woodwork-engine

Welcome to woodwork-engine, an AI Agent IaC tool that aims to make developing and deploying AI Agents easier.

Through defining components in a configuration language, an LLM will decompose the task into actionable steps, which can be executed using the supplied tools. We use latest research to inform design decisions, and we implement this as most of the setup is copy/paste across projects. Through only focussing on the necessary components of a system, this package should make designing custom, vertical agents much easier.

Table of Contents

• Features
• Installation
• Usage
• Examples
• Contributing
• License

Features

• A custom config language, woodwork (.ww files), allowing agent components to be declared
• Integrations and communication between components are handled
• Additional customisation or extension can be provided by implementing some of our interfaces

A roadmap is provided with details on future features.

Installation

1. Run pip install woodwork-engine: This gives access to the woodwork CLI tool, along with the ability to parse and deploy AI Agent components from .ww files
2. Install the Woodwork extension on VSCode if relevant: This provides syntax highlighting and intellisense for code in .ww files

Usage

1. Begin by duplicating the .env.example file and placing it in your project's root directory
2. Rename the file to .env (ensure this is part of your project's .gitignore so that it isn't committed)
3. Depending on the .ww configuration file, populate the .env file with your corresponding keys

See the .env.example file for further details.

Once you've configured your .ww config file and your .env file, there are two ways to run woodwork: A standalone application, or used as a dependency.

Standalone Application

1. Create a main.ww file and write some code: This file is where component declarations are read from. For some inspiration, consult the examples
2. Run woodwork init: This installs the necessary dependencies to run your components
3. Run woodwork: This activates the components and initializes a logger

As A Dependency

When using woodwork as a dependency, you will need to build your own logger implementation. Not building your own logger will result in no logs being generated but the application will still run.

1. (Optional) Create a ./config directory: This is where the logging configuration will live.
2. (Optional) Copy the log_config.json into your ./config directory: This configures your logger
3. In your file you'd like to utilize woodwork in, add from woodwork import __main__ as m
4. See dev-main.py for how to build your logger and configure calling woodwork

Developer Setup

If you are interested in contributing, the following steps are used to activate a developer environment.

1. Install pre-commit if needed via pip install pre-commit
2. Run pre-commit install [--hook-type pre-push] to run linting and formatting before commiting or pushing

Available Arguments

You can pass arguments to woodwork. For more details, see woodwork --help.

Argument	Options	Default	Notes
--mode	run, debug, embed, clear	run	Debug is deprecated, use Run instead. If using a workflow, you must use embed or clear.
--init	none, isolated, all	none
--workflow	none, add, remove, find	none	If specifying a workflow, you must provide a target.
--target	String	""	Defines the target for the workflow. For add workflows, specify the file path to the workflow. For remove workflows, specify the workflow ID. For find workflows, specify the search query.
--version	N/A	N/A	Prints the current version of Woodwork. Note: this will override any other arguments.

When calling your script, you can pass argument to the script as long as they do not conflict with woodworks arguments.

Logging

In log_config.json, you can set the desired logging levels for stdout and the generated log file. Do this by editing the respective level property in the json for the respective handler. Options include the standard logging levels: DEBUG, INFO, WARNING, ERROR, CRITICAL.

You can monitor the log file during execution by opening a terminal and entering tail -f logs/debug_log.log (or your custom log file name if modified in the log_config.json)

Examples

For some examples, consult the examples folder. ENV variables are denotes by a '$', place a .env file in the same directory as the main.ww file and populate it with the necessary variables.

Contributing

To view the contributing guide for woodwork, the CONTRIBUTING.md file in the meta repository contains more information. We would love your help! Additionally, if you prefer working on other projects aligned with language servers or web development, woodwork-language and woodwork-website could be worth taking a look at.

License

woodwork-engine uses a GPL license, which can be read in full here.