guides-valory-xyz-open-autonomy

Recall that the {{fsm_app}} skill is the core part of the agent blueprint that encodes the business logic of the AI agent. Developing the {{fsm_app}} is possibly most demanding step in the development process.

In order to simplify and speed up the development, the {{open_autonomy}} framework provides an {{fsm_app}} scaffold tool that auto-generates functional, template classes with all the boilerplate code in place. Therefore, you can focus on completing the actual operations executed at each state of the AI agent.

What you will learn

This guide covers step 3 of the development process. You will learn how to use the {{fsm_app}} scaffold tool to transform an FSM specification into template classes for the {{fsm_app}} skill, and identify what methods you need to implement to have a functional skill.

You must ensure that your machine satisfies the framework requirements, you have set up the framework, and you have a local registry populated with some default components. As a result you should have a Pipenv workspace folder with an initialized local registry (./packages) in it.

Step-by-step instructions

1. Generate the {{fsm_app}} template classes. Place the FSM specification fsm_specification.yaml that you have created in the previous step in the workspace folder. Use the scaffold tool to generate an {{fsm_app}} with template classes:

   bashCopy

   autonomy scaffold -tlr fsm your_fsm_app_abci --spec fsm_specification.yaml
   

   Make sure to end the name with _abci. This command will download the required packages to the local registry, and generate the {{fsm_app}} skill with a template for the required classes. The -tlr flag indicates that the {{fsm_app}} skill will be generated in the local registry (./packages). The actual path will be

   Copy

   ./packages/your_name/skills/your_fsm_app_abci/
   

   You can explore this path and take a look at the generated classes.

2. Fill in the business logic code of your {{fsm_app}}. By default, the generated classes don't execute any operation. It is your turn to define what actions are occurring at each state of the AI agent, by filling up the code of the template classes generated above. You can identify the main places where you should populate code by browsing for # TODO comments. These are the main files that you should edit:

   • behaviours.py,
   • rounds.py,
   • payloads.py.

   You should also define a number of test classes. You can review how the demo AI agents are implemented, or read about the internals of {{fsm_app}}s to learn how to populate the template classes.

   !!! tip

   Copy

    If you are a new user and want to skip coding the {{fsm_app}}, in the next guides we show you how you can use an existing {{fsm_app}} or an existing agent blueprint from the default [packages shipped with the framework](../package_list.md).