Show HN: Wut – a CLI that explains the output of your last command with an LLM (opens in new tab)

(github.com)

3 pointsjshobrook1y ago3 comments

Just type wut and an LLM will explain whatever's in your terminal. You’ll be surprised how useful this is. I use it mainly to understand and debug stack traces, but there are a bunch of other use cases:

- Deciphering error codes

- Fixing incorrect commands

- Summarizing logs

You can connect wut to OpenAI, Anthropic, or a local model via Ollama.

I hope y’all find this as helpful as I do!

3 comments

jas391y ago

Very useful tool. The tmux/screen requirement is a design flaw though.

jshobrookOP1y ago

Unfortunately, I don't see a way around it. There's no other way to capture the output buffer.

jas391y ago

Of course there is; tmux does it. Try redirecting the file descriptor so that you can capture stdout. For bash, something like: 'exec > >(tee -a "$LOG_FILE") 2>&1'.

This could be injected in .bashrc, and activated by default or by a bash function. The >(...) syntax is called process substitution.

j / k navigate · click thread line to collapse

Show HN: Wut – a CLI that explains the output of your last command with an LLM (opens in new tab)

(github.com)

3 pointsjshobrook1y ago3 comments

- Deciphering error codes

- Fixing incorrect commands

- Summarizing logs

You can connect wut to OpenAI, Anthropic, or a local model via Ollama.

I hope y’all find this as helpful as I do!

3 comments

jas391y ago

Very useful tool. The tmux/screen requirement is a design flaw though.

jshobrookOP1y ago

Unfortunately, I don't see a way around it. There's no other way to capture the output buffer.

jas391y ago

Of course there is; tmux does it. Try redirecting the file descriptor so that you can capture stdout. For bash, something like: 'exec > >(tee -a "$LOG_FILE") 2>&1'.

This could be injected in .bashrc, and activated by default or by a bash function. The >(...) syntax is called process substitution.

j / k navigate · click thread line to collapse