v0.6.19 v2.3.0
1Panel-dev/1PanelSep 10, 2026 → Sep 14, 20264 days between releases
What's New
Adds Ascend/AMD GPU support, SSH session preservation, and new AI providers like llmman.
Breaking Changes
- Rebuilt firewall system requires manual rule reconfiguration
New Features
- Added support for Ascend 910B and AMD accelerators in GPU monitoring
- Added llmman as a local model provider for model accounts
- Added DashScope text-to-image API support for custom model accounts
- Implemented SSH session preservation to recover connections after disconnections
- Added support for bulk environment variable imports in Runtime
Improvements
- OpenResty Brotli automatic compression when enabled
- Optimized Gzip defaults for websites
- Added support for changing the version of installed runtimes
- Added support for selecting Ascend device IDs in vLLM
- Added an Ascend Qwen3.6-35B-A3B model startup template for vLLM
Upgrade Notes
The firewall system has been rebuilt, so existing firewall rules may need manual reconfiguration. Users should back up environment variables before switching runtime versions. Note that new AI provider defaults may differ from previous configurations.
The Two Releases
v0.6.19
Sep 10, 2026· 1 month ago
This release focuses on stabilizing agent operations with bounded finalization and manual OAuth selection, while introducing support for Langfuse immutable spans and updating default AI provider configurations.
Full release notesv2.3.0
Sep 14, 2026· 1 month ago
This release introduces significant advancements in AI/LLM management, including support for Ascend and AMD accelerators, and new model providers. It also enhances infrastructure management with a rebuilt firewall system and SSH session preservation, alongside various UI and runtime improvements.
Full release notes