Bug: --v2 flag disables --host flag

[FordUniver]

Contact Details

No response

What happened?

It seems like the --v2 flag prohibits the use of --host in 0.9.0 (works without --v2.

Version

llamafile v0.9.0

What operating system are you seeing the problem on?

Linux

Relevant log output

uname@hostname:~$ ./llamafile --server --host 0.0.0.0 -m model.gguf --v2
./llamafile: --host unknown argument

[cjpais]

Try -l 0.0.0.0:<port>

This should be in the docs (-h)

[FordUniver]

Getting error: invalid parameter for argument: -l? The help menu and the server README.md also seem to only list --host and --port, or am I missing something?

[cjpais]

This works fine for me

./main -m ~/models/deepseek-distil/DeepSeek-R1-Distill-Llama-8B-Q4_K_M.gguf -l 0.0.0.0:8081 --server --v2

Though it does look like the wrong help menu is displayed. I found the correct param here for v2.

[FordUniver]

Right, I was forgetting the --v2

[cjpais]

Closing

[vadi2]

To enable programmatic launch from a C++ Qt application, I launch the /bin/sh binary with arguments /home/path/llava-v1.5-7b-q4.llamafile --server --v2 -l 127.0.0.1:8080 --gpu auto -ngl 9999 - and it seems that -l isn't recognised when launching this way:

/home/path/llava-v1.5-7b-q4.llamafile: -l 127.0.0.1:8080 unknown argument

This works fine when I use the complete command in a terminal though:

/bin/sh /home/path/llava-v1.5-7b-q4.llamafile --server --v2 -l 127.0.0.1:8080 --gpu auto -ngl 9999