Troubleshooting
Use the real HLVM paths when diagnosing failures. Do not bypass bootstrap by starting a separate system Ollama instance.
First Checks
hlvm --version
hlvm doctor
hlvm repair
hlvm ask --verbose "hello"
Runtime Fails To Start
Run:
hlvm doctor
hlvm repair
repair restores managed runtime assets. If the error still mentions
127.0.0.1:11439, another process may be occupying HLVM's managed local AI
endpoint.
Model Download Fails
Use the model command so the download stays inside HLVM's managed runtime:
hlvm model pull ollama/gemma4:12b
Then retry:
hlvm ask --verbose "hello"
Browser Bridge Fails
hlvm browser setup
hlvm browser status
Restart Chrome after setup. If status remains disconnected, open the extension
popup once and rerun hlvm browser status.
MCP Server Fails
hlvm mcp show <name>
hlvm mcp login <name>
For stdio servers, confirm the command works outside HLVM and that the hlvm mcp add command used -- before the stdio command.
HQL Command Fails
Use HQL's own docs:
hql run hello.hql
hql explain <CODE>