Tip

After using this for a few months I developed some opinions on how it should work that are better served by my new ai-completion bash/zsh function. Consider using that instead.

aicmd

use AI to generate swanky bash one-liners (supports Anthropic, OpenAI, and Ollama!)

usage

$ aicmd "find duplicate files in the current directory recursively, show their sizes, and group them by content"
generated command: find . -type f -exec sha256sum {} \; | sort | awk '{ if (last == $1) { if (count == 1) print prev; print $0; count++ } else { last = $1; prev = $0; count = 1 } }' | while read -r hash file; do size=$(du -h "$file" | cut -f1); echo -e "$size\t$file"; done | sort -h | awk '{ if (NR==1 || prev != $1) print "\nSize: " $1; prev=$1; print "  → " $2 }'

run it now? [Y/n]: y

Size: 4.2M
  → ./photos/vacation2023/IMG_1234.jpg
  → ./backup/old_photos/IMG_1234.jpg

Size: 8.7M
  → ./downloads/project-v1.0.0.tar.gz
  → ./archives/project-v1.0.0.tar.gz
  → ./desktop/project-backup.tar.gz

Size: 15M
  → ./videos/screencast.mp4
  → ./uploads/demo_video.mp4

installing

Pre-built binaries are available for Linux (AMD64/ARM64) and macOS (ARM64) on the releases page. Download them and put them in your path or see below to build them.

Set API keys must be set as environment variables, if multiple are set aicmd prefers anthropic, then openai, then ollama. You can also pass --provider <anthropic|openai|ollama> and --model <whatever>.

export ANTHROPIC_API_KEY="your-api-key"
aicmd "find duplicate files"

export OPENAI_API_KEY="your-api-key"
aicmd "find duplicate files"

export OLLAMA_API_BASE="http://localhost:11434"
export OLLAMA_MODEL="mistral"  # or any other model you've pulled
aicmd "find duplicate files"

building

1. Install Go: Make sure Go is installed. Download it here.

2. Get the Code: Clone the repo and navigate into it.

   git clone <repo-url>
   cd <repo-directory>

3. Set API Key: Choose your AI provider:

   For Anthropic (default):

   export ANTHROPIC_API_KEY="your-api-key"

   For OpenAI:

   export OPENAI_API_KEY="your-api-key"

   For Ollama:

   export OLLAMA_API_BASE="http://localhost:11434"
   export OLLAMA_MODEL="mistral"  # or whatever model you prefer

   You can also specify the provider explicitly with the --provider flag:

   aicmd --provider openai "your command"
   aicmd --provider anthropic "your command"
   aicmd --provider ollama "your command"

4. Build: Run the build command.

   go build -o aicmd main.go

5. Run: Use the tool.

   ./aicmd "describe your task"

6. Optional: Move aicmd to a directory in your PATH for easy access.

   sudo mv aicmd /usr/local/bin/