v0.2.16

feat: ddd identify - recover 'Artist - Title' for unnamed/mislabeled tracks via acoustic fingerprint (fpcalc/Chromaprint + AcoustID, free). GUI tab with per-track confirm (nothing renamed until you click), CLI --validate mode to calibrate confidence threshold against your own library. Requires a free AcoustID API key (acoustid.org/new-application) via Settings or 'ddd config set acoustid_api_key'. Note: even high-confidence matches can be wrong (samples/edits share fingerprints) - always review before applying.

v0.2.15

fix: manual override (checked track = force upgrade even if above bar or already in library, in-place deposit), fix: suspect forensic tracks never 'already good', fix: 'already good' status instead of stuck 'queued...', fix: auto-pick listen port (no more Hyper-V/WSL/Docker port clash), fix: all audio tracks checkable, fix: window centered on startup, feat: startup splash screen

v0.2.14

fix: real sldl error messages (no more fake slskd), feat: single-instance guard (no more double-click port clash), feat: startup splash screen, docs: TROUBLESHOOTING.md, fix: README slsknet.org account confusion

v0.2.13

feat: target format modes (MP3 320 / WAV-AIFF / FLAC only), download counter badge, Docker full pipeline (sldl bundled)

v0.2.12

feat: macOS Apple Silicon support (.app), forensic detector (multi-signal), editable folder path, Flet 0.28.2 pin. Windows title fix.

v0.2.11

feat: forensic lossless detector (multi-signal, opt-in via Settings). Cutoff + HF aliasing + MP3 comb + pre-echo + stereo collapse, consensus anchored. 73 false positives fixed, 1 hidden transcode caught. GUI: detector toggle, suspect badge, scan shows current folder. Default unchanged (legacy), switch in Settings.

v0.2.10

fix: cutoff max au lieu de min (reduit les faux positifs sur morceaux dynamiques), excludes spec -130Mo (pytest, onnxruntime.transformers, numba, llvmlite)

v0.2.9

fix: matching strict anti-remplacement, quick wins GUI, table cap 1000, notif mise a jour, compteur sort par fichier

v0.2.8

fix: credentials and quality preset were never read from DDD config (wrong import path: from .. import config -> from . import config). Users without slskd installed could not connect. Also fixes quality preset selection being silently ignored.

v0.2.7

feat: classif genre par analyse audio locale (Discogs-EffNet ONNX, 400 styles, zero cloud, ~0.5s/track)
feat: cascade complete tag ID3 -> Discogs -> audio-ML -> INBOX
feat: redesign GUI creme (palette warm, polices Anton + DM Mono, alignement baseline)
feat: nettoyage affichage tracks (strip Premiere, [Label], mots promo)
feat: tri par genre - tag ID3 combine avec Discogs (le plus specifique gagne)
feat: landing page recoloree + lightbox screenshots
fix: nav custom (remplace ft.Tabs), ListView, fix grey-block Flet
Credit modele audio : MTG/UPF (CC BY-NC)