[DevelopmentCats]

Moves the Unreleased section of CHANGELOG.md to ## [0.6.0] - 2026-11-04, adds a fresh empty Unreleased section and updates the compare links.

0.6.0 is a minor bump because a macOS system-scope LaunchDaemon installed by 0.5.0 has to be removed and set up again (docs/macos.md has the steps). The release also renames the tool to llm-helper, adds the MLX and llama.cpp engines, vllm-metal on macOS, and the catalogue of engines, recipes and hardware.

Upgrade from 0.5.0

• Profiles written by 0.5.0 read the same. The only new key is the optional PROVIDER, and a profile without it runs on vLLM.
• Of the service units, only the systemd unit's Description text changed.
• 0.5.0's upgrade matches archives by their _os_arch.tar.gz suffix and asks coder/vllm-helper, which redirects to this repository. Checked with upgrade --check on both channels. The archives carry the binary as both llm-helper and vllm-helper.

Checks

prettier, go vet, go test -race ./... and scripts/gendocs (no diff). The release notes preview from RELEASING.md prints the full 0.6.0 section.

After this merges, tagging v0.6.0 runs the release workflow, which publishes to coder/llm-helper (#56).

[Copilot]

🟢 Approval recommended

There are no active review issues for the 0.6.0 changelog update.

No open issues 🎉

What changed in this PR

Prepares the changelog for the 0.6.0 minor release, covering the expanded inference-engine support, llm-helper rename, and macOS migration requirement.

Changes:

• Moves pending release notes into a dated 0.6.0 section and leaves an empty Unreleased section.
• Updates version comparison links.
• Clarifies that hardware diagnostics warn when multiple definitions match a GPU.

File	Description
CHANGELOG.md	Finalizes 0.6.0 release notes, updates comparison links, and clarifies a hardware warning.

_{Review effort: Balanced}