Fetching metadata from the HF Docker repository...
May 30 08:55 fix: chown /app to su2a for auto-update write access 08:45 fix: write update temp file to /tmp instead of /app 08:37 feat: auto-update to latest GitHub release on restart 08:12 fix: use partial unique index syntax for admin ON CONFLICT 08:06 fix: create admin user from secrets on fresh DB 07:37 fix: bypass FUSE config corruption with env vars + enable registration on start 07:20 revert: restore original entrypoint from 5565702 07:06 fix: embed config in entrypoint, use DATA_DIR=/tmp to bypass FUSE entirely 06:59 fix: set PATH for hf in root context, unconditional config download 06:52 fix: download config via HF API instead of FUSE cp to avoid corruption 06:46 fix: use DATA_DIR=/tmp to bypass FUSE config read corruption 06:37 fix: copy config off FUSE in root context to avoid permission denied 06:31 fix: copy config off FUSE mount to prevent read corruption 06:17 revert: back to single-file backup (debug RUNTIME_ERROR) 06:05 feat: split DB backup into <2MB chunks to avoid FUSE write truncation 05:55 fix: use Python bcrypt at runtime 05:42 fix: use quoted heredoc + psql -v to prevent shell expansion of bcrypt hash 05:35 fix: use DO block for admin upsert, fix settings table column names 05:28 feat: auto-create admin user and enable registration on first start via psql 05:21 revert: restore original entrypoint 05:14 fix: run admin creation after server starts (tables must exist first) 05:07 feat: auto-create admin user from ADMIN_EMAIL/ADMIN_PASSWORD secrets on startup 03:24 fix: resolve RUNTIME_ERROR caused by FUSE bucket mount conflicts - Move REDIS_DIR from /app/data/redis (FUSE bucket mount) to /tmp/redis
- Add 2>/dev/null || true to mkdir to survive FUSE mount errors under set -e
- Wrap chown in background job with 2s timeout to prevent hang on bucket mount May 29 07:49 PG local + bucket backup/restore, DB: su2adb 
