Quickstart: diagridpy CLI

Initialize and deploy a durable agent to Catalyst in minutes using the diagridpy CLI. Choose your preferred agent framework — the CLI handles project scaffolding, Kubernetes provisioning, and deployment.

---

Prerequisites

Before you begin, ensure you have:

• Python 3.11+
• Docker
• kind (for local Kubernetes)
• An LLM API key (OpenAI or Google depending on framework)

---

Steps

1

Install the CLI

Install the Diagrid Python CLI:

pip install diagrid

Verify the installation:

diagridpy --help

2

Initialize your project

Run diagridpy init with your preferred framework:

Dapr Agents

diagridpy init my-project --framework dapr-agents

LangGraph

diagridpy init my-project --framework langgraph

Strands

diagridpy init my-project --framework strands

OpenAI Agents

diagridpy init my-project --framework openai-agents

CrewAI

diagridpy init my-project --framework crewai

Google ADK

diagridpy init my-project --framework adk

note

ADK uses GOOGLE_API_KEY instead of OPENAI_API_KEY. Pass it with --google-api-key or set the environment variable.

Pydantic AI

diagridpy init my-project --framework pydantic-ai

The init command performs 7 steps automatically:

1. Authenticate with Diagrid Catalyst
2. Validate LLM API key (OpenAI or Google)
3. Create a Catalyst project with agent infrastructure
4. Clone the quickstart template for your chosen framework
5. Provision a kind cluster (or use existing kubectl context)
6. Install the dapr-agents Helm chart with your LLM key
7. Create an App ID for your agent

3

Deploy and trigger

Navigate to your project and deploy:

cd my-project
diagridpy deploy --trigger "Plan a trip to Paris"

The deploy command:

1. Builds a Docker image from your project
2. Pushes it to the local kind registry
3. Deploys the agent as a Kubernetes pod
4. Connects it to Catalyst via the Dapr sidecar
5. Triggers the agent with your prompt via POST /agent/run

tip

You can use the --port option to specify a custom port for your agent container (default: 5001).

4

View results

After triggering, the CLI displays the workflow result. You can also check workflow status:

# Check the agent logs
kubectl logs -l app=my-project-agent -f

# Trigger the agent again via curl
curl -X POST http://localhost:5001/agent/run \
  -H "Content-Type: application/json" \
  -d '{"task": "Find flights to Tokyo"}'

---

What's in the quickstart template?

Each framework template includes:

File	Description
main.py	Agent definition with a sample tool and runner.serve()
requirements.txt	Python dependencies for the framework
Dockerfile	Container image build configuration

The main.py creates a minimal agent with a sample get_weather tool, wraps it in a DaprWorkflowAgentRunner, and starts an HTTP server with runner.serve() that exposes:

• POST /agent/run — trigger the agent with a prompt
• GET /agent/run/{workflow_id} — check workflow status

---

Next steps

• Durable Agent Quickstart — Build a full travel assistant with conversation memory
• CLI Reference: init — All diagridpy init options
• CLI Reference: deploy — All diagridpy deploy options