llm

This is alpha software. It is under active development, and its interface is unstable.

Overview

llm is a CLI wrapper for the OpenAI API. (Support for locally-hosted LLMs with a OpenAI-conformant interface is coming soon). Its purpose is to make LLM integrations convenient to those with a shell-centric workflow.

llm allows for the creation of multiple long-lived, user-created "sessions" (which correspond with a "chat" in the GPT-4 web interface), each of which may have a distinct "system" prompt.

Conversation history is stored in a local SQLite3 database, and thus searchable locally. Additionally, syntax highlighting is enabled by default.

Features

1. Shell-centric workflow that integrates well with shell-based text editors like vim and emacs
2. Ability to store and search conversation history locally
3. Syntax highlighting
4. Ability to "rewind" and "copy" sessions
5. Ability to "stack" and batch messages

Configuring

Export an OPENAI_KEY environment variable containing a valid API key:

export OPENAI_KEY=your-api-key

Usage

Usage:
    llm [options] db (path|destroy)
    llm [options] (q|question) (add|ask) [<msg>]
    llm [options] (q|question) (edit|rm) <id>
    llm [options] (q|question) send
    llm [options] (s|session) (new|edit|view|rm|use) <name>
    llm [options] (s|session) cp <orig> <name> [--all]
    llm [options] (s|session) log [<num>]
    llm [options] (s|session) ls
    llm [options] (s|session) rewind <id>
    llm [options] (s|session) search <string>
    llm [options] (s|session) truncate

Example workflows

Simple questions may be asked directly inline:

$ llm q ask 'Please demonstrate how to install a package on Debian Linux.'

Longer questions may be composed in your preferred editor:

$ llm q ask
# This will open $EDITOR or $VISUAL. Compose your question in the shell, then
save the document. On write and close, the question will be sent to the OpenAI
API.