Docker Model Runner v1.1.37

What's Changed

🚀 Features

• feat: add CNCF ModelPack type definitions (#859) @ilopezluna

🐛 Bug Fixes

• fix: add spirv-headers dependency for Vulkan build (#872) @ilopezluna
• fix: improve pull resume reliability and progress bar display on retry (#873) @ericcurtin
• Fix responses goroutine leak (#874) @ilopezluna
• fix: preserve Range header across registry redirects for resumable pulls (#877) @ilopezluna
• fix: allow vllm-metal cache dir in macOS sandbox (#876) @ilopezluna
• fix: keep Python include headers for vllm-metal Metal kernel compilation (#876) @ilopezluna
• fix: pre-compile vllm-metal paged_ops extension in tarball build (#876) @ilopezluna
• fix: allow file-map-executable for vllm-metal cache in macOS sandbox (#876) @ilopezluna

📦 Dependencies

• chore: bump llama.cpp (https://github.com/ggml-org/llama.cpp/releases/b8851) (#867) @app/github-actions
• chore(deps): bump the go-modules-root group with 2 updates (#868) @app/dependabot
• chore(deps): bump docker/cagent-action from 1.3.4 to 1.4.1 in the github-actions group (#869) @app/dependabot
• chore: update vllm-metal version to v0.2.0 and vllm version to 0.19.1 (#876) @ilopezluna

🔧 Maintenance

• Remove dead code (#870) @ilopezluna
• chore: pin GitHub Actions to commit SHA (#871) @quentin-laplanche-docker
• refactor: normalize ModelPack config to Docker format in API responses (#875) @ilopezluna

Full Changelog: v1.1.36...v1.1.37