Pipes like a Unix tool
Reads piped stdin as context and prints the answer to stdout. The positional prompt is the instruction; stdin is what it operates on — combined into one user message, prompt first.
heyllm
Talk to an LLM from your terminal. heyllm is a tiny, dependency-light CLI
that streams answers from any OpenAI-compatible Chat Completions API straight to
stdout — built for daily use and pipelines.
You ask a question, you get an answer. No plugins, no config files, no chat history — just one command that pipes well. And because it reads piped input as context, it slots naturally into shell workflows: hand it a diff, a log, or a file plus an instruction, and let the model do the rest.
$ heyllm "explain monads simply"A monad is a way to wrap a value along with a recipe for chaining...
$ git diff | heyllm "write a conventional commit message"feat(parser): handle SSE frames split across network chunksThe prompt is the instruction; piped stdin is the material it works on. That one
idea — <context> | heyllm "<instruction>" — is most of what makes heyllm
useful day to day.
Any OpenAI-compatible API
Speaks the OpenAI Chat Completions wire format, so --base-url points it at
OpenAI, OpenRouter, a local Ollama, or anything else that implements it.
Streams by default
Tokens print as they arrive, parsed from server-sent events. Pass
--no-stream when you’d rather wait for the full response in one shot.
Tiny & auditable
One dependency, no config files, no hidden state. A small binary that does
the obvious thing — and --json hands you the raw response for scripting.
Next steps
Section titled “Next steps” Quick start Install, set your API key, and run your first prompt in four steps.
CLI reference Every flag, stdin behavior, streaming, exit codes, and worked examples.
Configuration API keys, OpenAI-compatible providers, and model selection.
Recipes Useful one-liners — commit messages, command explainers, file summaries, and more.