Files
jellyfin-scene-search/scene_search/index_jobs.py
T
2026-10-07 13:52:44 +01:00

409 lines
15 KiB
Python

from __future__ import annotations
import threading
from copy import deepcopy
from datetime import datetime, timezone
from typing import Any
from .database import connect, init_db
from .index_report import load_index_report, save_index_report
from .jellyfin_index import (
build_jellyfin_client,
index_libraries,
reconcile_selected_libraries,
resolve_library_ids,
)
from .search_service import SearchService
from .sync import rebuild_meili
class IndexJobManager:
"""Runs one plugin-triggered Jellyfin indexing job at a time."""
def __init__(self, search_service: SearchService) -> None:
self.search_service = search_service
self._state_lock = threading.Lock()
self._thread: threading.Thread | None = None
self._state: dict[str, Any] = {
"running": False,
"stage": "idle",
"started_at": None,
"finished_at": None,
"error": None,
}
@property
def running(self) -> bool:
with self._state_lock:
return bool(self._state.get("running"))
def status(self) -> dict[str, Any]:
with self._state_lock:
return deepcopy(self._state)
def report(
self,
*,
category: str | None = None,
offset: int = 0,
limit: int = 200,
) -> dict[str, Any]:
report = load_index_report()
items = list(report.get("items") or [])
if category:
items = [
item
for item in items
if str(item.get("status") or "") == category
]
total = len(items)
page = items[offset:offset + limit]
return {
"available": bool(report.get("available", False)),
"kind": report.get("kind"),
"status": report.get("status"),
"started_at": report.get("started_at"),
"finished_at": report.get("finished_at"),
"error": report.get("error"),
"summary": report.get("summary") or {},
"category": category,
"offset": offset,
"limit": limit,
"total": total,
"items": page,
}
def libraries(self) -> list[dict[str, Any]]:
client = build_jellyfin_client()
libraries = client.media_folders()
init_db(self.search_service.db_path)
counts: dict[str, tuple[int, int]] = {}
with connect(self.search_service.db_path) as con:
rows = con.execute(
"""
SELECT m.jellyfin_library_id,
COUNT(DISTINCT m.id) AS indexed_items,
COUNT(f.id) AS indexed_frames
FROM media m
LEFT JOIN frames f ON f.media_id = m.id
WHERE m.source_kind = 'jellyfin'
AND m.jellyfin_library_id IS NOT NULL
GROUP BY m.jellyfin_library_id
"""
).fetchall()
for row in rows:
counts[str(row["jellyfin_library_id"])] = (
int(row["indexed_items"]),
int(row["indexed_frames"]),
)
result = []
for library in sorted(
libraries,
key=lambda item: str(item.get("Name") or "").casefold(),
):
library_id = str(library.get("Id") or "")
indexed_items, indexed_frames = counts.get(library_id, (0, 0))
result.append(
{
"id": library_id,
"name": str(library.get("Name") or library_id),
"indexed_items": indexed_items,
"indexed_frames": indexed_frames,
}
)
return result
def apply_library_selection(self, library_ids: list[str]) -> dict[str, object]:
with self._state_lock:
if self._state.get("running"):
raise RuntimeError("A Jellyfin indexing job is already running.")
selected_ids = {
str(value).strip()
for value in library_ids
if str(value).strip()
}
started_at = datetime.now(timezone.utc).isoformat()
result = reconcile_selected_libraries(selected_ids)
removed_items = list(result.pop("removed_items", []) or [])
removed_ids = [
str(value)
for value in (result.get("removed_library_ids") or [])
]
if removed_ids:
self.search_service.meili.delete_jellyfin_libraries(removed_ids)
print(
"Removed deselected Jellyfin libraries: "
+ ", ".join(removed_ids)
+ f" ({int(result.get('removed_media') or 0)} media item(s))"
)
finished_at = datetime.now(timezone.utc).isoformat()
save_index_report(
{
"available": True,
"kind": "library_selection",
"status": "complete",
"started_at": started_at,
"finished_at": finished_at,
"error": None,
"summary": {
"indexed": 0,
"skipped": 0,
"failed": 0,
"pruned": int(result.get("removed_media") or 0),
"removed_libraries": int(result.get("removed_libraries") or 0),
},
"items": removed_items,
}
)
return result
def start(self, library_ids: list[str], *, force: bool = False) -> dict[str, Any]:
with self._state_lock:
if self._state.get("running"):
raise RuntimeError("A Jellyfin indexing job is already running.")
requested_ids = [
str(value).strip()
for value in library_ids
if str(value).strip()
]
client = build_jellyfin_client()
libraries = (
resolve_library_ids(client, requested_ids)
if requested_ids
else []
)
selected_ids = {
str(item.get("Id") or "")
for item in libraries
if str(item.get("Id") or "")
}
library_item_totals = [
client.item_count(str(item.get("Id") or ""))
for item in libraries
]
overall_items_total = sum(library_item_totals)
with self._state_lock:
self._state = {
"running": True,
"stage": "starting",
"started_at": datetime.now(timezone.utc).isoformat(),
"finished_at": None,
"error": None,
"library_total": len(libraries),
"library_ids": sorted(selected_ids),
"library_item_totals": library_item_totals,
"overall_items_total": overall_items_total,
"overall_items_completed": 0,
"overall_percent": 0.0,
}
self._thread = threading.Thread(
target=self._run,
args=(libraries, selected_ids, force),
name="scene-search-index",
daemon=True,
)
self._thread.start()
return self.status()
def _progress(self, update: dict) -> None:
with self._state_lock:
self._state.update(update)
self._update_overall_progress_locked()
def _update_overall_progress_locked(self) -> None:
stage = str(self._state.get("stage") or "")
total = int(self._state.get("overall_items_total") or 0)
totals = [
int(value)
for value in (self._state.get("library_item_totals") or [])
]
if stage == "complete":
self._state["overall_items_completed"] = total
self._state["overall_percent"] = 100.0
return
if stage in {"indexed", "syncing", "sync", "synced"}:
self._state["overall_items_completed"] = total
self._state["overall_percent"] = 99.0 if total else 99.0
return
library_number = int(self._state.get("library_number") or 1)
library_index = max(0, min(library_number - 1, len(totals)))
before_library = sum(totals[:library_index])
item_number = int(self._state.get("item_number") or 0)
completed_before_item = max(item_number - 1, 0)
completed = min(before_library + completed_before_item, total)
fractional_item = 0.0
frames_total = int(self._state.get("frames_total") or 0)
frames_processed = int(self._state.get("frames_processed") or 0)
if stage == "embedding" and frames_total > 0:
fractional_item = min(max(frames_processed / frames_total, 0.0), 1.0)
position = min(completed + fractional_item, float(total))
self._state["overall_items_completed"] = completed
self._state["overall_percent"] = (
round((position / total) * 99.0, 1)
if total > 0
else 0.0
)
def _run(
self,
libraries: list[dict],
selected_ids: set[str],
force: bool,
) -> None:
report_items: list[dict[str, object]] = []
started_at = str(self.status().get("started_at") or datetime.now(timezone.utc).isoformat())
try:
removed = reconcile_selected_libraries(selected_ids)
report_items.extend(list(removed.pop("removed_items", []) or []))
removed_ids = [
str(value)
for value in (removed.get("removed_library_ids") or [])
]
if removed_ids:
self.search_service.meili.delete_jellyfin_libraries(removed_ids)
self._progress(
{
"removed_libraries": int(removed["removed_libraries"]),
"removed_library_media": int(removed["removed_media"]),
}
)
print(
"Removed deselected Jellyfin libraries: "
+ ", ".join(removed_ids)
+ f" ({int(removed.get('removed_media') or 0)} media item(s))"
)
if not libraries:
finished_at = datetime.now(timezone.utc).isoformat()
summary = {
"library_total": 0,
"indexed": 0,
"skipped": 0,
"failed": 0,
"pruned": int(removed.get("removed_media") or 0),
"removed_libraries": int(removed.get("removed_libraries") or 0),
}
self._progress(
{
"running": False,
"stage": "complete",
"finished_at": finished_at,
"documents_processed": 0,
**summary,
}
)
save_index_report(
{
"available": True,
"kind": "index",
"status": "complete",
"started_at": started_at,
"finished_at": finished_at,
"error": None,
"summary": summary,
"items": report_items,
}
)
return
# Search and indexing share one model instance and never use it concurrently.
with self.search_service.model_lock:
embedder = self.search_service.get_embedder()
_, summary = index_libraries(
build_jellyfin_client(),
libraries,
embedder=embedder,
force=force,
progress=self._progress,
report_items=report_items,
)
retained_index_changed = (
int(summary.get("indexed") or 0) > 0
or int(summary.get("pruned") or 0) > 0
)
summary["pruned"] = (
int(summary.get("pruned") or 0)
+ int(removed.get("removed_media") or 0)
)
summary["removed_libraries"] = int(
removed.get("removed_libraries") or 0
)
if retained_index_changed:
self._progress({"stage": "syncing", **summary})
documents = rebuild_meili(progress=self._progress)
else:
documents = 0
print(
"Selected libraries are unchanged; "
"skipping full Meilisearch rebuild."
)
finished_at = datetime.now(timezone.utc).isoformat()
self._progress(
{
"running": False,
"stage": "complete",
"finished_at": finished_at,
"documents_processed": documents,
"full_meili_rebuild": retained_index_changed,
**summary,
}
)
save_index_report(
{
"available": True,
"kind": "index",
"status": "complete",
"started_at": started_at,
"finished_at": finished_at,
"error": None,
"summary": {
**summary,
"documents_processed": documents,
"full_meili_rebuild": retained_index_changed,
},
"items": report_items,
}
)
except Exception as exc:
finished_at = datetime.now(timezone.utc).isoformat()
self._progress(
{
"running": False,
"stage": "failed",
"finished_at": finished_at,
"error": str(exc),
}
)
save_index_report(
{
"available": True,
"kind": "index",
"status": "failed",
"started_at": started_at,
"finished_at": finished_at,
"error": str(exc),
"summary": {
"indexed": sum(1 for item in report_items if item.get("status") == "indexed"),
"failed": sum(1 for item in report_items if item.get("status") == "failed"),
"pruned": sum(1 for item in report_items if item.get("status") == "removed"),
},
"items": report_items,
}
)