Run a model with hbay

1. Check the requirements

Open Terminal on macOS or Linux. You need curl and OpenSSL 3 or newer. Check:

curl --version
openssl version

If the second command reports LibreSSL or an older OpenSSL, follow the install requirements before continuing. Windows signed installation is not published; follow the Windows guide.

2. Install the native tool

The live install page publishes this command. It runs the installer, which verifies the release signature and archive hash:

curl -fsSL https://huggingbay.xyz/cli/install.sh | sh
export PATH="$HOME/.local/bin:$PATH"
hbay --version

The PATH line applies to this terminal session. If installation refuses a signature or prerequisite check, stop and use the troubleshooting guide. The earlier Node CLI packages are a different tool.

3. Inspect and download a small model

These commands use a model record checked on 17 October 2026:

hbay catalog "Qwen2.5-0.5B-Instruct-GGUF"
hbay info hf-model-qwen-qwen2-5-0-5b-instruct-gguf
hbay download hf-model-qwen-qwen2-5-0-5b-instruct-gguf

Read the license, chosen file and current readiness before downloading. Save the path and SHA-256 printed after success. If there is no ready file, return to the catalog; availability can change.

4. Run one prompt

In the next command, replace /path/from/download.gguf with the saved path and HASH_FROM_DOWNLOAD with the full 64-character SHA-256 from that download. Keep the quotes:

hbay prompt "/path/from/download.gguf" --sha256 "HASH_FROM_DOWNLOAD" --prompt "Explain a rainbow in three short sentences." --max-tokens 96

hbay checks the hash before loading. A successful prompt prints a reply. This is a CPU runner; a large model may be slow. hbay help prompt shows the options in your installed version.

About hbay run

The source docs describe hbay run as a 0.2.0 candidate requiring HBAY_ENABLE_RUN=1. This guide uses the documented download and prompt commands. A published installer does not prove that the one-shot candidate is in your installed release.