#LLM prompt somehow shares LLM API credentials across shell sessions..?

1 messages · Page 1 of 1 (latest)

lofty sage
#

Terminal 1:

$ dagger shell
Dagger interactive shell. Type ".help" for more information. Press Ctrl+D to exit.

claude-3-7-sonnet-20250219 ❯

Terminal 2

$ ( unset ANTHROPIC_API_KEY; unset ANTHROPIC_MODEL; env ) | grep ANTH
↑14:04:15 [0 1]
$ ( unset ANTHROPIC_API_KEY; unset ANTHROPIC_MODEL; dagger shell ) 
Dagger interactive shell. Type ".help" for more information. Press Ctrl+D to exit.

claude-3-7-sonnet-20250219 ❯

In parallel I do have a dagger shell session that IS using claude. I want to have another shell in parallel that would be using OpenAI. For some reason the new shell still uses claude, even though I unset all the env vars.

BUT if I exit the first shell (the one that I wanted to be using claude), then the second shell starts as expected -- using OpenAI.

What's happening...? I'm guessing this is a bug?

crystal lake
#

We're looking into it -- it's definitely a bug ahah 👀 🙏

lofty sage
#

Thanks! Btw, how do I step out of the prompt back into shell? W/o exitting dagger that is.

crystal lake
crystal lake