akili init

Initialize a new data product in the current directory by creating an .akili/ directory with 6 scaffold YAML manifest files. This is the fastest way to start building a data product.

Synopsis

akili init [--name <product-name>]

Description

akili init creates the standard Akili data product directory structure in the current working directory. It generates 6 manifest templates that define every aspect of a data product:

File	Purpose
.akili/product.yaml	Product identity, metadata, classification, domain
.akili/inputs.yaml	Input declarations (upstream products or connector resources)
.akili/output.yaml	Output schema contract (columns, types, roles)
.akili/quality.yaml	Quality checks and SLA definitions
.akili/serving.yaml	Serving intent and materialization config
.akili/compute.yaml	Runtime, engine, schedule, resources

If --name is not provided, the product name defaults to the current directory name.

Note

akili init is a local-only command. It does not contact the API and does not require authentication. The generated manifests are templates — you must fill in the details before deploying.

Flags

Flag	Short	Default	Description
--name		current directory name	Product name to use in the manifest

Examples

# Initialize in the current directory (uses directory name)
mkdir daily-orders && cd daily-orders
akili init
# Initialized data product 'daily-orders' with 6 manifest files
#   .akili/product.yaml
#   .akili/inputs.yaml
#   .akili/output.yaml
#   .akili/quality.yaml
#   .akili/serving.yaml
#   .akili/compute.yaml

# Initialize with a custom name
akili init --name customer-segmentation

Generated Manifests

product.yaml

apiVersion: akili/v1
kind: DataProduct
metadata:
  name: daily-orders
  domain: sales
  version: 1.0.0
  owner: data-engineering
  description: ""
  classification: internal
  #  Options: public, internal, confidential, restricted
  tags: []

inputs.yaml

apiVersion: akili/v1
kind: Inputs
inputs: []
#  - name: example_source
#    type: connector
#    connection: my_connection
#    ingestion:
#      strategy: incremental
#      cursor_field: updated_at

output.yaml

apiVersion: akili/v1
kind: Output
columns: []
#  - name: id
#    type: STRING
#    role: identity
#  - name: amount
#    type: DECIMAL
#    role: metric

quality.yaml

apiVersion: akili/v1
kind: Quality
checks: []
#  - name: not_null_check
#    type: not_null
#    column: id
sla:
  freshness: 24h
  completeness: 0.99

serving.yaml

apiVersion: akili/v1
kind: Serving
intent: analytical
#  Options: analytical, operational, streaming
materialization:
  enabled: false

compute.yaml

apiVersion: akili/v1
kind: Compute
engine: sql
schedule: "@daily"
resources:
  cpu: "1"
  memory: 2Gi
#  entrypoint: logic/transform.sql

Classification and access control are declared in product.yaml under metadata.classification. There is no separate governance file.

Behavior

• If .akili/ already exists, the command prints a warning and exits without modifying any files. This prevents accidental overwrites of existing manifest work.
• The command is idempotent in the sense that running it twice has no effect on the second run.
• No network connection is required.

Workflow

After initializing, the typical workflow is:

1. Edit the manifest files in .akili/ to define your data product
2. Validate the manifests: akili validate .akili/
3. Create the product on the platform: akili product create --name daily-orders --namespace sales
4. Deploy: akili product deploy daily-orders

Related

• akili scaffold — creates a new directory with manifests (vs. init which works in the current directory)
• akili validate — validates manifest files
• akili product create — registers a product on the platform