Project Commands

The project commands cover the local development lifecycle: scaffolding a new project, adding components, running them locally, and pushing to the cloud. These commands require daita-agents to be installed in addition to daita-cli.

#init

Scaffold a new Daita project in the current directory.

bash

daita init [project_name]

If project_name is omitted you are prompted for it interactively.

Options:

Flag	Default	Description
`--type`	`basic`	Project template: `basic`, `analysis`, or `pipeline`
`--force`	—	Overwrite an existing project without confirmation

Example:

bash

daita init my-data-pipeline --type pipeline

This creates:

python

my-data-pipeline/
├── .daita/
├── agents/
│   └── my_agent.py         # starter agent with factory function
├── workflows/
│   └── my_workflow.py      # starter workflow
├── skills/
│   └── example_skill.py    # reusable instructions and tools
├── evals/
│   └── starter-agent.yaml  # starter eval suite config
├── data/
├── tests/
│   └── test_agents.py
├── daita-project.yaml
├── requirements.txt
└── .gitignore

Each generated agent and workflow includes a create_agent() / create_workflow() factory function that daita test and daita push use to load your components. The starter eval config can be run with daita eval locally and is registered for cloud eval runs after deployment.

#create

Add a new agent or workflow to the current project.

bash

daita create agent <name>
daita create workflow <name>

Example:

bash

daita create agent customer_support
daita create workflow ingestion_pipeline

A new Python file is created under agents/ or workflows/ with a boilerplate class and factory function. The component is also registered in daita-project.yaml.

#test

Run agents and workflows locally using your current environment.

bash

daita test [target]

If target is omitted, all agents and workflows in the project are tested.

Options:

Flag	Description
`--data <file>`	JSON or text file to use as input data
`--watch`	Watch for file changes and re-run automatically (requires `watchdog`)

Example:

bash

# Test everything
daita test
 
# Test a specific agent with custom input
daita test my_agent --data sample.json
 
# Watch mode during development
daita test --watch

The CLI dynamically imports each agent or workflow via importlib, calls its factory function, and runs it. Timing and token metrics are printed on success.

#push

Package and deploy the current project to the Daita cloud.

bash

daita push

Options:

Flag	Description
`--force`	Skip the confirmation prompt
`--dry-run`	Show what would be packaged and deployed without actually deploying

Requirements: DAITA_API_KEY must be set.

Example:

bash

daita push --dry-run    # preview the deployment
daita push              # deploy for real

Push creates a zip package from your project, uploads it, and deploys it. The following paths are excluded from the package:

.daita/, .git/, .pytest_cache/, .mypy_cache/
__pycache__/, venv/, env/, .venv/, node_modules/
tests/, data/

A deployment ID is returned on success. Use daita status or daita deployments show <id> to monitor the rollout.