Add semantic tagging, search, favourite/NSFW filtering, and LLM job queue

Replaces old list-format tags (which duplicated prompt content) with structured
dict tags per category (origin_series, outfit_type, participants, style_type,
scene_type, etc.). Tags are now purely organizational metadata — removed from
the prompt pipeline entirely.

Adds is_favourite and is_nsfw columns to all 8 resource models. Favourite is
DB-only (user preference); NSFW is mirrored in JSON tags for rescan persistence.
All library pages get filter controls and favourites-first sorting.

Introduces a parallel LLM job queue (_enqueue_task + _llm_queue_worker) for
background tag regeneration, with the same status polling UI as ComfyUI jobs.
Fixes call_llm() to use has_request_context() fallback for background threads.

Adds global search (/search) across resources and gallery images, with navbar
search bar. Adds gallery image sidecar JSON for per-image favourite/NSFW metadata.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Aodhan Collins
2026-03-21 03:22:09 +00:00
parent 7d79e626a5
commit 32a73b02f5
72 changed files with 3163 additions and 2212 deletions

View File

@@ -1,7 +1,6 @@
import json
import os
import re
import time
import logging
from flask import render_template, request, redirect, url_for, flash, session, current_app
@@ -10,7 +9,7 @@ from sqlalchemy.orm.attributes import flag_modified
from models import db, Checkpoint, Character, Settings
from services.workflow import _prepare_workflow, _get_default_checkpoint, _apply_checkpoint_settings, _log_workflow_prompts
from services.job_queue import _enqueue_job, _make_finalize
from services.job_queue import _enqueue_job, _make_finalize, _enqueue_task
from services.prompts import build_prompt, _resolve_character, _ensure_character_fields, _append_background
from services.sync import sync_checkpoints, _default_checkpoint_data
from services.file_io import get_available_checkpoints
@@ -57,8 +56,17 @@ def register_routes(app):
@app.route('/checkpoints')
def checkpoints_index():
checkpoints = Checkpoint.query.order_by(Checkpoint.name).all()
return render_template('checkpoints/index.html', checkpoints=checkpoints)
query = Checkpoint.query
fav = request.args.get('favourite')
nsfw = request.args.get('nsfw', 'all')
if fav == 'on':
query = query.filter_by(is_favourite=True)
if nsfw == 'sfw':
query = query.filter_by(is_nsfw=False)
elif nsfw == 'nsfw':
query = query.filter_by(is_nsfw=True)
checkpoints = query.order_by(Checkpoint.is_favourite.desc(), Checkpoint.name).all()
return render_template('checkpoints/index.html', checkpoints=checkpoints, favourite_filter=fav or '', nsfw_filter=nsfw)
@app.route('/checkpoints/rescan', methods=['POST'])
def rescan_checkpoints():
@@ -189,9 +197,9 @@ def register_routes(app):
os.makedirs(checkpoints_dir, exist_ok=True)
overwrite = request.form.get('overwrite') == 'true'
created_count = 0
skipped_count = 0
overwritten_count = 0
skipped = 0
written_directly = 0
job_ids = []
system_prompt = load_prompt('checkpoint_system.txt')
if not system_prompt:
@@ -219,7 +227,7 @@ def register_routes(app):
is_existing = os.path.exists(json_path)
if is_existing and not overwrite:
skipped_count += 1
skipped += 1
continue
# Look for a matching HTML file alongside the model file
@@ -235,52 +243,72 @@ def register_routes(app):
clean_html = re.sub(r'<[^>]+>', ' ', clean_html)
html_content = ' '.join(clean_html.split())
except Exception as e:
print(f"Error reading HTML for {filename}: {e}")
logger.error("Error reading HTML for %s: %s", filename, e)
defaults = _default_checkpoint_data(checkpoint_path, filename)
if html_content:
try:
print(f"Asking LLM to describe checkpoint: {filename}")
prompt = (
f"Generate checkpoint metadata JSON for the model file: '{filename}' "
f"(checkpoint_path: '{checkpoint_path}').\n\n"
f"Here is descriptive text extracted from an associated HTML file:\n###\n{html_content[:3000]}\n###"
)
llm_response = call_llm(prompt, system_prompt)
clean_json = llm_response.replace('```json', '').replace('```', '').strip()
ckpt_data = json.loads(clean_json)
# Enforce fixed fields
ckpt_data['checkpoint_path'] = checkpoint_path
ckpt_data['checkpoint_name'] = filename
# Fill missing fields with defaults
for key, val in defaults.items():
if key not in ckpt_data or ckpt_data[key] is None:
ckpt_data[key] = val
time.sleep(0.5)
except Exception as e:
print(f"LLM error for {filename}: {e}. Using defaults.")
ckpt_data = defaults
# Has HTML companion — enqueue LLM task
def make_task(filename, checkpoint_path, json_path, html_content, system_prompt, defaults, is_existing):
def task_fn(job):
prompt = (
f"Generate checkpoint metadata JSON for the model file: '{filename}' "
f"(checkpoint_path: '{checkpoint_path}').\n\n"
f"Here is descriptive text extracted from an associated HTML file:\n###\n{html_content[:3000]}\n###"
)
try:
llm_response = call_llm(prompt, system_prompt)
clean_json = llm_response.replace('```json', '').replace('```', '').strip()
ckpt_data = json.loads(clean_json)
ckpt_data['checkpoint_path'] = checkpoint_path
ckpt_data['checkpoint_name'] = filename
for key, val in defaults.items():
if key not in ckpt_data or ckpt_data[key] is None:
ckpt_data[key] = val
except Exception as e:
logger.error("LLM error for %s: %s. Using defaults.", filename, e)
ckpt_data = defaults
with open(json_path, 'w') as f:
json.dump(ckpt_data, f, indent=2)
job['result'] = {'name': filename, 'action': 'overwritten' if is_existing else 'created'}
return task_fn
job = _enqueue_task(f"Create checkpoint: {filename}", make_task(filename, checkpoint_path, json_path, html_content, system_prompt, defaults, is_existing))
job_ids.append(job['id'])
else:
ckpt_data = defaults
# No HTML — write defaults directly (no LLM needed)
try:
with open(json_path, 'w') as f:
json.dump(defaults, f, indent=2)
written_directly += 1
except Exception as e:
logger.error("Error saving JSON for %s: %s", filename, e)
try:
with open(json_path, 'w') as f:
json.dump(ckpt_data, f, indent=2)
if is_existing:
overwritten_count += 1
else:
created_count += 1
except Exception as e:
print(f"Error saving JSON for {filename}: {e}")
needs_sync = len(job_ids) > 0 or written_directly > 0
if created_count > 0 or overwritten_count > 0:
sync_checkpoints()
msg = f'Successfully processed checkpoints: {created_count} created, {overwritten_count} overwritten.'
if skipped_count > 0:
msg += f' (Skipped {skipped_count} existing)'
flash(msg)
else:
flash(f'No checkpoints created or overwritten. {skipped_count} existing entries found.')
if needs_sync:
if job_ids:
# Sync after all LLM tasks complete
def sync_task(job):
sync_checkpoints()
job['result'] = {'synced': True}
_enqueue_task("Sync checkpoints DB", sync_task)
else:
# No LLM tasks — sync immediately
sync_checkpoints()
if request.headers.get('X-Requested-With') == 'XMLHttpRequest':
return {'success': True, 'queued': len(job_ids), 'written_directly': written_directly, 'skipped': skipped}
flash(f'Queued {len(job_ids)} checkpoint tasks, {written_directly} written directly ({skipped} skipped).')
return redirect(url_for('checkpoints_index'))
@app.route('/checkpoint/<path:slug>/favourite', methods=['POST'])
def toggle_checkpoint_favourite(slug):
ckpt = Checkpoint.query.filter_by(slug=slug).first_or_404()
ckpt.is_favourite = not ckpt.is_favourite
db.session.commit()
if request.headers.get('X-Requested-With') == 'XMLHttpRequest':
return {'success': True, 'is_favourite': ckpt.is_favourite}
return redirect(url_for('checkpoint_detail', slug=slug))