Skip to content

Scaffolding a new FSM agent.

It can be difficult to scaffold an initial agent to begin development of the fsm.

The tools within the fsm subcommand are used to scaffold a new agent.

We start with a simple fsm;

cat auto_dev/data/fsm/samples/fsm_specification.yaml

Output: 

alphabet_in:
- DONE
- NO_MAJORITY
- TIMEOUT
default_start_state: FirstRound
final_states:
- ErrorRound
- FinalRound
label: DummyAbciApp
start_states:
- FirstRound
states:
- FirstRound
- FinalRound
- ErrorRound
transition_func:
  (FirstRound, DONE): FinalRound
  (FirstRound, NO_MAJORITY): ErrorRound
  (FirstRound, TIMEOUT): ErrorRound

We can then convert this to a mermaid diagram using the adev tool.

adev fsm from-file auto_dev/data/fsm/samples/fsm_specification.yaml --output mermaid DemoAbciApp

Output: 

graph TD
  FirstRound
  FirstRound
  FinalRound
  ErrorRound
  FirstRound -->|DONE| FinalRound
  FirstRound -->|NO_MAJORITY| ErrorRound
  FirstRound -->|TIMEOUT| ErrorRound
which can be rendered as a mermaid diagram as so;

graph TD
  FirstRound
  FirstRound
  FinalRound
  ErrorRound
  FirstRound -->|DONE| FinalRound
  FirstRound -->|NO_MAJORITY| ErrorRound
  FirstRound -->|TIMEOUT| ErrorRound

We now scaffold the agent.

adev create --no-clean-up -t eightballer/base author/new_agent
We now have a new agent.

cd new_agent

This creates a new agent without any real skills.

Once we have a new agent, we can first scaffold the skill using the CORE autonomy libraries.

This reduces the amount of code we need to write to get a skill up and means that we have no need to write any code to re-implement the wheel.

adev create_from_fsm author/new_agent whale_watcher_fsm.yaml