dtoro
574d71371f
Drops face recognition, OCR, object detection, and semantic embeddings. The sole remaining vision task is a CLIP-based binary classifier (photography vs other); photos in "other" get needs_review=true so screenshots, documents, memes and scans can be triaged from a new filter pill in the UI. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
35 lines
862 B
YAML
35 lines
862 B
YAML
# Mulita configuration file
|
|
#
|
|
# Source roots and discard handling are owned by the database — manage them
|
|
# from the UI (left sidebar → "+ Add Source Folder") or via the API. Only
|
|
# operational tuning lives here.
|
|
|
|
thumbnails:
|
|
small: 240 # px, longest edge
|
|
medium: 640
|
|
large: 1280
|
|
quality: 85 # JPEG/WebP quality
|
|
format: webp # output format for thumbs
|
|
|
|
scanner:
|
|
watch: true # use watchfiles inotify
|
|
initial_scan_on_start: true
|
|
batch_size: 100
|
|
concurrent_workers: 4
|
|
|
|
performance:
|
|
max_concurrent_thumbnails: 10
|
|
cache_ttl: 3600
|
|
db_pool_size: 20
|
|
db_pool_recycle: 3600
|
|
|
|
# Vision pipeline — single binary classifier (photography vs other).
|
|
# Photos landing in 'other' get needs_review=true.
|
|
vision:
|
|
enabled: true
|
|
backend: onnx
|
|
models_dir: /data/models
|
|
classifier:
|
|
min_confidence: 0.3
|
|
worker_concurrency: 2
|