Skip to content

simonw/llm

Repository files navigation

LLM

PyPI Documentation Changelog Tests License Discord Homebrew

A CLI utility and Python library for interacting with Large Language Models, both via remote APIs and models that can be installed and run on your own machine.

Run prompts from the command-line, store the results in SQLite, generate embeddings and more.

Consult the LLM plugins directory for plugins that provide access to remote and local models.

Full documentation: llm.datasette.io

Background on this project:

Installation

Install this tool using pip:

pip install llm

Or using Homebrew:

brew install llm

Detailed installation instructions.

Getting started

If you have an OpenAI API key you can get started using the OpenAI models right away.

As an alternative to OpenAI, you can install plugins to access models by other providers, including models that can be installed and run on your own device.

Save your OpenAI API key like this:

llm keys set openai

This will prompt you for your key like so:

Enter key: <paste here>

Now that you've saved a key you can run a prompt like this:

llm "Five cute names for a pet penguin"
1. Waddles
2. Pebbles
3. Bubbles
4. Flappy
5. Chilly

Read the usage instructions for more.

Installing a model that runs on your own machine

LLM plugins can add support for alternative models, including models that run on your own machine.

To download and run Mistral 7B Instruct locally, you can install the llm-gpt4all plugin:

llm install llm-gpt4all

Then run this command to see which models it makes available:

llm models
gpt4all: all-MiniLM-L6-v2-f16 - SBert, 43.76MB download, needs 1GB RAM
gpt4all: orca-mini-3b-gguf2-q4_0 - Mini Orca (Small), 1.84GB download, needs 4GB RAM
gpt4all: mistral-7b-instruct-v0 - Mistral Instruct, 3.83GB download, needs 8GB RAM
...

Each model file will be downloaded once the first time you use it. Try Mistral out like this:

llm -m mistral-7b-instruct-v0 'difference between a pelican and a walrus'

You can also start a chat session with the model using the llm chat command:

llm chat -m mistral-7b-instruct-v0
Chatting with mistral-7b-instruct-v0
Type 'exit' or 'quit' to exit
Type '!multi' to enter multiple lines, then '!end' to finish
> 

Using a system prompt

You can use the -s/--system option to set a system prompt, providing instructions for processing other input to the tool.

To describe how the code in a file works, try this:

cat mycode.py | llm -s "Explain this code"

Help

For help, run:

llm --help

You can also use:

python -m llm --help