Skip to content

Troubleshooting

Start with read-only commands. They usually show whether the problem is mise activation, a missing profile, or a project override:

Terminal window
mise --version
mise doctor
mise config
printf '%s\n' "${MISE_ENV:-<not set>}"
mise ls --current

Check that mise is activated in your shell startup file. For zsh, the line is:

Terminal window
eval "$(mise activate zsh)"

Use bash instead of zsh when needed. Restart the shell, or test without activation:

Terminal window
mise exec -- rg --version

mise config should list conf.d/pyahu-toolchain.toml and one config.<profile>.toml file for each installed profile. MISE_ENV must also contain the profile name:

Terminal window
export MISE_ENV=workstation,node,cloud
mise install

Kimi and Pi need both node and ai. The AI profile is rolling, so install it without --locked.

Preview the exact destination first:

Terminal window
./install.sh --dry-run workstation node

The destination is $MISE_CONFIG_DIR when set. Otherwise it is ${XDG_CONFIG_HOME:-$HOME/.config}/mise.

The installer will not replace an unrelated file or symlink. Inspect and move it yourself, or use --force to preserve it beside the destination as .bak, .bak.1, and so on. Do not delete a configuration you do not recognize just to make installation pass.

Check the profile prerequisites, network or proxy access, free disk space, and the upstream service. Then inspect the selected backend and retry with useful diagnostics:

Terminal window
mise config
mise ls-remote TOOL
mise install --verbose TOOL@VERSION

For a stable project checkout, keep the reviewed lockfile intact:

Terminal window
git status --short
git pull --ff-only
mise install --locked

Do not regenerate a release lockfile merely to bypass a failed download.

Run mise config from the directory where the problem happens. A project-level mise.toml normally wins over the global Pyahu setup. That is expected and lets each repository choose its own runtime.

Cloud or local-model tools fail after installation

Section titled “Cloud or local-model tools fail after installation”
  • kind, k3d, and pyahu up need a running Docker engine.
  • Kubernetes and cloud CLIs need credentials and contexts configured outside this project.
  • Ollama model requirements depend on the model and machine; installing the CLI does not download a model.
  • kubectl ctx and kubectl ns need the cloud profile and the repository bin directory in the mise environment.

Switch a versioned checkout to an older release. Existing installer links follow the checkout:

Terminal window
git fetch --tags
git switch --detach vX.Y.Z
mise install

Prefer a known-good tag over editing a generated lockfile.

Preview first, then remove only links owned by this checkout:

Terminal window
./install.sh --dry-run --uninstall
./install.sh --uninstall

Backups created by the installer are restored. Downloaded tools stay in the mise cache but are no longer active.