true}`,
       `${ITEM('i1')} {"needed":false}`,
     ])
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(screen.getByText('2 / 3')).toBeInTheDocument()
     expect(screen.queryByRole('alert')).not.toBeInTheDocument()
   })
@@ -609,21 +566,21 @@ describe('offline toggle queue, replay', () => {
   it('tries a toggle again once the server is back, with no online event from the browser', async () => {
     const { fetchMock } = stubServer()
     await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
 
     // The browser says online throughout; the server cannot be reached, then can.
     reachable = false
     fireEvent.click(checkbox('Corn tortillas'))
-    expect(await until(() => within(row('Corn tortillas')).getByText('will sync'))).toBeInTheDocument()
+    expect(await within(row('Corn tortillas')).findByText('will sync')).toBeInTheDocument()
     reachable = true
     expect(patches(fetchMock)).toEqual([])
 
     // Nothing says the server is back: only the wait's timer tries it again.
     await retrySecondTry()
-    await until(() => expect(queue()).toEqual([]))
+    await waitFor(() => expect(queue()).toEqual([]))
     // The first try failed at GET /api/v1/me, before its PATCH.
     expect(patches(fetchMock)).toEqual([`${ITEM('i2')} {"needed":false}`])
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(screen.getByText('1 / 3')).toBeInTheDocument()
   })
 
@@ -631,20 +588,20 @@ describe('offline toggle queue, replay', () => {
     const answers = held()
     stubServer(undefined, answers.hold)
     const page = await openPage()
-    expect(await until(() => screen.getByText('2 / 3'))).toBeInTheDocument()
+    expect(await screen.findByText('2 / 3')).toBeInTheDocument()
     goOffline()
     fireEvent.click(checkbox('Ground beef'))
     expect(within(row('Ground beef')).getByText('will sync')).toBeInTheDocument()
 
     page.leave()
     goOnline()
-    await until(() => expect(answers.waiting()).toBe(1))
+    await waitFor(() => expect(answers.waiting()).toBe(1))
     page.reopen()
-    expect(await until(() => within(row('Ground beef')).getByText('will sync'))).toBeInTheDocument()
+    expect(await within(row('Ground beef')).findByText('will sync')).toBeInTheDocument()
     expect(queue()).toEqual(['i1 false'])
 
     act(() => answers.release())
-    await until(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
+    await waitFor(() => expect(screen.queryByText('will sync')).not.toBeInTheDocument())
     expect(queue()).toEqual([])
     expect(screen.getByText('1 / 3')).toBeInTheDocument()
   })
@@ -656,10 +613,10 @@ describe('offline toggle queue, replay', () => {
     const getItem = vi.spyOn(Storage.prototype, 'getItem')
     await openPage()
 
-    expect(await until(() => screen.getByText(/Couldn't load the household settings/))).toBeInTheDocument()
+    expect(await screen.findByText(/Couldn't load the household settings/)).toBeInTheDocument()
     expect(screen.getByText(/can't be kept on this device while offline/)).toBeInTheDocument()
     const reads = getItem.mock.calls.filter(([key]) => key === SAVED_KEY).length
-    await settle()
+    await new Promise((resolve) => setTimeout(resolve, 50))
     expect(getItem.mock.calls.filter(([key]) => key === SAVED_KEY)).toHaveLength(reads)
   })
 })

web/src/states.test.tsx

@@ -1,601 +0,0 @@
-import { QueryClientProvider } from '@tanstack/react-query'
-import { fireEvent, render, screen, waitFor, within } from '@testing-library/react'
-import type { ReactNode } from 'react'
-import { Route, Routes } from 'react-router'
-import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest'
-import { ApiProvider, createApiClient, createQueryClient, createVerifiedSession } from './api/index.ts'
-import { DinnerDetailPage } from './dinners/DinnerDetailPage.tsx'
-import { DiscoverPage } from './discover/DiscoverPage.tsx'
-import { HomePage } from './home/HomePage.tsx'
-import { CookFromPantryPage } from './ideas/CookFromPantryPage.tsx'
-import { QuickDinnersPage } from './ideas/QuickDinnersPage.tsx'
-import { OnboardingPage } from './onboarding/OnboardingPage.tsx'
-import { createNetworkStatus } from './pwa/networkStatus.ts'
-import { PlanPage } from './plan/PlanPage.tsx'
-import { defaultWeek } from './plan/week.ts'
-import { AISettingsPage } from './settings/AISettingsPage.tsx'
-import { CalendarSettingsPage } from './settings/CalendarSettingsPage.tsx'
-import { HouseholdSettingsPage } from './settings/HouseholdSettingsPage.tsx'
-import { NotificationsPage } from './settings/NotificationsPage.tsx'
-import { RegularsSettingsPage } from './settings/RegularsSettingsPage.tsx'
-import { SettingsPage } from './settings/SettingsPage.tsx'
-import { SupplyArchivePage } from './shopping/SupplyArchivePage.tsx'
-import { SupplyRunPage } from './shopping/SupplyRunPage.tsx'
-import {
-  AI_MODELS,
-  AI_SETTINGS,
-  CHAT_WEBHOOK_NONE,
-  GOOGLE_NOT_CONNECTED,
-  HOME,
-  HOUSEHOLD_SETTINGS,
-  NOTIFICATION_SETTINGS,
-  RECIPE_SEARCH_SETTINGS,
-  SETTINGS_API,
-} from './test/ApiProviders.tsx'
-import { InDataRouter } from './test/dataRouter.tsx'
-import { dinnerBody } from './test/dinners.ts'
-import { errorResponse, fakeLocation, jsonResponse, ME } from './test/fetchStub.ts'
-import { mockViewport } from './test/matchMedia.ts'
-import skeletonStyles from './ui/Skeleton.module.css'
-
-// The empty, loading and error state of every route, as docs/STATES.md
-// lists them: each route renders its loading, empty and error variants
-// against a stubbed API.
-
-type Responder = () => Response | Promise<Response>
-
-// pending is a response that never comes, so the page stays loading.
-const pending: Responder = () => new Promise<Response>(() => {})
-
-// failThen answers the first request with a 500, and every later one with
-// then: Retry reads again and gets it.
-function failThen(then: Responder): Responder {
-  let failed = false
-  return () => {
-    if (failed) return then()
-    failed = true
-    return errorResponse(500, 'internal')
-  }
-}
-
-const json = (body: unknown): Responder => () => jsonResponse(200, body)
-
-// startJobs answers each POST that starts a job with the next job id, j1,
-// j2 and so on.
-function startJobs(): Responder {
-  let n = 0
-  return () => jsonResponse(202, { jobId: `j${++n}` })
-}
-
-// serve answers "METHOD path" requests from responders, with /me and the
-// household settings; any other request fails the test.
-function serve(responders: Record<string, Responder>) {
-  const all: Record<string, Responder> = {
-    'GET /api/v1/me': json(ME),
-    'GET /api/v1/settings/household': json(HOUSEHOLD_SETTINGS),
-    ...responders,
-  }
-  const fetchMock = vi.fn<typeof fetch>((input, init) => {
-    const key = `${init?.method ?? 'GET'} ${String(input)}`
-    const respond = Object.hasOwn(all, key) ? all[key] : undefined
-    if (respond === undefined) throw new Error(`unexpected fetch: ${key}`)
-    return Promise.resolve(respond())
-  })
-  vi.stubGlobal('fetch', fetchMock)
-  return fetchMock
-}
-
-// renderRoute renders element on its route pattern at path, under a data
-// router as App renders it, once /me has loaded as AuthBootstrap loads it.
-async function renderRoute(path: string, pattern: string, element: ReactNode) {
-  const api = createApiClient(fakeLocation(path), createNetworkStatus(window), createVerifiedSession(localStorage))
-  await api.loadMe()
-  return render(
-    <ApiProvider client={api}>
-      <QueryClientProvider client={createQueryClient()}>
-        <InDataRouter path={path}>
-          <Routes>
-            <Route path={pattern} element={element} />
-          </Routes>
-        </InDataRouter>
-      </QueryClientProvider>
-    </ApiProvider>,
-  )
-}
-
-function skeletons() {
-  return document.querySelectorAll(`.${skeletonStyles.skeleton}`)
-}
-
-// expectLoading waits for the pill skeletons and checks nothing failed.
-async function expectLoading() {
-  await waitFor(() => expect(skeletons().length).toBeGreaterThan(0))
-  expect(screen.queryByRole('alert')).not.toBeInTheDocument()
-}
-
-// expectOnlyLoadingBefore checks, once a page's independent sections have
-// settled, that the one skeleton left is the held read's: it sits before
-// next, the first settled section after it, and nothing failed.
-function expectOnlyLoadingBefore(next: Element) {
-  expect(skeletons()).toHaveLength(1)
-  const left = skeletons().item(0)
-  if (left === null) throw new Error('no skeleton left')
-  expect(next.contains(left)).toBe(false)
-  expect(left.compareDocumentPosition(next) & Node.DOCUMENT_POSITION_FOLLOWING).toBeTruthy()
-  expect(screen.queryByRole('alert')).not.toBeInTheDocument()
-}
-
-// expectRetry finds the error banner saying said, presses its Retry and
-// waits for recovered.
-async function expectRetry(said: string, recovered: () => Promise<unknown>) {
-  const alert = await screen.findByRole('alert')
-  expect(alert).toHaveTextContent(said)
-  fireEvent.click(within(alert).getByRole('button', { name: 'Retry' }))
-  await recovered()
-  await waitFor(() => expect(screen.queryByRole('alert')).not.toBeInTheDocument())
-}
-
-const WEEK = () => defaultWeek(HOUSEHOLD_SETTINGS, new Date())
-
-beforeEach(() => {
-  mockViewport(390)
-})
-
-afterEach(() => {
-  Reflect.deleteProperty(window, 'matchMedia')
-  vi.unstubAllGlobals()
-})
-
-describe('/ Home', () => {
-  const NOTHING_TONIGHT = { ...HOME, tonight: null, sideQuestCount: 0, recentUpdates: { unread: false } }
-  const page = async (home: Responder) => {
-    serve({ 'GET /api/v1/home': home, 'GET /api/v1/rotation-reviews': json({ reviews: [] }) })
-    await renderRoute('/', '/', <HomePage />)
-  }
-
-  it('loading: skeletons', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: no quest planned tonight, with Plan this week', async () => {
-    await page(json(NOTHING_TONIGHT))
-    expect(await screen.findByRole('heading', { name: 'No quest planned tonight' })).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Plan this week' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json(NOTHING_TONIGHT)))
-    await expectRetry("Couldn't load tonight's quest. internal message", () =>
-      screen.findByRole('heading', { name: 'No quest planned tonight' }),
-    )
-  })
-})
-
-describe('/onboarding Welcome step', () => {
-  it('reads nothing, so it has no empty, loading or error state', async () => {
-    const fetchMock = serve({})
-    await renderRoute('/onboarding', '/onboarding/*', <OnboardingPage />)
-    expect(screen.getByRole('heading', { name: 'Welcome to MealQuest' })).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: "Let's start" })).toBeEnabled()
-    expect(skeletons()).toHaveLength(0)
-    expect(fetchMock.mock.calls.map(([input]) => String(input))).toEqual(['/api/v1/me'])
-  })
-})
-
-describe('/onboarding/regulars Regulars step', () => {
-  const page = async (regulars: Responder) => {
-    serve({ 'GET /api/v1/dinners?status=regular': regulars })
-    await renderRoute('/onboarding/regulars', '/onboarding/*', <OnboardingPage />)
-  }
-
-  it('loading: skeleton', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: no Regulars yet, common picks offered and Next disabled', async () => {
-    await page(json({ dinners: [] }))
-    expect(await screen.findByRole('heading', { name: 'Common picks' })).toBeInTheDocument()
-    expect(screen.getByRole('heading', { name: 'Your regulars · 0' })).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Next: notifications' })).toBeDisabled()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json({ dinners: [] })))
-    await expectRetry('Your Regulars could not load: internal message', () =>
-      screen.findByRole('heading', { name: 'Common picks' }),
-    )
-  })
-})
-
-describe('/onboarding/notifications Notifications step', () => {
-  const page = async (finish: Responder) => {
-    serve({ 'PATCH /api/v1/settings/household': finish })
-    await renderRoute('/onboarding/notifications', '/onboarding/*', <OnboardingPage />)
-    fireEvent.click(screen.getByRole('button', { name: 'Maybe later' }))
-  }
-
-  it("loading: the step's buttons wait while it finishes", async () => {
-    await page(pending)
-    await waitFor(() => expect(screen.getByRole('button', { name: 'Maybe later' })).toBeDisabled())
-    expect(screen.getByRole('button', { name: 'Turn on notifications' })).toBeDisabled()
-    expect(screen.queryByRole('alert')).not.toBeInTheDocument()
-  })
-
-  it('error: says finishing failed and lets it be tried again', async () => {
-    await page(failThen(pending))
-    expect(await screen.findByRole('alert')).toHaveTextContent('Could not finish setting up: internal message')
-    expect(screen.getByRole('button', { name: 'Maybe later' })).toBeEnabled()
-  })
-})
-
-describe('/plan Draft board', () => {
-  const page = async (plan: Responder) => {
-    serve({ [`GET /api/v1/plans/${WEEK()}`]: plan, [`GET /api/v1/shopping/${WEEK()}`]: () => errorResponse(404, 'not_found') })
-    await renderRoute('/plan', '/plan', <PlanPage />)
-  }
-  const noPlan = () => errorResponse(404, 'not_found')
-
-  it('loading: skeleton tiles', async () => {
-    await page(pending)
-    expect(await screen.findByRole('region', { name: 'Plan loading' })).toBeInTheDocument()
-    await expectLoading()
-  })
-
-  it('empty: nothing planned yet, with Draft this week', async () => {
-    await page(noPlan)
-    expect(await screen.findByRole('heading', { name: 'Nothing planned yet' })).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Draft this week' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(noPlan))
-    await expectRetry("Couldn't load this week's plan. internal message", () =>
-      screen.findByRole('heading', { name: 'Nothing planned yet' }),
-    )
-  })
-})
-
-describe('/plan/shopping Supply run', () => {
-  const page = async (run: Responder) => {
-    serve({ [`GET /api/v1/shopping/${WEEK()}`]: run })
-    await renderRoute('/plan/shopping', '/plan/shopping', <SupplyRunPage />)
-  }
-  const noList = () => errorResponse(404, 'not_found')
-
-  it('loading: skeleton cards', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: no shopping list yet, with Make the shopping list', async () => {
-    await page(noList)
-    expect(await screen.findByRole('heading', { name: 'No shopping list yet' })).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Make the shopping list' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(noList))
-    await expectRetry("Couldn't load the shopping list. internal message", () =>
-      screen.findByRole('heading', { name: 'No shopping list yet' }),
-    )
-  })
-})
-
-describe('/plan/shopping/archive Supply archive', () => {
-  const page = async (archive: Responder) => {
-    serve({ 'GET /api/v1/shopping/archive?by=week': archive })
-    await renderRoute('/plan/shopping/archive', '/plan/shopping/archive', <SupplyArchivePage />)
-  }
-  const nothing = json({ by: 'week', periods: [] })
-
-  it('loading: skeleton posts', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: nothing archived yet', async () => {
-    await page(nothing)
-    expect(await screen.findByRole('heading', { name: 'Nothing archived yet' })).toBeInTheDocument()
-    expect(screen.getByText("Each week's Supply run becomes a post here once it's made.")).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(nothing))
-    await expectRetry("Couldn't load the Supply archive. internal message", () =>
-      screen.findByRole('heading', { name: 'Nothing archived yet' }),
-    )
-  })
-})
-
-describe('/discover Side quests', () => {
-  const page = async (deck: Responder) => {
-    serve({ 'GET /api/v1/suggestions/deck': deck, 'GET /api/v1/dinners?status=quest_log': json({ dinners: [] }) })
-    await renderRoute('/discover', '/discover', <DiscoverPage />)
-  }
-  const empty = json({ suggestions: [], active_batch: null, failed_batch: null })
-
-  it('loading: skeleton cards', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: All caught up / New side quests arrive Sunday., with Get more now', async () => {
-    await page(empty)
-    expect(await screen.findByRole('heading', { name: 'All caught up' })).toBeInTheDocument()
-    expect(screen.getByText('New side quests arrive Sunday.')).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Get more now' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(empty))
-    await expectRetry("Couldn't load your side quests. internal message", () =>
-      screen.findByRole('heading', { name: 'All caught up' }),
-    )
-  })
-})
-
-describe('/quick Quick dinners', () => {
-  const QUICK = 'POST /api/v1/ideas/quick'
-  const job = (status: string, error: string | null) => ({
-    id: 'j1',
-    kind: 'quick_ideas',
-    status,
-    result: { stage: 'grounding', kind: 'under30', exclude: [], ideas: [], dropped: 0 },
-    error,
-  })
-  const page = async (quick: Responder, poll: Responder = pending) => {
-    serve({ [QUICK]: quick, 'GET /api/v1/jobs/j1': poll, 'GET /api/v1/jobs/j2': pending })
-    await renderRoute('/quick', '/quick', <QuickDinnersPage />)
-  }
-
-  it('loading: skeleton rows', async () => {
-    await page(pending)
-    await waitFor(() => expect(screen.getAllByTestId('skeleton-row')).toHaveLength(6))
-    await expectLoading()
-  })
-
-  it('empty: no ideas found, with Retry', async () => {
-    await page(startJobs(), json(job('failed', 'no_ideas_found')))
-    await expectRetry("Couldn't find Quick dinners. No ideas came back with a recipe to match.", () =>
-      waitFor(() => expect(skeletons().length).toBeGreaterThan(0)),
-    )
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(pending))
-    await expectRetry("Couldn't start looking for Quick dinners. internal message", () =>
-      waitFor(() => expect(screen.getAllByTestId('skeleton-row')).toHaveLength(6)),
-    )
-  })
-})
-
-describe('/cook-from-pantry Cook from what I have', () => {
-  const FIND = 'POST /api/v1/ideas/from-ingredients'
-  const job = {
-    id: 'j1',
-    kind: 'ideas_from_ingredients',
-    status: 'failed',
-    result: { stage: 'grounding', ingredients: ['Rice'], quick_only: false, ideas: [], dropped: 0 },
-    error: 'no_ideas_found',
-  }
-  const page = async (find: Responder, poll: Responder = pending) => {
-    serve({ [FIND]: find, 'GET /api/v1/jobs/j1': poll, 'GET /api/v1/jobs/j2': pending })
-    await renderRoute('/cook-from-pantry', '/cook-from-pantry', <CookFromPantryPage />)
-  }
-  const search = () => {
-    fireEvent.change(screen.getByRole('textbox', { name: 'Add ingredient' }), { target: { value: 'Rice' } })
-    fireEvent.click(screen.getByRole('button', { name: 'Find ideas' }))
-  }
-
-  it('loading: skeleton cards', async () => {
-    await page(pending)
-    search()
-    expect(await screen.findByRole('region', { name: 'Ideas loading' })).toBeInTheDocument()
-    await expectLoading()
-  })
-
-  it('empty: no ingredients yet, with Find ideas disabled; no ideas found, with Retry', async () => {
-    await page(startJobs(), json(job))
-    expect(screen.getByRole('button', { name: 'Find ideas' })).toBeDisabled()
-    search()
-    await expectRetry("Couldn't find ideas. No ideas came back with a recipe to match.", () =>
-      screen.findByRole('region', { name: 'Ideas' }),
-    )
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(pending))
-    search()
-    await expectRetry("Couldn't start looking for ideas. internal message", () =>
-      screen.findByRole('region', { name: 'Ideas loading' }),
-    )
-  })
-})
-
-describe('/dinner/:id Dinner', () => {
-  const page = async (dinner: Responder) => {
-    serve({ 'GET /api/v1/dinners/d1': dinner })
-    await renderRoute('/dinner/d1', '/dinner/:id', <DinnerDetailPage />)
-  }
-  const notFound = () => errorResponse(404, 'not_found')
-
-  it('loading: skeletons', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: dinner not found', async () => {
-    await page(notFound)
-    expect(await screen.findByRole('heading', { name: 'Dinner not found' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(notFound))
-    await expectRetry("Couldn't load this dinner. internal message", () =>
-      screen.findByRole('heading', { name: 'Dinner not found' }),
-    )
-  })
-})
-
-describe('/settings Settings home', () => {
-  const page = async (calendar: Responder) => {
-    const api = Object.fromEntries(Object.entries(SETTINGS_API).map(([path, respond]) => [`GET ${path}`, respond]))
-    serve({ ...api, 'GET /api/v1/dinners?status=regular': json({ dinners: [] }), 'GET /api/v1/integrations/google': calendar })
-    await renderRoute('/settings', '/settings', <SettingsPage />)
-  }
-  const notConnected = json(GOOGLE_NOT_CONNECTED)
-
-  it('loading: status skeletons', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: no Regulars, calendar not connected, Google Chat not connected', async () => {
-    await page(notConnected)
-    expect(await screen.findByText('None on rotation yet')).toBeInTheDocument()
-    expect(await screen.findByText('Not connected')).toBeInTheDocument()
-    expect(await screen.findByText(/Google Chat not connected/)).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(notConnected))
-    await expectRetry('Settings could not load: internal message', () => screen.findByText('Not connected'))
-  })
-})
-
-describe('/settings/notifications Notifications', () => {
-  const page = async (prefs: Responder) => {
-    serve({ 'GET /api/v1/settings/notifications': prefs, 'GET /api/v1/settings/chat-webhook': json(CHAT_WEBHOOK_NONE) })
-    await renderRoute('/settings/notifications', '/settings/notifications', <NotificationsPage />)
-  }
-
-  it('loading: skeleton for the channel matrix', async () => {
-    await page(pending)
-    // The webhook card and this device's push status read on their own;
-    // once they settle, the one skeleton left is the channel matrix's.
-    await screen.findByRole('button', { name: 'Save webhook' })
-    const device = screen.getByRole('group', { name: 'Notifications on this device' })
-    await waitFor(() => expect(device.querySelectorAll(`.${skeletonStyles.skeleton}`)).toHaveLength(0))
-    expectOnlyLoadingBefore(device)
-    expect(screen.queryByRole('group', { name: 'Notification channels' })).not.toBeInTheDocument()
-  })
-
-  it('empty: no Google Chat webhook stored', async () => {
-    await page(json(NOTIFICATION_SETTINGS))
-    expect(await screen.findByRole('button', { name: 'Save webhook' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json(NOTIFICATION_SETTINGS)))
-    await expectRetry('Notification settings could not load: internal message', () =>
-      screen.findByRole('group', { name: 'Notification channels' }),
-    )
-  })
-})
-
-describe('/settings/ai AI assistant', () => {
-  const NO_KEY = { ...AI_SETTINGS, api_key_last4: null, key_verified_at: null }
-  const page = async (settings: Responder) => {
-    serve({
-      'GET /api/v1/settings/ai': settings,
-      'GET /api/v1/settings/ai/models': json(AI_MODELS),
-      'GET /api/v1/settings/recipe-search': json(RECIPE_SEARCH_SETTINGS),
-    })
-    await renderRoute('/settings/ai', '/settings/ai', <AISettingsPage />)
-  }
-
-  it('loading: skeleton for the AI assistant form', async () => {
-    await page(pending)
-    // The Recipe search card reads on its own; once it settles, the one
-    // skeleton left is the AI assistant form's.
-    await screen.findByRole('button', { name: 'Save recipe search' })
-    expectOnlyLoadingBefore(screen.getByRole('region', { name: 'Recipe search' }))
-    expect(screen.queryByText('No key stored')).not.toBeInTheDocument()
-  })
-
-  it('empty: no key stored, with Add key', async () => {
-    await page(json(NO_KEY))
-    expect(await screen.findByText('No key stored')).toBeInTheDocument()
-    expect(screen.getByRole('button', { name: 'Add key' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json(NO_KEY)))
-    await expectRetry('AI settings could not load: internal message', () => screen.findByText('No key stored'))
-  })
-})
-
-describe('/settings/household Household', () => {
-  const page = async (settings: Responder) => {
-    serve({ 'GET /api/v1/settings/household': settings })
-    await renderRoute('/settings/household', '/settings/household', <HouseholdSettingsPage />)
-  }
-
-  it('loading: skeleton', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  // The household's settings always exist (seeded on first start), so its
-  // "empty" state is the seeded defaults.
-  it('empty: the seeded settings', async () => {
-    await page(json(HOUSEHOLD_SETTINGS))
-    expect(await screen.findByLabelText('Dinner time')).toHaveValue('17:30')
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json(HOUSEHOLD_SETTINGS)))
-    await expectRetry('Household settings could not load: internal message', () => screen.findByLabelText('Dinner time'))
-  })
-})
-
-describe('/settings/regulars Regular dinners', () => {
-  const page = async (regulars: Responder) => {
-    serve({ 'GET /api/v1/dinners?status=regular': regulars })
-    await renderRoute('/settings/regulars', '/settings/regulars', <RegularsSettingsPage />)
-  }
-  const none = json({ dinners: [] })
-
-  it('loading: skeleton', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: no Regulars yet', async () => {
-    await page(none)
-    expect(await screen.findByText('No Regulars yet. Add the dinners your household makes on repeat.')).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(json({ dinners: [dinnerBody('d1', 'Beef tacos')] })))
-    await expectRetry('Your Regulars could not load: internal message', () => screen.findByText('Beef tacos'))
-  })
-})
-
-describe('/settings/calendar Calendar', () => {
-  const page = async (connection: Responder) => {
-    serve({ 'GET /api/v1/integrations/google': connection })
-    await renderRoute('/settings/calendar', '/settings/calendar', <CalendarSettingsPage />)
-  }
-  const notConnected = json(GOOGLE_NOT_CONNECTED)
-
-  it('loading: skeleton', async () => {
-    await page(pending)
-    await expectLoading()
-  })
-
-  it('empty: not connected, with Connect Google Calendar', async () => {
-    await page(notConnected)
-    expect(await screen.findByRole('heading', { name: 'Not connected' })).toBeInTheDocument()
-    expect(screen.getByRole('link', { name: 'Connect Google Calendar' })).toBeInTheDocument()
-  })
-
-  it('error: banner with Retry', async () => {
-    await page(failThen(notConnected))
-    await expectRetry('Google Calendar status could not load: internal message', () =>
-      screen.findByRole('heading', { name: 'Not connected' }),
-    )
-  })
-})

web/src/test/dinners.ts

@@ -13,6 +13,7 @@ export type DinnerBody = {
   photo_url: string | null
   grounding_status: string
   grounding_error: string | null
+  photo_backfill_eligible: boolean
   updated_by: string | null
   created_at: string
   updated_at: string
@@ -35,6 +36,7 @@ export function dinnerBody(id: string, name: string, overrides: Partial<DinnerBo
     photo_url: null,
     grounding_status: 'grounded',
     grounding_error: null,
+    photo_backfill_eligible: false,
     updated_by: null,
     created_at: '2026-10-01T15:00:00.000Z',
     updated_at: '2026-10-01T15:00:00.000Z',