llama-stack-client configure --endpoint http://localhost:8321 requires a server change, too.

[deanwampler]

I tried this command in https://llama-stack.readthedocs.io/en/latest/getting_started/detailed_tutorial.html#step-3-run-client-cli (including the option --api-key none). It changes ~/.llama/client/config.yaml to this, as expected:

api_key: none
endpoint: http://localhost:8321

However, llama-stack-client can no longer connect to the server:

$  uv run --with llama-stack llama-stack-client models list
╭────────────────────────────────╮
│ Failed to list models          │
│                                │
│ Error Type: APIConnectionError │
│ Details: Connection error.     │
╰────────────────────────────────╯

Actually, this is to be expected, because the server defaults to port 5001, which appears to be set in ~/.llamastackrc, so you have to edit that file:

export LLAMA_STACK_PORT=8321

Then . ~/.llamastackrc (It doesn't appear to be sourced when you start the server, true??), then rerun the server. The server output shows it's now using 8321 and the client works again.

I'm thinking of writing a PR to add this additional documentation. Make sense or is this already covered somewhere else?