Metadata-Version: 2.2
Name: llm-council
Version: 0.1.2
Summary: Use LLM to generate and execute commands in your shell
Author: Simon Willison
License: MIT
Project-URL: homepage, https://github.com/nuwandavek/llm-council
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: llm
Requires-Dist: prompt_toolkit>=3.0.43
Requires-Dist: llm-claude-3
Requires-Dist: llm-gemini

# llm-council

Get a council of LLMs to advise consult for you!

## Installation

This plugin should be installed in the same environment as [LLM](https://llm.datasette.io/).
```bash
llm install llm-council
```

![council](https://raw.githubusercontent.com/nuwandavek/llm-council/refs/heads/master/assets/council.png)

### Supported models/providers
The models themselves are fixed as of now with:
- `openai`: `gpt-4o`
- `anthropic`: `clause-3-5-sonnet-latest`
- `gemini`: `gemini-1.5-flash-latest`

The necessary `llm plugins` are already installed. But you still need to set the keys
```bash
llm keys set openai
llm keys set claude
llm keys set gemini
```

## Usage

I usually run every query on all LLMs just to see what they have to say. And I love the llm library. You can now assemble your own council of advisors by simply running `llm council` like this:

```bash
llm council 'whats the california traffic law around double white lines?'
```

By default, it uses `openai` and `anthropic`. But you can specify the providers by:

```bash
llm council -p openai -p anthropic 'tell me a joke'
```
Press Q or Ctrl + C to exit.

## The system prompt

This is the prompt used by this tool:

> Keep your answers brief and to the point.

Feel free to modify it by passing the `--system` arg.

## Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:
```bash
cd llm-council
uv venv
source .venv/bin/activate

uv pip install -r pyproject.toml
```
Now install the plugin with:
```bash
llm install -e .
```
