Quickstart

Requirements

• A running NATS.io cluster
• A running Postgres DB (See included docker-compose.yml file)
• Poetry
• (Optional for now) A Weaviate instance (See included docker-compose.yml file)

Quickstart

1. Clone the repo, and install dependencies with poetry.

   poetry install
   

2. Run the NATS Jetstream cluster.

   You can simply use the NATS CLI to run a test cluster:

   nats server --jetstream
   

3. Run Postgres.

   You can use the included docker-compose.yml file:

   docker compose --profile postgres up -d
   

4. Create a .env file in ./ based on the included .env.example file.

5. Run the coordinator process.

   poetry run germinate coordinator
   

   The coordinator updates state, and triggers state transitions/enqueues tasks as appropriate based on your Workflow state machine + tasks DAGs definitions.

6. Run one or more workers.

   poetry run germinate worker
   

   Workers execute tasks that are assigned to them.

7. Run a workflow.

   poetry run germinate workflow germinate_ai.workflows.metagpt.main '{"product_requirements": "A simple tictactoe game using Pygame"}' 
   

TEMP: Ideally, you would be able to create workflows in your own project and use Germinate as a framework. But, until we implement pickling and sending task executors over the wire to remote workers, its easiest to write workflows in ./germinate_ai/workflows and execute them as shown above.

The last argument is the input to the first set of tasks in the workflows initial state's tasks DAG.

Refer to ./germinate_ai/workflows/metagpt to see our (WIP) first example workflow.

Commands

Temporary limited CLI:

• germinate coordinator - Start the Coordinator process.
• germinate worker - Start the Worker process.
• germinate workflow germinate_ai.workflows.metagpt.main - Enqueue the simple MetaGPT-(ish) workflow.
• germinate db - Create all Postgres tables.
• germinate --help - Print help message and exit.