First Run

The first useful test is one AI prompt through the real managed runtime path.

hlvm ask "summarize the main points of REST in five bullets"

On a fresh machine, HLVM may prepare its managed local runtime and download the default local model before the answer appears.

What Happens

1. HLVM starts or connects to the local runtime host on 127.0.0.1:11435.
2. The host prepares the managed local AI endpoint on 127.0.0.1:11439.
3. The bootstrap selector chooses the local model for the machine.
4. The agent runs the request and streams the answer.

The current local model tier map is source-owned by src/hlvm/runtime/embedded/model-tiers.json: gemma4:12b at 16 GiB RAM and above, gemma4:e2b below that.

Expected Result

A successful run prints only the answer by default. Use --verbose when you want to inspect agent routing, tool labels, and trace output:

hlvm ask --verbose "summarize the main points of REST"

If It Fails

Run the checks in order:

hlvm doctor
hlvm repair
hlvm ask --verbose "hello"

Do not start a system Ollama process on port 11434 to bypass the issue. HLVM's default local AI path is the managed runtime.

Next

• Ask
• Models
• Troubleshooting