feat(translation): re-translate on model change — artist-scoped + global re-run (#146)

retranslate_posts resets the 5 translation columns to NULL for a scoped set of posts (all, or WHERE artist_id IN ids) then reuses the untranslated sweep to re-run them, chasing the tail until drained (run-until-done). Interpreter cache keys on engine_version so a changed model re-translates, an unchanged one is cache-fast. Reset only happens when the service is configured+healthy so translations are never wiped when they can't be rebuilt. New POST /settings/translation/retranslate (artist_id | all=true). UI: per-artist 'Re-translate posts' on the Artist Management tab + 'Re-translate all' in the Settings Translation card, both with confirm dialogs. No migration (reuses m143 columns).

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

tests/test_api_settings.py

@@ -97,6 +97,68 @@ async def test_translation_run_enqueues_when_configured(client, monkeypatch):
     assert (await resp.get_json())["celery_task_id"] == "t1"
 
 
+def _fake_retranslate(monkeypatch, sink):
+    # Record the kwargs the endpoint enqueues with, return a fake AsyncResult.
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.retranslate_posts.delay",
+        lambda *a, **k: sink.update(k) or type("R", (), {"id": "r1"})(),
+    )
+
+
+@pytest.mark.asyncio
+async def test_translation_retranslate_requires_config(client):
+    # Disabled → 400 even with an explicit scope (no reset is enqueued).
+    resp = await client.post(
+        "/api/settings/translation/retranslate", json={"all": True})
+    assert resp.status_code == 400
+
+
+@pytest.mark.asyncio
+async def test_translation_retranslate_requires_target(client, monkeypatch):
+    # 'all' must be explicit: an empty body is a 400, so a stray call can't
+    # wipe every translation.
+    _fake_retranslate(monkeypatch, {})
+    await client.patch("/api/settings/import", json={
+        "translation_enabled": True, "interpreter_base_url": "http://i.lan",
+    })
+    resp = await client.post("/api/settings/translation/retranslate")
+    assert resp.status_code == 400
+
+
+@pytest.mark.asyncio
+async def test_translation_retranslate_artist_scope(client, monkeypatch):
+    sink = {}
+    _fake_retranslate(monkeypatch, sink)
+    await client.patch("/api/settings/import", json={
+        "translation_enabled": True, "interpreter_base_url": "http://i.lan",
+    })
+    resp = await client.post(
+        "/api/settings/translation/retranslate", json={"artist_id": 7})
+    assert resp.status_code == 202
+    assert (await resp.get_json())["celery_task_id"] == "r1"
+    assert sink["artist_ids"] == [7]
+
+
+@pytest.mark.asyncio
+async def test_translation_retranslate_all_scope(client, monkeypatch):
+    sink = {}
+    _fake_retranslate(monkeypatch, sink)
+    await client.patch("/api/settings/import", json={
+        "translation_enabled": True, "interpreter_base_url": "http://i.lan",
+    })
+    resp = await client.post(
+        "/api/settings/translation/retranslate", json={"all": True})
+    assert resp.status_code == 202
+    assert sink["artist_ids"] is None
+
+
+@pytest.mark.asyncio
+async def test_translation_retranslate_bad_artist_id(client):
+    resp = await client.post(
+        "/api/settings/translation/retranslate", json={"artist_id": "abc"})
+    assert resp.status_code == 400
+
+
 @pytest.mark.asyncio
 async def test_patch_rejects_non_object(client):
     resp = await client.patch("/api/settings/import", json=[1, 2, 3])

tests/test_translate_posts.py

@@ -4,7 +4,7 @@ import pytest
 from sqlalchemy import select
 
 from backend.app.models import Artist, ImportSettings, Post
-from backend.app.tasks.translation import translate_posts
+from backend.app.tasks.translation import retranslate_posts, translate_posts
 
 pytestmark = pytest.mark.integration
 
@@ -28,8 +28,8 @@ def _sf(db_sync):
     return _SM()
 
 
-def _artist(db):
-    a = Artist(name="A", slug="a")
+def _artist(db, name="A", slug="a"):
+    a = Artist(name=name, slug=slug)
     db.add(a)
     db.flush()
     return a
@@ -45,6 +45,15 @@ def _post(db, artist_id, *, title=None, desc=None, ext="p1"):
     return p
 
 
+def _mark_translated(p, *, lang="ja", ver="v1"):
+    """Simulate a prior (old-model) translation so retranslate has something to
+    reset."""
+    p.post_title_translated = "OLD"
+    p.description_translated = "OLD"
+    p.translated_source_lang = lang
+    p.translation_engine_version = ver
+
+
 def _enable(db, url="http://i.lan"):
     cfg = db.execute(
         select(ImportSettings).where(ImportSettings.id == 1)
@@ -125,3 +134,108 @@ def test_translate_posts_service_down_leaves_untranslated(db_sync, monkeypatch):
     assert translate_posts() == "interpreter unavailable"
     db_sync.refresh(p)
     assert p.translated_source_lang is None       # will retry next run
+
+
+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)
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.ic.translate",
+        lambda texts, **k: {
+            "translations": [f"NEW:{t}" for t in texts],
+            "detected_lang": "ja", "engine": "llm", "engine_version": ver,
+        },
+    )
+
+
+def test_retranslate_resets_and_reruns(db_sync, monkeypatch):
+    # A post translated by the old model gets cleared + re-run through the new
+    # one (new engine_version proves the reset happened, not a cache short-circuit).
+    _patch(monkeypatch, db_sync)
+    _mock_new_model(monkeypatch)
+    a = _artist(db_sync)
+    p = _post(db_sync, a.id, title="ねこ", desc="かわいい")
+    _mark_translated(p)
+    _enable(db_sync)
+    db_sync.commit()
+
+    retranslate_posts(artist_ids=[a.id])
+    db_sync.refresh(p)
+    assert p.post_title_translated == "NEW:ねこ"
+    assert p.description_translated == "NEW:かわいい"
+    assert p.translation_engine_version == "v2"
+
+
+def test_retranslate_artist_scope_leaves_others(db_sync, monkeypatch):
+    # Scoping to one artist must not touch another artist's translations.
+    _patch(monkeypatch, db_sync)
+    _mock_new_model(monkeypatch)
+    a1 = _artist(db_sync)
+    a2 = _artist(db_sync, name="B", slug="b")
+    p1 = _post(db_sync, a1.id, title="ねこ", ext="p1")
+    p2 = _post(db_sync, a2.id, title="いぬ", ext="p2")
+    _mark_translated(p1)
+    _mark_translated(p2)
+    _enable(db_sync)
+    db_sync.commit()
+
+    retranslate_posts(artist_ids=[a1.id])
+    db_sync.refresh(p1)
+    db_sync.refresh(p2)
+    assert p1.post_title_translated == "NEW:ねこ"     # re-run
+    assert p1.translation_engine_version == "v2"
+    assert p2.post_title_translated == "OLD"          # untouched
+    assert p2.translation_engine_version == "v1"
+
+
+def test_retranslate_runs_until_done(db_sync, monkeypatch):
+    # With the chunk capped at 1 and 2 posts to redo, the first chunk re-enqueues
+    # itself (with _reset_done=True so it doesn't wipe the fresh work) to finish.
+    _patch(monkeypatch, db_sync)
+    _mock_new_model(monkeypatch)
+    monkeypatch.setattr("backend.app.tasks.translation._MAX_POSTS_PER_RUN", 1)
+    calls = []
+    monkeypatch.setattr(
+        "backend.app.tasks.translation.retranslate_posts.apply_async",
+        lambda *a, **k: calls.append((a, k)),
+    )
+    a = _artist(db_sync)
+    p1 = _post(db_sync, a.id, title="ねこ", ext="p1")
+    p2 = _post(db_sync, a.id, title="いぬ", ext="p2")
+    _mark_translated(p1)
+    _mark_translated(p2)
+    _enable(db_sync)
+    db_sync.commit()
+
+    retranslate_posts(artist_ids=[a.id])
+    assert len(calls) == 1                             # one re-enqueue for the tail
+    args, _kw = calls[0]
+    assert args[1]["_reset_done"] is True
+    assert args[1]["artist_ids"] == [a.id]
+
+
+def test_retranslate_disabled_does_not_reset(db_sync, monkeypatch):
+    # Never wipe translations we can't rebuild — a disabled service leaves them.
+    _patch(monkeypatch, db_sync)
+    a = _artist(db_sync)
+    p = _post(db_sync, a.id, title="ねこ", ext="p1")
+    _mark_translated(p)
+    db_sync.commit()   # translation_enabled defaults False
+    assert retranslate_posts(artist_ids=[a.id]) == "disabled"
+    db_sync.refresh(p)
+    assert p.translated_source_lang == "ja"
+    assert p.post_title_translated == "OLD"
+
+
+def test_retranslate_unhealthy_does_not_reset(db_sync, monkeypatch):
+    _patch(monkeypatch, db_sync)
+    monkeypatch.setattr("backend.app.tasks.translation.ic.health", lambda *a, **k: False)
+    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 retranslate_posts(artist_ids=[a.id]) == "interpreter unavailable"
+    db_sync.refresh(p)
+    assert p.translated_source_lang == "ja"
+    assert p.post_title_translated == "OLD"