fauxberry CLI

A Node.js CLI application that implements a "fauxberry" style thinking loop using Anthropic's Claude AI model. This tool helps break down complex tasks into smaller reasoning steps before providing a final response. It simulates a stripped down version of the reasoning loop found in OpenAI's o1 family of models.

Features

• Implements the fauxberry thinking methodology
• Supports Claude 3.5 and other Anthropic models
• Accepts tasks via command line or file input
• Configurable API key via environment variable or command line
• Detailed step-by-step reasoning process
• Customizable model selection

Installation

Global Installation

npm install -g fauxberry-cli

Local Installation

git clone https://github.com/Stevenic/convo.git
cd convo/examples/fauxberry-cli
npm install

Configuration

Before using fauxberry, you need an Anthropic API key. You can provide it in two ways:

1. Environment variable: Create a .env file in the project root:

   ANTHROPIC_API_KEY=your_anthropic_api_key

2. Command line argument:

   fauxberry-cli --apiKey "your_anthropic_api_key"

Usage

Basic Usage

fauxberry-cli --task "count backwards from 100 to 1 skipping every 3rd and 14th number"

Using a File Input

fauxberry-cli --file "path/to/task.txt"

Specifying a Different Model

fauxberry-cli --task "Your task" --model "claude-3-5-sonnet-20240620"

Command Line Options

Options:
  --apiKey, -k     Anthropic API key                             [string]
  --model, -m      Model to use                                  [string] [default: "claude-3-5-sonnet-20241022"]
  --task, -t       Task to perform                               [string]
  --file, -f       File containing the task                      [string]
  --help           Show help                                     [boolean]

Example

$ fauxberry-cli --task "count backwards from 100 to 1 skipping every 3rd and 14th number"