Files
FabledCurator/backend/app/services/series_match_service.py
T
bvandeusen 59746d213d
CI / lint (push) Successful in 3s
CI / frontend-build (push) Successful in 21s
CI / backend-lint-and-test (push) Successful in 35s
CI / integration (push) Successful in 3m13s
feat(series): flat series sequence + cosmetic chapter dividers (#789 Phase 1)
Reframe a series from "ordered chapters that own pages" to ONE flat,
series-global ordered run of pages with optional cosmetic chapter DIVIDERS
over it. A chapter no longer wraps content — it's a labeled divider anchored
to the page that begins it; a page's chapter is derived as the nearest
preceding divider. This is what lets installments assembled from multiple
sources sit in one continuous, correctly-numbered sequence (operator's
Goblin Juice case).

- migration 0047: flatten each series to a series-global page_number
  (preserving today's reading order); convert each existing chapter to a
  divider anchored at its first page (keeping title/stated_part); drop
  series_page.chapter_id; reshape series_chapter (anchor_page_id UNIQUE FK,
  drop chapter_number/is_placeholder/stated_page_start/end). Loss-safe for
  content; drops empty placeholder chapters + a redundant page-1 divider.
- series_page: page_number is now the series-global order; no chapter_id.
- series_chapter: anchored divider (anchor_page_id, title, stated_part).
- series_service: flat list_pages (one run + derived dividers + per-page
  source_post + part_gaps), series-wide reorder/renumber, divider CRUD
  (create/update/move/delete); retired per-chapter reorder/merge/placement.
- api/tags: drop chapter_id from add; /chapters endpoints are divider
  create/update/delete (removed chapter reorder/merge/page-reorder).
- series_match_service: series "end" reads max(series_page.stated_page);
  accept appends via add_post. tag_service series-merge appends src's pages
  after tgt's max so the merged series stays one clean run.
- frontend: seriesManage store + SeriesManageView → one continuous
  drag-reorder grid with inline divider bars + series-global page numbers;
  reader walks the flat run, headings from dividers; PostSeriesMenu copy.
- tests reworked across the series suite for the divider model.

Phase 2 (pending staging for add-from-post) is separate.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-11 21:30:01 -04:00

339 lines
12 KiB
Python

"""FC-6.3 — assisted continuation matcher.
Scores a (post, candidate series) pair from several weighted signals; above the
configured threshold it records a *pending* SeriesSuggestion. Confirm-only — the
operator accepts (post becomes a chapter) or dismisses. No single signal gates;
the score is an additive weighted sum, each signal a 0..1 strength.
The learned "title pattern" isn't persisted — it's derived on the fly from the
post titles already in a series, so it sharpens automatically as more posts are
confirmed into the series.
"""
import re
import time
from sqlalchemy import and_, func, select
from sqlalchemy.dialects.postgresql import insert as pg_insert
from sqlalchemy.ext.asyncio import AsyncSession
from ..models import ImageRecord, Post, Tag, TagKind
from ..models.series_page import SeriesPage
from ..models.series_suggestion import SeriesSuggestion
from ..models.tag import image_tag
from .page_number_parser import parse_page_range
from .series_service import SeriesError, SeriesService
# Additive signal weights (sum to 1.0 → max score 1.0). Kept as constants;
# only the on/off + threshold are operator-tunable (sensitivity is the knob
# that matters; per-signal weights are an over-tune for v1).
WEIGHTS = {"title": 0.40, "artist": 0.20, "pages": 0.25, "tags": 0.15}
_DISTINCTIVE_KINDS = (TagKind.character, TagKind.series, TagKind.fandom)
_MAX_CANDIDATES = 50
# Strip installment markers so titles collapse to their stable stem.
_PAGE_TOKEN = re.compile(r"\b(?:pages?|pgs?|pp\.?)\s*\d+(?:\s*[-–—/]\s*\d+)?", re.I)
_BRACKET_NUM = re.compile(r"[\[(]\s*\d+\s*(?:/\s*\d+)?\s*[\])]")
_TRAILING_NUM = re.compile(r"[\s\-_#]*\d+\s*$")
def normalize_title(title: str | None) -> str:
if not title:
return ""
t = _PAGE_TOKEN.sub("", title)
t = _BRACKET_NUM.sub("", t)
t = _TRAILING_NUM.sub("", t)
return " ".join(t.split()).strip().lower()
def _common_prefix(strings: list[str]) -> str:
strings = [s for s in strings if s]
if not strings:
return ""
pre = strings[0]
for s in strings[1:]:
i = 0
while i < len(pre) and i < len(s) and pre[i] == s[i]:
i += 1
pre = pre[:i]
if not pre:
break
return pre.strip()
def title_signal(series_titles: list[str], post_title: str | None) -> float:
"""Overlap of the post title against the series' title stem (the longest
common prefix of its known titles, page/installment markers removed)."""
norm = [normalize_title(t) for t in series_titles]
norm = [t for t in norm if t]
pt = normalize_title(post_title)
if not norm or not pt:
return 0.0
stem = _common_prefix(norm)
if len(stem) < 3:
stem = max(norm, key=len)
i = 0
while i < len(stem) and i < len(pt) and stem[i] == pt[i]:
i += 1
return min(1.0, i / max(len(stem), 4))
def pages_signal(series_max_stated_end: int | None, post_start: int | None) -> float:
"""1.0 when the post's first page continues right after the series' last
stated page; partial for a near-continuation; 0 otherwise."""
if series_max_stated_end is None or post_start is None:
return 0.0
diff = post_start - series_max_stated_end
if diff == 1:
return 1.0
if 2 <= diff <= 3:
return 0.5
return 0.0
def tags_signal(shared_distinctive: int) -> float:
if shared_distinctive <= 0:
return 0.0
return min(1.0, shared_distinctive / 3.0)
def weighted_score(signals: dict) -> float:
return round(sum(WEIGHTS[k] * signals.get(k, 0.0) for k in WEIGHTS), 4)
class SeriesMatchService:
def __init__(self, session: AsyncSession):
self.session = session
async def _post_image_ids(self, post_id: int) -> list[int]:
rows = (
await self.session.execute(
select(ImageRecord.id).where(ImageRecord.primary_post_id == post_id)
)
).scalars().all()
return list(rows)
async def _series_image_ids(self, series_tag_id: int) -> set[int]:
rows = (
await self.session.execute(
select(SeriesPage.image_id).where(
SeriesPage.series_tag_id == series_tag_id
)
)
).scalars().all()
return set(rows)
async def _series_post_titles(self, series_tag_id: int) -> list[str]:
rows = (
await self.session.execute(
select(Post.post_title)
.select_from(SeriesPage)
.join(ImageRecord, ImageRecord.id == SeriesPage.image_id)
.join(Post, Post.id == ImageRecord.primary_post_id)
.where(
and_(
SeriesPage.series_tag_id == series_tag_id,
Post.post_title.isnot(None),
)
)
.distinct()
)
).scalars().all()
return [t for t in rows if t]
async def _series_max_stated_end(self, series_tag_id: int) -> int | None:
# Highest printed page across the series' pages (FC-6.x: stated pages
# live on series_page now; chapters are cosmetic dividers).
return await self.session.scalar(
select(func.max(SeriesPage.stated_page)).where(
SeriesPage.series_tag_id == series_tag_id
)
)
async def _distinctive_tags(self, image_ids: list[int] | set[int]) -> set[int]:
ids = list(image_ids)
if not ids:
return set()
rows = (
await self.session.execute(
select(image_tag.c.tag_id)
.join(Tag, Tag.id == image_tag.c.tag_id)
.where(
and_(
image_tag.c.image_record_id.in_(ids),
Tag.kind.in_(_DISTINCTIVE_KINDS),
)
)
.distinct()
)
).scalars().all()
return set(rows)
async def _candidate_series(self, artist_id: int) -> list[int]:
"""Series that already contain a page by this artist — cross-artist
series are rare, so same-artist is the candidate bound."""
rows = (
await self.session.execute(
select(SeriesPage.series_tag_id)
.join(ImageRecord, ImageRecord.id == SeriesPage.image_id)
.where(ImageRecord.artist_id == artist_id)
.distinct()
)
).scalars().all()
return list(rows)
async def _decided_series(self, post_id: int) -> set[int]:
rows = (
await self.session.execute(
select(SeriesSuggestion.series_tag_id).where(
and_(
SeriesSuggestion.post_id == post_id,
SeriesSuggestion.status.in_(["added", "dismissed"]),
)
)
)
).scalars().all()
return set(rows)
async def match_post(self, post_id: int, *, threshold: float) -> int:
"""Score a post against its artist's series; upsert pending suggestions
for those at/above threshold. Returns the number written."""
post = await self.session.get(Post, post_id)
if post is None or post.artist_id is None:
return 0
post_images = await self._post_image_ids(post_id)
if not post_images:
return 0
post_image_set = set(post_images)
rng = parse_page_range(f"{post.post_title or ''} {post.description or ''}")
post_start = rng[0] if rng else None
post_dtags = await self._distinctive_tags(post_images)
decided = await self._decided_series(post_id)
candidates = await self._candidate_series(post.artist_id)
made = 0
for sid in candidates[:_MAX_CANDIDATES]:
if sid in decided:
continue
s_images = await self._series_image_ids(sid)
if post_image_set & s_images:
continue # the post is already (partly) in this series
signals = {
"title": title_signal(
await self._series_post_titles(sid), post.post_title
),
"artist": 1.0, # candidates are same-artist by construction
"pages": pages_signal(
await self._series_max_stated_end(sid), post_start
),
"tags": tags_signal(
len(post_dtags & await self._distinctive_tags(s_images))
),
}
score = weighted_score(signals)
if score < threshold:
continue
await self.session.execute(
pg_insert(SeriesSuggestion)
.values(
post_id=post_id, series_tag_id=sid,
score=score, signals=signals, status="pending",
)
.on_conflict_do_update(
constraint="uq_series_suggestion_post_series",
set_={"score": score, "signals": signals, "status": "pending"},
where=SeriesSuggestion.status == "pending",
)
)
made += 1
return made
# ---- queue ops --------------------------------------------------------
async def list_pending(self) -> list[dict]:
rows = (
await self.session.execute(
select(
SeriesSuggestion.id,
SeriesSuggestion.post_id,
SeriesSuggestion.series_tag_id,
SeriesSuggestion.score,
SeriesSuggestion.signals,
Post.post_title,
Post.external_post_id,
Tag.name.label("series_name"),
)
.join(Post, Post.id == SeriesSuggestion.post_id)
.join(Tag, Tag.id == SeriesSuggestion.series_tag_id)
.where(SeriesSuggestion.status == "pending")
.order_by(SeriesSuggestion.score.desc())
)
).all()
return [
{
"id": r.id,
"post_id": r.post_id,
"series_tag_id": r.series_tag_id,
"series_name": r.series_name,
"post_title": r.post_title or f"Post {r.external_post_id}",
"score": r.score,
"signals": r.signals or {},
}
for r in rows
]
async def accept(self, suggestion_id: int) -> dict:
s = await self.session.get(SeriesSuggestion, suggestion_id)
if s is None:
raise SeriesError(f"suggestion {suggestion_id} not found")
if s.status != "pending":
raise SeriesError(f"suggestion {suggestion_id} is already {s.status}")
out = await SeriesService(self.session).add_post(
s.series_tag_id, s.post_id
)
s.status = "added"
self.session.add(s)
return out
async def dismiss(self, suggestion_id: int) -> None:
s = await self.session.get(SeriesSuggestion, suggestion_id)
if s is None:
raise SeriesError(f"suggestion {suggestion_id} not found")
if s.status == "pending":
s.status = "dismissed"
self.session.add(s)
async def rescan(
self, *, threshold: float, time_budget_seconds: float | None = None,
after_post_id: int = 0,
) -> dict:
"""Score every post (id > after_post_id) against its artist's series,
time-boxed + resumable like the other long maintenance sweeps."""
post_ids = (
await self.session.execute(
select(Post.id)
.where(Post.id > after_post_id)
.order_by(Post.id.asc())
)
).scalars().all()
summary = {
"scanned": 0, "suggested": 0,
"partial": False, "resume_after_id": after_post_id,
}
start = time.monotonic()
for pid in post_ids:
summary["scanned"] += 1
summary["resume_after_id"] = pid
summary["suggested"] += await self.match_post(pid, threshold=threshold)
await self.session.commit() # commit per post so progress survives
if (
time_budget_seconds is not None
and time.monotonic() - start >= time_budget_seconds
):
summary["partial"] = True
break
else:
summary["partial"] = False
return summary