I'm not quite sure what this is exactly.

It's a toy. Maybe a toddler's first coding agent. A poor man's video generator. An experiment in isolating an LLM in a way that it is fun and harmless.

You type what you want to see. A 64x32 character grid comes alive. Each prompt builds on what's already running. The model sees the current code and extends it. /clear to start over.

I would love to see someone livecode VJing with droeftoeter on a beamer during an algorave.

https://youtu.be/-atrhQ-9Vy8

Binaries at Releases.

Or build from source:

go build -o droeftoeter .

On first launch you pick a provider. Works with:

• Groq (free) — Llama models, fast
• Gemini (free) — Google's API
• OpenAI-compatible — any endpoint (OpenRouter, DeepSeek, etc.)
• Anthropic — Claude
• Ollama (local) — your own hardware, no API key

Change anytime with /config, or set environment variables:

DROEFTOETER_PROVIDER=openai
DROEFTOETER_API_KEY=your-key
DROEFTOETER_BASE_URL=https://api.groq.com/openai/v1
DROEFTOETER_MODEL=llama-3.3-70b-versatile

Or use a config.toml (see config.toml.example).

MIT