Update CHANGELOG.md (#17916 )

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Fix URL for downloading OneDev latest release (#17915 )
2026-10-13 19:40:25 -04:00 · 2026-10-13 22:56:14 +00:00 · 2026-10-14 00:55:44 +02:00 · 2026-10-13 16:10:40 +00:00 · 2026-10-13 18:10:07 +02:00
5 changed files with 19 additions and 2 deletions
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -568,6 +568,8 @@ Exercise vigilance regarding copycat or coat-tailing sites that seek to exploit

  - #### 🐞 Bug Fixes

+    - Fix URL for downloading OneDev latest release [@ryansully](https://github.com/ryansully) ([#17915](https://github.com/community-scripts/ProxmoxVE/pull/17915))
+    - LocalAI: pin the CPU backends when the container has no GPU runtime [@MickLesk](https://github.com/MickLesk) ([#17909](https://github.com/community-scripts/ProxmoxVE/pull/17909))
    - [Gluetun] Update GitHub repository source install and update. [@CerberusStyle](https://github.com/CerberusStyle) ([#17899](https://github.com/community-scripts/ProxmoxVE/pull/17899))
    - Immich: Bump to 3.3.1 and restore (failed) geodata/countryInfo.txt on update  [@MickLesk](https://github.com/MickLesk) ([#17894](https://github.com/community-scripts/ProxmoxVE/pull/17894))
    - WikiJS: use v2 Repo until v3 is stable [@MickLesk](https://github.com/MickLesk) ([#17895](https://github.com/community-scripts/ProxmoxVE/pull/17895))
--- a/ct/localai.sh
+++ b/ct/localai.sh
@@ -37,6 +37,15 @@ function update_script() {
    exit
  fi

+  # Older installs have no CPU pin; on a host with a GPU but no driver in the
+  # container every model then ends in "backend not found".
+  if ! grep -q '^LOCALAI_FORCE_META_BACKEND_CAPABILITY=' /opt/localai.env 2>/dev/null &&
+    ! ldconfig -p 2>/dev/null | grep -qE 'libcuda[.]so|libamdhip64[.]so|libze_intel_gpu[.]so|libze_loader[.]so'; then
+    echo "LOCALAI_FORCE_META_BACKEND_CAPABILITY=default" >>/opt/localai.env
+    systemctl restart localai
+    msg_ok "Pinned LocalAI to the CPU backends (no GPU runtime in the container)"
+  fi
+
  if check_for_gh_release "localai" "mudler/LocalAI"; then
    msg_info "Stopping Service"
    systemctl stop localai
--- a/ct/onedev.sh
+++ b/ct/onedev.sh
@@ -39,7 +39,7 @@ function update_script() {

    msg_info "Updating OneDev"
    cd /opt
-    curl -fsSL "https://code.onedev.io/onedev/server/~site/onedev-latest.tar.gz" -o onedev-latest.tar.gz
+    curl -fsSL "https://code.onedev.io/onedev/server/~sites/public/onedev-latest.tar.gz" -o onedev-latest.tar.gz
    tar -xzf onedev-latest.tar.gz
    $STD /opt/onedev-latest/bin/upgrade.sh /opt/onedev
    rm -rf /opt/onedev-latest
--- a/install/localai-install.sh
+++ b/install/localai-install.sh
@@ -39,6 +39,12 @@ LOCALAI_MODELS_PATH=/opt/localai_data/models
 LOCALAI_BACKENDS_PATH=/opt/localai_data/backends
 LOCALAI_LOG_LEVEL=info
 EOF
+# LocalAI picks the backend variant from the GPU it sees in sysfs, which is the
+# host's card even when no driver was installed in here, and a CUDA/ROCm/SYCL
+# backend then never starts. Pin the CPU variants unless a runtime is present.
+if ! ldconfig -p 2>/dev/null | grep -qE 'libcuda[.]so|libamdhip64[.]so|libze_intel_gpu[.]so|libze_loader[.]so'; then
+  echo "LOCALAI_FORCE_META_BACKEND_CAPABILITY=default" >>/opt/localai.env
+fi
 # LocalAI refuses to start on a wildcard bind with nothing to identify callers,
 # and binding the wildcard is what makes the container reachable at all.
 if [[ "$var_auth" == "yes" ]]; then
--- a/install/onedev-install.sh
+++ b/install/onedev-install.sh
@@ -24,7 +24,7 @@ JAVA_VERSION="21" setup_java
 msg_info "Installing OneDev"
 RELEASE=$(get_latest_github_release "theonedev/onedev" false)
 cd /opt
-curl -fsSL "https://code.onedev.io/onedev/server/~site/onedev-latest.tar.gz" -o onedev-latest.tar.gz
+curl -fsSL "https://code.onedev.io/onedev/server/~sites/public/onedev-latest.tar.gz" -o onedev-latest.tar.gz
 tar -xzf onedev-latest.tar.gz
 mv /opt/onedev-latest /opt/onedev
 $STD /opt/onedev/bin/server.sh install
Author	SHA1	Message	Date
community-scripts-pr-app[bot]	bbb81b0d21	Update CHANGELOG.md (#17916 ) Some checks are pending Create Changelog Pull Request / update-changelog-pull-request (push) Waiting to run Details Set state to is_deleted in pocketbase / delete-pocketbase-entry (push) Waiting to run Details Sync ct/install to Incus / dispatch (push) Waiting to run Details Update script timestamp on .sh changes / update-script-timestamp (push) Waiting to run Details Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>	2026-10-13 22:56:14 +00:00
Ryan Sullivan	ecbc302c68	Fix URL for downloading OneDev latest release (#17915 ) Published sites use a new path, per https://code.onedev.io/onedev/server/~issues/3179#IssueComment-11050	2026-10-14 00:55:44 +02:00
community-scripts-pr-app[bot]	1e4b435351	Update CHANGELOG.md (#17910 ) Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>	2026-10-13 16:10:40 +00:00
CanbiZ (MickLesk)	6234161462	LocalAI: pin the CPU backends when the container has no GPU runtime (#17909 ) LocalAI resolves meta backends from the GPU vendor it reads out of sysfs, which is the host's card even with no driver installed in the container, so llama-cpp became cuda12-llama-cpp and never started (#17908). Force the default capability unless a CUDA, ROCm or Level Zero runtime is present; update adds the pin to existing installs.	2026-10-13 18:10:07 +02:00