abiosoft/colima · error

failed to pull model %q: %w

Error message

failed to pull model %q: %w

What it means

EnsureModelAvailable runs `ramalama pull <model>` inside the Lima VM through guest.RunInteractive, after ramalamaModelExists reported the model absent locally; the error means that interactive shell command exited non-zero. %q carries the requested model name and %w the shell exit error from the VM.

Source

Thrown at model/ramalama.go:141

	return name
}

// EnsureRamalamaModel ensures a ramalama model is available, pulling if necessary.
func EnsureRamalamaModel(modelName string) error {
	if ramalamaModelExists(modelName) {
		return nil
	}

	// Model not found locally, pull it
	guest := lima.New(host.New())
	shellCmd := fmt.Sprintf(
		`export RAMALAMA_CONTAINER_ENGINE=docker PATH="$HOME/.local/bin:$PATH"; ramalama pull %s`,
		modelName,
	)

	if err := guest.RunInteractive("sh", "-c", shellCmd); err != nil {
		return fmt.Errorf("failed to pull model %q: %w", modelName, err)
	}

	return nil
}

// ProvisionRamalama installs ramalama and its dependencies in the VM.
func ProvisionRamalama() error {
	guest := lima.New(host.New())

	script := `set -e
export PATH="$HOME/.local/bin:$PATH"

# ensure pipx is available
sudo apt-get update -y && sudo apt-get install -y pipx

# install ramalama via pipx; upgrade if ramalama is already installed
if command -v ramalama >/dev/null 2>&1; then
  pipx upgrade ramalama

View on GitHub (pinned to c3a5f9184d)

Solutions

1. Reproduce inside the VM to see ramalama's own error: colima ssh -- sh -c 'PATH=$HOME/.local/bin:$PATH ramalama pull <model>'
2. Check VM disk headroom: colima ssh -- df -h, and grow with colima stop && colima start --disk if full
3. Verify VM internet egress and DNS: colima ssh -- curl -I https://ollama.com
4. Use the exact model identifiers listed by colima model list / ramalama ls

// sanity-check the reference shape before triggering a pull
var modelRef = regexp.MustCompile(`^(ollama://|huggingface://|hf.co/|quay.io/)[\w./:-]+$`)

func looksLikeModelRef(name string) bool {
	return modelRef.MatchString(name)
}

When it happens

Trigger: The model identifier does not exist on its source registry (typo in an ollama/hf.co/quay reference); the VM has no internet egress; registry authentication is required; the VM disk is full; the user aborts the interactive pull with Ctrl-C.

Common situations: Typo'd model names passed to colima model run; VPN cut-offs isolating the VM; VM disk (colima disk size) exhausted after pulling several multi-GB models.

Related errors

• error setting up AI model runner: %w
• error listing models: %w
• dependency check failed for VM: %w
• error retrieving current runtime: empty value
• error listing networks: %w

AI-assisted analysis of abiosoft/colima@c3a5f9184d (2026-08-15). Data as JSON: /api/errors/749f7bad25bf3921. Report an issue: GitHub.