# Upgrading [CHANGELOG.md](../CHANGELOG.md) lists every change. Notes for bigger steps: ## To 1.0 - Nothing to do. Your `PLEX_TOKEN`, `PLEXMIND_API_KEY` and webhook secret keep working and win over the new defaults. You can now sign in to the dashboard with Plex as the server owner. - **Unraid:** if you installed from the template before v0.9.15, edit the container once and set Repository to `ghcr.io/finechainesium/plexmind:latest`, so Unraid offers updates. ## To v0.9.0: one container v0.9.0 replaced the four PlexMind containers (`plexmind`, `plexmind-recommendation-worker`, `plexmind-scripts`, `plexmind-docker-broker`) with one `plexmind` container. - **Compose:** `docker compose up -d --remove-orphans`. - **Unraid or hand-made containers:** remove the four old containers, then create `plexmind` from the template with the old scripts worker's media mounts, the Docker socket and `--stop-timeout=40`. Keep the same `/app/data` folder. - `docker exec plexmind-scripts ...` becomes `docker exec plexmind ...`. ## To v0.8.22 - llama.cpp and Whisper host ports bind to `127.0.0.1`. If Bazarr or another LAN client uses them, set `LLAMA_CPP_BIND_ADDRESS` / `WHISPER_BIND_ADDRESS` to `0.0.0.0`. - Translation sends 8 cues per request (`TRANSLATION_BATCH_CUES`); `1` restores one per request. - Rotate `PLEXMIND_WEBHOOK_SECRET` and the TMDB key if old container logs could have been read: earlier versions logged both. ## Version numbers The project briefly published `v2.1.0` and `v2.1.1` while it was still being hardened. Those numbers overstated its maturity, so the line was reset to `v0.7.0` / `v0.7.1` and continued from there to 1.0. That was a renumbering, not a code rollback.