code.oscarkilo.com/klex-git/one/README.md

one: runs an LLM on one input

Before running this script, create a klex.json file in the root of your git repo, with these contents:

{
  "project_name": "**name**",
  "owner_username": "**your oscarkilo.com username**",
  "reader_username": "**your oscarkilo.com username**",
  "datasets_dir": "**optional**",
  "klex_url": "https://las.oscarkilo.com/klex",
  "api_key_file": "klex.key"
}

Then put your Klex API key into the klex.key file.

Then run this command once:

go get oscarkilo.com/klex-git

Now you’re ready for the examples below.

Hello World example

go run oscarkilo.com/klex-git/one <<EOF
{
  "model": "Any big LLM",
  "messages": [{
    "role": "user",
    "content": [{
      "type": "text",
      "text": "Which band played Hotel California in Lebowski?"
    }]
  }]
}

EOF

Reading prompts from files

echo "Tell me it's going to rain tomorrow." > prompt.txt
echo "You're a folksy pirate; speak like it." > system.txt
echo "{}" | go run oscarkilo.com/klex-git/one \
  -model=llama3 \
  -system_file=system.txt \
  -prompt_file=prompt.txt
rm prompt.txt system.txt

Analyzing images

echo "{}" | go run oscarkilo.com/klex-git/one \
  -model=gpt-4o \
  -image_file=<(curl -s https://upload.wikimedia.org/wikipedia/commons/thumb/3/35/Tux.svg/1024px-Tux.svg.png) \
  -prompt_file=<(echo "What is this guy's name?")