Merge pull request 'Graceful shutdown + resilience & Explore polish (m147)' (#203) from dev into main

translation/status now reports `active` (a translate/retranslate sweep is
running, from the TaskRun table) and `last_run` (the most recent finished run's
task + status). The Settings card polls live while a sweep runs, showing a
spinner + "Translating… N remaining" that ticks down, and flags a last run that
ended in error/timeout. No migration.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

backend/app/api/gallery.py

@@ -145,6 +145,9 @@ async def similar():
         filters, _sort = _parse_filters()
     except (KeyError, ValueError):
         return jsonify({"error": "similar_to query param required"}), 400
+    # Explore passes exclude_wip=1 to also drop work-in-progress from the
+    # rabbit-hole; the gallery's own "similar" button omits it (keeps wip, #1274).
+    exclude_wip = request.args.get("exclude_wip") in ("1", "true", "True")
     # post_id is the exclusive post-detail view — not a similarity scope.
     # include_hidden is a gallery-browse flag; similar() has its OWN presentation
     # exclusion (a similarity-quality concern, #1274), so drop it here (#141).
@@ -154,7 +157,8 @@ async def similar():
     async with get_session() as session:
         svc = GalleryService(session)
         try:
-            images = await svc.similar(image_id=similar_to, limit=limit, **scope)
+            images = await svc.similar(
+                image_id=similar_to, limit=limit, exclude_wip=exclude_wip, **scope)
         except ValueError as exc:
             return jsonify({"error": str(exc)}), 400
     if images is None:

backend/app/api/settings.py

@@ -16,6 +16,7 @@ from ..models import (
     ImportTask,
     Post,
     Tag,
+    TaskRun,
 )
 from ..services import interpreter_client as ic
 
@@ -306,6 +307,10 @@ async def translation_status():
     """For the Settings card: is it on, is a URL set, is the service reachable,
     and how many posts still await translation. Health runs the sync client in a
     thread so the event loop isn't blocked."""
+    translation_tasks = (
+        "backend.app.tasks.translation.translate_posts",
+        "backend.app.tasks.translation.retranslate_posts",
+    )
     async with get_session() as session:
         cfg = await ImportSettings.load(session)
         untranslated = (await session.execute(
@@ -315,6 +320,21 @@ async def translation_status():
                 Post.post_title.is_not(None), Post.description.is_not(None),
             ))
         )).scalar_one()
+        # Live progress: is a sweep running now, and what did the last one do?
+        # (run-until-done re-enqueues itself, so `active` stays true across a
+        # bulk re-translate; `last_run` surfaces a completed run's outcome.)
+        active = (await session.execute(
+            select(func.count(TaskRun.id))
+            .where(TaskRun.task_name.in_(translation_tasks))
+            .where(TaskRun.status == "running")
+        )).scalar_one()
+        last = (await session.execute(
+            select(TaskRun.task_name, TaskRun.status, TaskRun.finished_at)
+            .where(TaskRun.task_name.in_(translation_tasks))
+            .where(TaskRun.finished_at.is_not(None))
+            .order_by(TaskRun.finished_at.desc())
+            .limit(1)
+        )).first()
     base_url = (cfg.interpreter_base_url or "").strip()
     healthy = await asyncio.to_thread(ic.health, base_url) if base_url else False
     return jsonify({
@@ -322,6 +342,12 @@ async def translation_status():
         "base_url_set": bool(base_url),
         "healthy": healthy,
         "untranslated_count": int(untranslated),
+        "active": int(active) > 0,
+        "last_run": {
+            "task": last[0].rsplit(".", 1)[-1],
+            "status": last[1],
+            "finished_at": last[2].isoformat() if last[2] else None,
+        } if last else None,
     })
 
 

backend/app/celery_app.py

@@ -70,6 +70,14 @@ def make_celery() -> Celery:
         },
         # Heavy ML tasks need fair dispatch — see ImageRepo's precedent.
         task_acks_late=True,
+        # Deploy graceful-shutdown safety: with acks_late, a task killed because
+        # it outran the container's stop-grace window (SIGKILL) is re-queued
+        # rather than silently lost. Safe because our long tasks are idempotent +
+        # chunked (translation per-post commit, downloads terminal-status, audits
+        # chunk) and the 5-min recovery sweeps re-drive anything left non-terminal
+        # — a re-run resumes cleanly and never corrupts. No redeliver-loop risk:
+        # heavy GPU work is tombstoned via gpu_queue, not run inline in a worker.
+        task_reject_on_worker_lost=True,
         worker_prefetch_multiplier=1,
         # Broker resilience (2026-06-24): a swarm overlay-network blip after a
         # redeploy left Redis healthy but transiently unreachable, and a worker
@@ -107,6 +115,11 @@ def make_celery() -> Celery:
                 "task": "backend.app.tasks.maintenance.cleanup_old_tasks",
                 "schedule": 86400.0,  # daily
             },
+            "cleanup-orphaned-temp-files": {
+                "task": "backend.app.tasks.maintenance.cleanup_orphaned_temp_files",
+                "schedule": 86400.0,  # daily — sweep .part/.partial left by a
+                # download/import killed mid-write (graceful-shutdown fallout)
+            },
             "train-heads-nightly": {
                 "task": "backend.app.tasks.ml.scheduled_train_heads",
                 "schedule": 86400.0,  # passive cadence; manual retrain stays available

backend/app/models/tag.py

@@ -48,6 +48,10 @@ class TagKind(StrEnum):
 # content. `wip` is real art: only the training pipelines exclude it.
 SYSTEM_TAG_NAMES = ("wip", "banner", "editor screenshot")
 PRESENTATION_SYSTEM_TAGS = ("banner", "editor screenshot")
+# `wip` marks real-but-unfinished art. It's kept in the gallery's own "similar"
+# results (#1274), but the Explore rabbit-hole opts to hide it (exclude_wip) so a
+# browse doesn't keep surfacing work-in-progress (operator, 2026-07-08).
+WIP_SYSTEM_TAG = "wip"
 
 image_tag = Table(
     "image_tag",

backend/app/services/gallery_service.py

@@ -31,7 +31,7 @@ from ..models import (
     Tag,
     TagPositiveConfirmation,
 )
-from ..models.tag import PRESENTATION_SYSTEM_TAGS, image_tag
+from ..models.tag import PRESENTATION_SYSTEM_TAGS, WIP_SYSTEM_TAG, image_tag
 from .pagination import decode_cursor, encode_cursor
 from .tag_query import (
     fandom_join_alias,
@@ -715,6 +715,7 @@ class GalleryService:
         platform: str | None = None,
         untagged: bool = False, no_artist: bool = False,
         date_from: datetime | None = None, date_to: datetime | None = None,
+        exclude_wip: bool = False,
     ) -> list[GalleryImage] | None:
         """Visual "more like this": images near `image_id`'s SigLIP embedding
         (pgvector, HNSW-indexed — alembic 0036), then DIVERSIFIED so the result
@@ -751,14 +752,18 @@ class GalleryService:
         # Presentation images (banner / editor-screenshot system tags, #128)
         # cluster on UI chrome rather than content, so near any one of them
         # they'd fill the grid. Excluded from CANDIDATES only — the anchor
-        # itself may be a banner, and `wip` stays surfaced (real art; only
-        # the training pipelines exclude it).
+        # itself may be a banner. `wip` stays surfaced here by default (real art;
+        # only the training pipelines exclude it), but the Explore rabbit-hole
+        # passes exclude_wip to also drop work-in-progress (operator, 2026-07-08).
+        excluded_system_tags = PRESENTATION_SYSTEM_TAGS
+        if exclude_wip:
+            excluded_system_tags = (*PRESENTATION_SYSTEM_TAGS, WIP_SYSTEM_TAG)
         presentation = (
             select(image_tag.c.image_record_id)
             .join(Tag, Tag.id == image_tag.c.tag_id)
             .where(
                 Tag.is_system.is_(True),
-                Tag.name.in_(PRESENTATION_SYSTEM_TAGS),
+                Tag.name.in_(excluded_system_tags),
             )
         )
         stmt = stmt.where(

backend/app/services/interpreter_client.py

@@ -8,12 +8,24 @@ behind a reverse proxy). Full API contract: Scribe note #1347.
 """
 from __future__ import annotations
 
+from datetime import UTC, datetime
+from email.utils import parsedate_to_datetime
+
 import requests
+from requests.adapters import HTTPAdapter
+from urllib3.util.retry import Retry
 
 
 class InterpreterUnavailable(Exception):
-    """The translation engine is down / unreachable (HTTP 503 or a connection
-    error) — retry later, don't drop the item."""
+    """The translation engine is down / unreachable / draining — a connection
+    error, HTTP 429, or a 5xx (commonly 502/503/504 through a reverse proxy while
+    the service restarts). Retry later, don't drop the item. ``retry_after`` holds
+    the server's Retry-After hint in seconds when it sent one, so the caller can
+    back off exactly as long as it's asked to."""
+
+    def __init__(self, message: str, *, retry_after: float | None = None):
+        super().__init__(message)
+        self.retry_after = retry_after
 
 
 class InterpreterBadRequest(Exception):
@@ -24,13 +36,52 @@ def _url(base_url: str, path: str) -> str:
     return f"{base_url.rstrip('/')}{path}"
 
 
+def _parse_retry_after(resp) -> float | None:
+    """Parse a Retry-After header (RFC 7231: delta-seconds or an HTTP-date) into
+    non-negative seconds, or None if absent/unparseable — so a gracefully-
+    draining Interpreter can tell Curator exactly how long to wait before it
+    tries again."""
+    raw = (resp.headers.get("Retry-After") or "").strip()
+    if not raw:
+        return None
+    try:
+        return max(0.0, float(int(raw)))       # delta-seconds form
+    except ValueError:
+        pass
+    try:                                        # HTTP-date form
+        when = parsedate_to_datetime(raw)
+    except (TypeError, ValueError):
+        return None
+    if when is None:
+        return None
+    if when.tzinfo is None:
+        when = when.replace(tzinfo=UTC)
+    return max(0.0, (when - datetime.now(UTC)).total_seconds())
+
+
+# A shared session pools the keep-alive connection across the per-post sweep
+# calls, and retries CONNECT failures only (connect=2, short backoff) — smoothing
+# the instant a reverse proxy reloads. Status codes are deliberately NOT retried
+# (status=0, raise_on_status=False): translate() maps 429/5xx → InterpreterUnavailable
+# itself, and letting urllib3 retry a draining 503 would defeat the Retry-After
+# backoff we honour upstream.
+_retry = Retry(
+    total=None, connect=2, read=0, redirect=0, status=0,
+    backoff_factor=0.3, raise_on_status=False,
+)
+session = requests.Session()
+_adapter = HTTPAdapter(max_retries=_retry)
+session.mount("http://", _adapter)
+session.mount("https://", _adapter)
+
+
 def health(base_url: str, *, timeout: float = 5.0) -> bool:
     """True iff the Interpreter LLM engine is up. Any error (unset URL, network,
     non-200, engine down) → False, so the sweep just no-ops rather than raising."""
     if not base_url:
         return False
     try:
-        r = requests.get(_url(base_url, "/v1/health"), timeout=timeout)
+        r = session.get(_url(base_url, "/v1/health"), timeout=timeout)
     except requests.RequestException:
         return False
     if r.status_code != 200:
@@ -53,14 +104,15 @@ def translate(
     Interpreter batch metadata is aggregate (first item only) — fine for one
     post's ``[title, description]`` batch since they share a language. Passthrough
     items (already target-language / emoji-only) come back UNCHANGED in their
-    slot. Raises InterpreterUnavailable on 503 / connection error (retry later),
-    InterpreterBadRequest on 400. (Scribe note #1347.)
+    slot. Raises InterpreterUnavailable on a connection error / 429 / 5xx (retry
+    later, honouring Retry-After), InterpreterBadRequest on 400. (Scribe note
+    #1347.)
     """
     if not texts:
         return {"translations": [], "detected_lang": None,
                 "engine": None, "engine_version": None}
     try:
-        r = requests.post(
+        r = session.post(
             _url(base_url, "/v1/translate"),
             json={"q": list(texts), "source": source,
                   "target": target, "engine": "auto"},
@@ -68,10 +120,18 @@ def translate(
         )
     except requests.RequestException as e:
         raise InterpreterUnavailable(str(e)) from e
-    if r.status_code == 503:
-        raise InterpreterUnavailable("interpreter engine unavailable")
     if r.status_code == 400:
         raise InterpreterBadRequest((r.json() or {}).get("error", "bad request"))
+    # A gracefully-draining service — often behind a reverse proxy — returns 429
+    # or a 5xx gateway error (502/503/504), not always a clean 503. Treat every
+    # one as "unavailable, retry later" and honour any Retry-After it sends, so a
+    # rolling Interpreter restart interrupts the sweep cleanly (instead of raising
+    # an opaque HTTPError) and resumes exactly when the service says it's ready.
+    if r.status_code == 429 or r.status_code >= 500:
+        raise InterpreterUnavailable(
+            f"interpreter unavailable (HTTP {r.status_code})",
+            retry_after=_parse_retry_after(r),
+        )
     r.raise_for_status()
     body = r.json() or {}
     out = body.get("translatedText")

backend/app/tasks/maintenance.py

@@ -79,6 +79,14 @@ VERIFY_PAGE = 200
 FFPROBE_TIMEOUT_SECONDS = 10
 TASK_RUN_KEEP_OK_SECONDS = 24 * 3600       # 24 h
 TASK_RUN_KEEP_FAILURE_SECONDS = 7 * 24 * 3600  # 7 days
+# Orphaned staging files: downloads/imports stage into <name>.part / <name>.partial
+# then os.replace() into place (importer / external_fetch / native_ingest_common /
+# attachment_store), so a kill mid-write leaves a discardable temp, never a corrupt
+# final. cleanup_orphaned_temp_files sweeps ones left behind; the min-age guard
+# keeps it from deleting an in-flight download's staging file mid-write.
+IMAGES_ROOT = Path("/images")
+TEMP_STAGING_SUFFIXES = (".part", ".partial")
+ORPHAN_TEMP_MIN_AGE_HOURS = 6
 
 # Audit 2026-06-02: per-entity recovery sweep thresholds. Each must be
 # > the entity's longest legitimate runtime (its task's time_limit + a
@@ -339,6 +347,34 @@ def cleanup_old_tasks() -> int:
         return result.rowcount or 0
 
 
+@celery.task(name="backend.app.tasks.maintenance.cleanup_orphaned_temp_files")
+def cleanup_orphaned_temp_files() -> int:
+    """Delete orphaned .part/.partial staging files under the images root, left by
+    a download/import killed mid-write. Only removes files older than
+    ORPHAN_TEMP_MIN_AGE_HOURS so an in-flight download's staging file is never
+    pulled out from under it. Returns the count removed."""
+    if not IMAGES_ROOT.is_dir():
+        return 0
+    cutoff = datetime.now(UTC).timestamp() - ORPHAN_TEMP_MIN_AGE_HOURS * 3600
+    removed = 0
+    for path in IMAGES_ROOT.rglob("*"):
+        if path.suffix not in TEMP_STAGING_SUFFIXES or not path.is_file():
+            continue
+        try:
+            if path.stat().st_mtime >= cutoff:
+                continue          # too fresh — may be an active download
+            path.unlink()
+            removed += 1
+        except OSError as exc:
+            log.warning("cleanup_orphaned_temp_files: %s: %s", path, exc)
+    if removed:
+        log.info(
+            "cleanup_orphaned_temp_files: removed %d orphaned staging file(s)",
+            removed,
+        )
+    return removed
+
+
 @celery.task(name="backend.app.tasks.maintenance.recover_stalled_task_runs")
 def recover_stalled_task_runs() -> int:
     """Flip task_run rows stuck in 'running' past their queue-specific

backend/app/tasks/translation.py

@@ -36,6 +36,20 @@ _MAX_POSTS_PER_RUN = 300
 # Short gap between run-until-done chunks so a big re-translate finishes on its
 # own without monopolising the maintenance_long worker in one uninterrupted run.
 _RETRANSLATE_COUNTDOWN = 5
+# When Interpreter drains mid-chunk (graceful restart), resume the bulk
+# re-translate after the service's Retry-After hint if it sent one, else this
+# default; capped so a bogus/huge hint can't park the work longer than the daily
+# beat's own safety net would.
+_INTERRUPT_BACKOFF = 60
+_INTERRUPT_BACKOFF_MAX = 900
+
+
+def _interrupt_backoff(retry_after) -> int:
+    """Seconds to wait before resuming after an Interpreter interruption: the
+    server's Retry-After hint (capped), else the default."""
+    if retry_after and retry_after > 0:
+        return int(min(retry_after, _INTERRUPT_BACKOFF_MAX))
+    return _INTERRUPT_BACKOFF
 
 
 @celery.task(
@@ -56,7 +70,21 @@ def translate_posts() -> str:
             return ready
         base_url, target = ready
         posts = _select_untranslated(session, None, _MAX_POSTS_PER_RUN)
-        status, translated = _translate_batch(session, posts, base_url, target)
+        status, translated, retry_after = _translate_batch(
+            session, posts, base_url, target,
+        )
+        # Steady-state daily sweep: one chunk per run on success (the beat drives
+        # the rest — no tail-chase). But if Interpreter drained mid-chunk, don't
+        # make a manual "Translate now" wait a whole day — re-enqueue after its
+        # Retry-After hint (or the default backoff). Self-terminating: once the
+        # service is down the health gate returns "interpreter unavailable" early.
+        if status == "interrupted" and _count_untranslated(session, None):
+            delay = _interrupt_backoff(retry_after)
+            translate_posts.apply_async((), {}, countdown=delay)
+            log.info(
+                "translate_posts: interrupted (service draining) → retry in %ss",
+                delay,
+            )
     return _summary(status, translated, len(posts))
 
 
@@ -88,11 +116,11 @@ def retranslate_posts(artist_ids=None, _reset_done=False) -> str:
             )
 
         posts = _select_untranslated(session, artist_ids, _MAX_POSTS_PER_RUN)
-        status, translated = _translate_batch(session, posts, base_url, target)
+        status, translated, retry_after = _translate_batch(
+            session, posts, base_url, target,
+        )
 
-        # run-until-done: only chase the tail when the chunk finished cleanly. A
-        # service interruption / bad request stops here and leaves the rest NULL
-        # for the daily beat or a fresh trigger to resume (rule-89 recovery).
+        # run-until-done: chase the tail when the chunk finished cleanly.
         # Termination is guaranteed: every handled post leaves the NULL set, so
         # the remaining count strictly decreases each chunk.
         if status == "ok":
@@ -106,6 +134,23 @@ def retranslate_posts(artist_ids=None, _reset_done=False) -> str:
                 log.info(
                     "retranslate_posts: %d remaining → re-enqueued", remaining,
                 )
+        # Interpreter drained mid-chunk (graceful restart / 429 / 5xx): don't
+        # stall the bulk re-translate until the daily beat — resume it after the
+        # service's Retry-After hint (or the default backoff), with
+        # _reset_done=True so the fresh work is never re-wiped. Self-terminating:
+        # if the service is still down next run, _translation_config's health gate
+        # returns "interpreter unavailable" early and nothing re-enqueues.
+        elif status == "interrupted" and _count_untranslated(session, artist_ids):
+            delay = _interrupt_backoff(retry_after)
+            retranslate_posts.apply_async(
+                (),
+                {"artist_ids": artist_ids, "_reset_done": True},
+                countdown=delay,
+            )
+            log.info(
+                "retranslate_posts: interrupted (service draining) → retry in %ss",
+                delay,
+            )
     return _summary(status, translated, len(posts))
 
 
@@ -174,23 +219,25 @@ def _reset_translations(session, artist_ids) -> int:
 
 def _translate_batch(session, posts, base_url: str, target: str):
     """Translate each post in the chunk with a per-post commit (an interrupted
-    run keeps progress). Returns (status, translated) where status is one of
-    "ok" / "interrupted" (503/conn) / "stopped" (400) / "timeout"."""
+    run keeps progress). Returns (status, translated, retry_after) where status is
+    one of "ok" / "interrupted" (unavailable/drain) / "stopped" (400) / "timeout";
+    retry_after is the server's Retry-After hint in seconds on an interrupt, else
+    None."""
     translated = 0
     for post in posts:
         try:
             translated += _translate_one(session, post, base_url, target)
             session.commit()
-        except ic.InterpreterUnavailable:
+        except ic.InterpreterUnavailable as e:
             session.rollback()
-            return "interrupted", translated
+            return "interrupted", translated, getattr(e, "retry_after", None)
         except ic.InterpreterBadRequest as e:
             session.rollback()
             log.warning("translate bad request: %s", e)
-            return "stopped", translated
+            return "stopped", translated, None
         except SoftTimeLimitExceeded:
-            return "timeout", translated
-    return "ok", translated
+            return "timeout", translated, None
+    return "ok", translated, None
 
 
 def _summary(status: str, translated: int, scanned: int) -> str:

docker-compose.yml

@@ -8,6 +8,35 @@
 # run `docker compose up` from this directory and switches images to
 # local builds + DEBUG logging.
 
+# Rolling-deploy safety (Swarm / `docker stack deploy`): update one task at a
+# time, START the new task before stopping the old (zero-downtime via the ingress
+# mesh), and if the new task doesn't reach a healthy state within `monitor`, roll
+# back to the previous image automatically. `monitor` is sized above web's
+# healthcheck start_period so a broken image that never goes healthy is caught.
+# Plain `docker compose up` ignores `deploy:` (it warns + skips), so the dev
+# override is unaffected. Referenced by each long-lived service below.
+x-deploy-policy: &deploy_policy
+  update_config:
+    order: start-first
+    failure_action: rollback
+    monitor: 90s
+  rollback_config:
+    order: start-first
+  restart_policy:
+    condition: any
+    delay: 10s
+
+# Worker liveness: ping THIS container's celery node over the broker. Lenient
+# (60s interval, 3 retries, 60s start_period) so a transient broker blip never
+# false-flags a worker into a rollback. `$$HOSTNAME` → `$HOSTNAME` for the shell;
+# celery's default node name is celery@<hostname> (the container id).
+x-celery-healthcheck: &celery_healthcheck
+  test: ["CMD-SHELL", "celery -A backend.app.celery_app:celery inspect ping -d celery@$$HOSTNAME --timeout 10 >/dev/null 2>&1"]
+  interval: 60s
+  timeout: 15s
+  retries: 3
+  start_period: 60s
+
 services:
   redis:
     image: redis:7-alpine
@@ -47,6 +76,24 @@ services:
   web:
     image: git.fabledsword.com/bvandeusen/fabledcurator:dev
     command: ["web"]
+    # Graceful shutdown: give the container time to drain in-flight work on a
+    # deploy (docker SIGTERMs, then SIGKILLs after this window — default is only
+    # 10s, far too short for real jobs). Hypercorn/Celery both warm-shut-down on
+    # SIGTERM; per-lane values sized to typical task length. Anything that still
+    # outruns the window is re-queued (task_reject_on_worker_lost) and re-driven
+    # by the 5-min recovery sweeps, so a kill never corrupts. web = short HTTP
+    # requests + the occasional file download.
+    stop_grace_period: 30s
+    # Liveness for rolling deploys: /api/health is a no-DB 200 (just proves the
+    # app booted + serves HTTP after `alembic upgrade head`). start_period covers
+    # the migration + boot so a slow start isn't mis-flagged.
+    healthcheck:
+      test: ["CMD-SHELL", "python -c \"import urllib.request,sys; sys.exit(0 if urllib.request.urlopen('http://localhost:8080/api/health', timeout=5).status==200 else 1)\""]
+      interval: 15s
+      timeout: 6s
+      retries: 3
+      start_period: 40s
+    deploy: *deploy_policy
     ports:
       - "${PORT:-8080}:8080"
     environment: &app_env
@@ -77,6 +124,10 @@ services:
   worker:
     image: git.fabledsword.com/bvandeusen/fabledcurator:dev
     command: ["worker"]
+    # Drain in-flight import/thumbnail/download tasks before SIGKILL on deploy.
+    stop_grace_period: 90s
+    healthcheck: *celery_healthcheck
+    deploy: *deploy_policy
     environment:
       <<: *app_env
       CELERY_QUEUES: default,import,thumbnail,download
@@ -93,6 +144,10 @@ services:
   scheduler:
     image: git.fabledsword.com/bvandeusen/fabledcurator:dev
     command: ["scheduler"]
+    # Quick maintenance/scan lane + beat — short tasks, modest drain window.
+    stop_grace_period: 60s
+    healthcheck: *celery_healthcheck
+    deploy: *deploy_policy
     environment:
       <<: *app_env
       CELERY_QUEUES: maintenance,scan
@@ -110,6 +165,12 @@ services:
   maintenance-long:
     image: git.fabledsword.com/bvandeusen/fabledcurator:dev
     command: ["worker"]
+    # Longest lane (DB backups, library audits, translation backfill) — give it
+    # the most room to finish a chunk gracefully. Chunked + idempotent, so a job
+    # that still outruns this resumes cleanly next run rather than corrupting.
+    stop_grace_period: 180s
+    healthcheck: *celery_healthcheck
+    deploy: *deploy_policy
     environment:
       <<: *app_env
       CELERY_QUEUES: maintenance_long
@@ -125,6 +186,10 @@ services:
   ml-worker:
     image: git.fabledsword.com/bvandeusen/fabledcurator-ml:dev
     command: ["ml-worker"]
+    # A single GPU inference pass can run tens of seconds — let it finish.
+    stop_grace_period: 120s
+    healthcheck: *celery_healthcheck
+    deploy: *deploy_policy
     environment:
       <<: *app_env
     volumes:

frontend/src/components/settings/TranslationCard.vue

@@ -52,11 +52,25 @@
         prepend-icon="mdi-refresh" :loading="retranslating"
         :disabled="!enabled || !baseUrl" @click="confirmAll = true"
       >Re-translate all</v-btn>
-      <span v-if="status" class="fc-muted text-caption">
+      <span
+        v-if="status && status.active"
+        class="fc-muted text-caption d-inline-flex align-center" style="gap: 6px;"
+      >
+        <v-progress-circular indeterminate size="12" width="2" color="accent" />
+        Translating… {{ status.untranslated_count }} remaining
+      </span>
+      <span v-else-if="status" class="fc-muted text-caption">
         {{ status.untranslated_count }}
         post{{ status.untranslated_count === 1 ? '' : 's' }} awaiting translation
       </span>
     </div>
+    <p
+      v-if="status && status.last_run && ['error', 'timeout'].includes(status.last_run.status)"
+      class="text-caption text-error mt-1 mb-0"
+    >
+      Last translation run ended with “{{ status.last_run.status }}”. It resumes on
+      the next sweep; check the Interpreter connection if it persists.
+    </p>
     <p class="fc-muted text-caption mt-2 mb-0">
       Use <strong>Re-translate all</strong> after switching the Interpreter model —
       it clears every stored translation and re-runs it through the new model
@@ -94,7 +108,7 @@
 </template>
 
 <script setup>
-import { computed, onMounted, ref } from 'vue'
+import { computed, onMounted, onUnmounted, ref } from 'vue'
 import { toast } from '../../utils/toast.js'
 import { useApi } from '../../composables/useApi.js'
 import { useImportStore } from '../../stores/import.js'
@@ -126,10 +140,16 @@ const statusText = computed(() => {
   return status.value.healthy ? 'Interpreter reachable' : 'Interpreter unreachable'
 })
 
+let pollTimer = null
 async function loadStatus() {
   try { status.value = await api.get('/api/settings/translation/status') }
   catch { status.value = null }
+  // Poll live while a sweep is running so the remaining count ticks down; stop
+  // when idle (a fresh trigger restarts it via its own setTimeout(loadStatus)).
+  clearTimeout(pollTimer)
+  if (status.value?.active) pollTimer = setTimeout(loadStatus, 3000)
 }
+onUnmounted(() => clearTimeout(pollTimer))
 
 async function onTest() {
   // Ping /v1/health for the CURRENTLY-typed URL (not the saved one), so a new

frontend/src/stores/explore.js

@@ -44,7 +44,9 @@ export const useExploreStore = defineStore('explore', () => {
       // empty and let the view explain why (anchor.has_embedding === false).
       if (detail.has_embedding) {
         const body = await api.get('/api/gallery/similar', {
-          params: { similar_to: numId, limit: NEIGHBOR_LIMIT },
+          // exclude_wip: keep work-in-progress out of the Explore rabbit-hole
+          // (the gallery's own "similar" button still shows it) — operator 2026-07-08.
+          params: { similar_to: numId, limit: NEIGHBOR_LIMIT, exclude_wip: 1 },
         })
         if (!t.isCurrent()) return
         neighbors.value = body.images || []

tests/test_api_settings.py

@@ -61,6 +61,34 @@ async def test_translation_status_defaults(client):
     assert body["base_url_set"] is False
     assert body["healthy"] is False
     assert "untranslated_count" in body
+    assert body["active"] is False      # no sweep running
+    assert body["last_run"] is None
+
+
+@pytest.mark.asyncio
+async def test_translation_status_reports_active_and_last_run(client, db):
+    # A running translate/retranslate TaskRun → active True; the most recent
+    # finished one → last_run (task basename + status).
+    from datetime import UTC, datetime
+
+    from backend.app.models import TaskRun
+
+    db.add(TaskRun(
+        celery_task_id="r-run", queue="maintenance_long",
+        task_name="backend.app.tasks.translation.retranslate_posts",
+        started_at=datetime.now(UTC), status="running",
+    ))
+    db.add(TaskRun(
+        celery_task_id="r-done", queue="maintenance_long",
+        task_name="backend.app.tasks.translation.translate_posts",
+        started_at=datetime.now(UTC), finished_at=datetime.now(UTC),
+        status="success",
+    ))
+    await db.commit()
+    body = await (await client.get("/api/settings/translation/status")).get_json()
+    assert body["active"] is True
+    assert body["last_run"]["task"] == "translate_posts"
+    assert body["last_run"]["status"] == "success"
 
 
 @pytest.mark.asyncio

tests/test_gallery_similar.py

@@ -98,6 +98,30 @@ async def test_similar_excludes_presentation_tagged_images(db):
     assert {i.id for i in res_from_banner} == {src.id, wipped.id, plain.id}
 
 
+@pytest.mark.asyncio
+async def test_similar_exclude_wip_drops_wip_neighbors(db):
+    """Explore passes exclude_wip=True to also hide work-in-progress from the
+    rabbit-hole (banner is always hidden; wip only when asked)."""
+    src = await _img(db, 1, _vec(1, 0))
+    bannered = await _img(db, 2, _vec(1, 0.02))   # always hidden
+    wipped = await _img(db, 3, _vec(1, 0.3))      # hidden only with exclude_wip
+    plain = await _img(db, 4, _vec(1, 0.6))
+    banner_tag = (await db.execute(select(Tag).where(
+        Tag.is_system.is_(True), Tag.name == "banner"))).scalar_one()
+    wip_tag = (await db.execute(select(Tag).where(
+        Tag.is_system.is_(True), Tag.name == "wip"))).scalar_one()
+    await db.execute(image_tag.insert().values(
+        image_record_id=bannered.id, tag_id=banner_tag.id, source="manual"))
+    await db.execute(image_tag.insert().values(
+        image_record_id=wipped.id, tag_id=wip_tag.id, source="manual"))
+    svc = GalleryService(db)
+    res = await svc.similar(src.id, limit=10, exclude_wip=True)
+    assert [i.id for i in res] == [plain.id]        # wip + banner both gone
+    # Default (gallery "similar" button) still keeps wip (#1274).
+    res_default = await svc.similar(src.id, limit=10)
+    assert wipped.id in {i.id for i in res_default}
+
+
 @pytest.mark.asyncio
 async def test_similar_composes_with_tag_filter(db):
     src = await _img(db, 1, _vec(1, 0))

tests/test_interpreter_client.py

@@ -6,9 +6,10 @@ from backend.app.services import interpreter_client as ic
 
 
 class _Resp:
-    def __init__(self, status_code=200, payload=None):
+    def __init__(self, status_code=200, payload=None, headers=None):
         self.status_code = status_code
         self._payload = payload or {}
+        self.headers = headers or {}
 
     def json(self):
         return self._payload
@@ -29,7 +30,7 @@ def test_translate_maps_batch_in_order(monkeypatch):
             "interpreter": {"engine": "llm", "engine_version": "ollama:x:12b"},
         })
 
-    monkeypatch.setattr(ic.requests, "post", fake_post)
+    monkeypatch.setattr(ic.session, "post", fake_post)
     out = ic.translate(["ねこが可愛い", "金髪ギャル！"], base_url="http://i.lan")
     assert out["translations"] == ["The cat is cute", "Blonde gal!"]
     assert out["detected_lang"] == "ja"
@@ -40,7 +41,7 @@ def test_translate_maps_batch_in_order(monkeypatch):
 
 def test_translate_passthrough_unchanged(monkeypatch):
     # Already-English items come back unchanged in their slot (engine "none").
-    monkeypatch.setattr(ic.requests, "post", lambda *a, **k: _Resp(200, {
+    monkeypatch.setattr(ic.session, "post", lambda *a, **k: _Resp(200, {
         "translatedText": ["already english"],
         "detectedLanguage": {"language": "en"},
         "interpreter": {"engine": "none", "engine_version": None},
@@ -52,7 +53,7 @@ def test_translate_passthrough_unchanged(monkeypatch):
 
 
 def test_translate_503_raises_unavailable(monkeypatch):
-    monkeypatch.setattr(ic.requests, "post", lambda *a, **k: _Resp(503, {}))
+    monkeypatch.setattr(ic.session, "post", lambda *a, **k: _Resp(503, {}))
     with pytest.raises(ic.InterpreterUnavailable):
         ic.translate(["x"], base_url="http://i.lan")
 
@@ -61,14 +62,46 @@ def test_translate_connection_error_raises_unavailable(monkeypatch):
     def boom(*a, **k):
         raise ic.requests.ConnectionError("refused")
 
-    monkeypatch.setattr(ic.requests, "post", boom)
+    monkeypatch.setattr(ic.session, "post", boom)
     with pytest.raises(ic.InterpreterUnavailable):
         ic.translate(["x"], base_url="http://i.lan")
 
 
+@pytest.mark.parametrize("code", [429, 500, 502, 503, 504])
+def test_translate_429_and_5xx_raise_unavailable(monkeypatch, code):
+    # A gracefully-draining service behind a reverse proxy returns 429/502/503/504
+    # — all mean "retry later", not an opaque error.
+    monkeypatch.setattr(ic.session, "post", lambda *a, **k: _Resp(code, {}))
+    with pytest.raises(ic.InterpreterUnavailable) as ei:
+        ic.translate(["x"], base_url="http://i.lan")
+    assert ei.value.retry_after is None
+
+
+def test_translate_honours_retry_after_seconds(monkeypatch):
+    monkeypatch.setattr(
+        ic.session, "post",
+        lambda *a, **k: _Resp(503, {}, headers={"Retry-After": "42"}),
+    )
+    with pytest.raises(ic.InterpreterUnavailable) as ei:
+        ic.translate(["x"], base_url="http://i.lan")
+    assert ei.value.retry_after == 42
+
+
+def test_translate_retry_after_past_http_date_clamps_to_zero(monkeypatch):
+    # HTTP-date form; a past date → non-negative clamp to 0 (deterministic).
+    monkeypatch.setattr(
+        ic.session, "post",
+        lambda *a, **k: _Resp(
+            503, {}, headers={"Retry-After": "Wed, 21 Oct 2015 07:28:00 GMT"}),
+    )
+    with pytest.raises(ic.InterpreterUnavailable) as ei:
+        ic.translate(["x"], base_url="http://i.lan")
+    assert ei.value.retry_after == 0.0
+
+
 def test_translate_400_raises_bad_request(monkeypatch):
     monkeypatch.setattr(
-        ic.requests, "post", lambda *a, **k: _Resp(400, {"error": "bad target"})
+        ic.session, "post", lambda *a, **k: _Resp(400, {"error": "bad target"})
     )
     with pytest.raises(ic.InterpreterBadRequest):
         ic.translate(["x"], base_url="http://i.lan")
@@ -78,20 +111,20 @@ def test_translate_empty_is_noop(monkeypatch):
     def boom(*a, **k):
         raise AssertionError("should not call the service for an empty batch")
 
-    monkeypatch.setattr(ic.requests, "post", boom)
+    monkeypatch.setattr(ic.session, "post", boom)
     assert ic.translate([], base_url="http://i.lan")["translations"] == []
 
 
 def test_health_true_when_llm_up(monkeypatch):
     monkeypatch.setattr(
-        ic.requests, "get", lambda *a, **k: _Resp(200, {"engines": {"llm": True}})
+        ic.session, "get", lambda *a, **k: _Resp(200, {"engines": {"llm": True}})
     )
     assert ic.health("http://i.lan") is True
 
 
 def test_health_false_when_engine_down(monkeypatch):
     monkeypatch.setattr(
-        ic.requests, "get", lambda *a, **k: _Resp(200, {"engines": {"llm": False}})
+        ic.session, "get", lambda *a, **k: _Resp(200, {"engines": {"llm": False}})
     )
     assert ic.health("http://i.lan") is False
 
@@ -100,7 +133,7 @@ def test_health_false_on_error(monkeypatch):
     def boom(*a, **k):
         raise ic.requests.ConnectionError("refused")
 
-    monkeypatch.setattr(ic.requests, "get", boom)
+    monkeypatch.setattr(ic.session, "get", boom)
     assert ic.health("http://i.lan") is False
 
 

tests/test_maintenance.py

@@ -26,6 +26,28 @@ def _make_batch(session) -> int:
     return batch.id
 
 
+def test_cleanup_orphaned_temp_files_removes_stale_only(tmp_path, monkeypatch):
+    import os
+
+    from backend.app.tasks import maintenance as m
+
+    monkeypatch.setattr(m, "IMAGES_ROOT", tmp_path)
+    stale = tmp_path / "artist" / "img.jpg.part"      # killed download → orphan
+    stale.parent.mkdir(parents=True)
+    stale.write_bytes(b"x")
+    old = datetime.now(UTC).timestamp() - 8 * 3600     # older than the 6h guard
+    os.utime(stale, (old, old))
+    fresh = tmp_path / "in_progress.jpg.partial"       # active download → keep
+    fresh.write_bytes(b"x")
+    keep = tmp_path / "real.jpg"                        # a real image → keep
+    keep.write_bytes(b"x")
+
+    assert m.cleanup_orphaned_temp_files() == 1
+    assert not stale.exists()
+    assert fresh.exists()
+    assert keep.exists()
+
+
 def test_recover_interrupted_only_old(db_sync, monkeypatch):
     batch_id = _make_batch(db_sync)
     now = datetime.now(UTC)

tests/test_translate_posts.py

@@ -4,6 +4,7 @@ import pytest
 from sqlalchemy import select
 
 from backend.app.models import Artist, ImportSettings, Post
+from backend.app.services import interpreter_client as ic
 from backend.app.tasks.translation import retranslate_posts, translate_posts
 
 pytestmark = pytest.mark.integration
@@ -136,6 +137,31 @@ def test_translate_posts_service_down_leaves_untranslated(db_sync, monkeypatch):
     assert p.translated_source_lang is None       # will retry next run
 
 
+def test_translate_posts_interrupt_reenqueues_after_backoff(db_sync, monkeypatch):
+    # A drain mid-sweep (health passed, translate 503s w/ Retry-After) re-enqueues
+    # the daily sweep after the backoff instead of waiting for tomorrow's beat.
+    _patch(monkeypatch, db_sync)
+    monkeypatch.setattr("backend.app.tasks.translation.ic.health", lambda *a, **k: True)
+
+    def _drain(texts, **k):
+        raise ic.InterpreterUnavailable("draining", retry_after=30)
+
+    monkeypatch.setattr("backend.app.tasks.translation.ic.translate", _drain)
+    calls = []
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.translate_posts.apply_async",
+        lambda *a, **k: calls.append((a, k)),
+    )
+    a = _artist(db_sync)
+    _post(db_sync, a.id, title="ねこ", ext="p1")
+    _enable(db_sync)
+    db_sync.commit()
+
+    assert "interrupted" in translate_posts()
+    assert len(calls) == 1
+    assert calls[0][1]["countdown"] == 30
+
+
 def _mock_new_model(monkeypatch, *, ver="v2"):
     """Interpreter is up and translates with a NEW engine version."""
     monkeypatch.setattr("backend.app.tasks.translation.ic.health", lambda *a, **k: True)
@@ -239,3 +265,61 @@ def test_retranslate_unhealthy_does_not_reset(db_sync, monkeypatch):
     db_sync.refresh(p)
     assert p.translated_source_lang == "ja"
     assert p.post_title_translated == "OLD"
+
+
+def test_retranslate_interrupt_resumes_after_retry_after(db_sync, monkeypatch):
+    # Interpreter drains mid-chunk (health passed, then translate 503s with a
+    # Retry-After): the bulk re-translate re-enqueues itself after the hinted
+    # backoff, with _reset_done=True so it never re-wipes fresh work — instead of
+    # stalling until the daily beat.
+    _patch(monkeypatch, db_sync)
+    monkeypatch.setattr("backend.app.tasks.translation.ic.health", lambda *a, **k: True)
+
+    def _drain(texts, **k):
+        raise ic.InterpreterUnavailable("draining", retry_after=42)
+
+    monkeypatch.setattr("backend.app.tasks.translation.ic.translate", _drain)
+    calls = []
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.retranslate_posts.apply_async",
+        lambda *a, **k: calls.append((a, k)),
+    )
+    a = _artist(db_sync)
+    p = _post(db_sync, a.id, title="ねこ", ext="p1")
+    _mark_translated(p)
+    _enable(db_sync)
+    db_sync.commit()
+
+    assert "interrupted" in retranslate_posts(artist_ids=[a.id])
+    assert len(calls) == 1                             # resume re-enqueued
+    args, kw = calls[0]
+    assert args[1]["_reset_done"] is True
+    assert args[1]["artist_ids"] == [a.id]
+    assert kw["countdown"] == 42                       # honoured the Retry-After
+    db_sync.refresh(p)
+    assert p.translated_source_lang is None            # reset happened, not re-run
+
+
+def test_retranslate_interrupt_default_backoff_without_hint(db_sync, monkeypatch):
+    # No Retry-After header → fall back to the default backoff (60s).
+    _patch(monkeypatch, db_sync)
+    monkeypatch.setattr("backend.app.tasks.translation.ic.health", lambda *a, **k: True)
+
+    def _drain(texts, **k):
+        raise ic.InterpreterUnavailable("draining")
+
+    monkeypatch.setattr("backend.app.tasks.translation.ic.translate", _drain)
+    calls = []
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.retranslate_posts.apply_async",
+        lambda *a, **k: calls.append((a, k)),
+    )
+    a = _artist(db_sync)
+    p = _post(db_sync, a.id, title="ねこ", ext="p1")
+    _mark_translated(p)
+    _enable(db_sync)
+    db_sync.commit()
+
+    retranslate_posts(artist_ids=[a.id])
+    assert len(calls) == 1
+    assert calls[0][1]["countdown"] == 60