Make launching RamaLama boring 🚀

RamaLama is a Python-based CLI tool that may require additional Python dependencies and setup, especially on Windows. Native support on Windows can be tricky due to path and environment assumptions.

To simplify this experience, rmlm CLI acts as a thin wrapper that runs RamaLama CLI using a containerized approach with Podman to ensure consistency across platforms.

Head to the Releases Page and download the binary for your platform.

oneliner (assuming you have the jq utility) that will fetch the latest macOS binary and rename it to rmlm

curl -L $(curl -s https://api.github.com/repos/benoitf/rmlm/releases/latest | jq -r '.assets[] | select(.name | test("rmlm-mac-universal")) | .browser_download_url') -o rmlm && chmod u+x rmlm

Option: copy the binary to /usr/local/bin

Note: The binary is not code-signed. You can either build it yourself or remove the quarantine attribute manually:

xattr -d com.apple.quarantine rmlm-mac-arm64

oneliner (for PowerShell):

Invoke-WebRequest -Uri ((Invoke-RestMethod -Uri "https://api.github.com/repos/benoitf/rmlm/releases/latest" -Headers @{ "User-Agent" = "PowerShell" }).assets | Where-Object { $_.name -eq "rmlm-win-x64.exe" }).browser_download_url -OutFile "rmlm.exe"

oneliner (assuming you have the jq utility) that will fetch the latest Linux binary based on your arch and rename it to rmlm

curl -L $(curl -s https://api.github.com/repos/benoitf/rmlm/releases/latest | jq -r --arg arch "$(uname -m)" '.assets[] | select(.name | test("rmlm-linux-" + ($arch | if . == "aarch64" then "arm64" else "x64" end))) | .browser_download_url') -o rmlm && chmod u+x rmlm

Option: copy the binary to /usr/local/bin

• A working Podman installation on Windows or macOS.
  • Podman should be installed and configured to run containers (e.g., rootless or with a virtual machine on macOS/Windows).

Once installed, you can use the rmlm command just like you would use ramalama:

./rmlm <RamaLama command>

This command launches a small LLM and gives you the familiar interactive RamaLama prompt:

rmlm run tinyllama
🦭 >

Start a model server and access it from your browser:

rmlm serve tinyllama | grep http://
main: server is listening on http://0.0.0.0:8080 - starting the main loop

Then open the displayed URL (e.g., http://0.0.0.0:8080) to chat with the model interactively in your browser.

If you’d like to help improve RMLM, check out CONTRIBUTING.md for setup instructions, development tips, and contribution guidelines.

Whether it’s fixing bugs, improving documentation, or adding new features, all kinds of contributions are welcome! 💡

This project is licensed under the Apache-2 Liense.

Open an issue or start a discussion