Schema

lit.sdk.data.schema

This module provides classes and functions for managing schemas.

Schema

Represents the configuration for a model and contains information required for compiling the model.

name = name instance-attribute

The name of the schema.

path = Path('/data') / self.team / 'schemas' / f'{self.name}.json' instance-attribute

The path to the JSON representation of the canvas on disk.

team = team instance-attribute

The name of the team.

__init__(team, name)

Initializes a new instance of Schema.

Parameters:

Name	Type	Description	Default
team	str	The team which owns the Schema.	required
name	str	The name of the Schema.	required

build()

Starts a build process in a new screen session.

create(team, name, *, dataset=None, features=None, from_schema=None, paths=None, resolution=None, num_chunks=None, hours_filter=None, split=None, split_date=None, split_percent=None, precache_count=None, seed=None, tags=None, project=None) classmethod

Create a new schema with the given parameters.

This provides a programmatic way to create schemas, either from scratch or by cloning an existing schema and overriding specific parameters.

Parameters:

Name	Type	Description	Default
team	str	The team which will own the schema.	required
name	str	The name for the new schema.	required
dataset	Optional[str]	Dataset name from raw/ folder (e.g., 'aapl'). Required unless using from_schema.	None
features	Optional[list[str]]	List of feature names from features/_index.json. Required unless using from_schema.	None
from_schema	Optional[str]	Clone from an existing schema. Only override the params you specify.	None
paths	Optional[list[str]]	Explicit list of data file paths. If omitted, auto-discovers from dataset folder.	None
resolution	Optional[int]	Data resolution in minutes (default: 1).	None
num_chunks	Optional[int]	Number of chunks for parallel processing (default: 80). Fewer chunks = larger chunks = better parallelism for fast operations.	None
hours_filter	Optional[str]	Filter by trading hours: 'all', 'market', or 'extended' (default: 'all').	None
split	Optional[str]	Split method: 'date' or 'percent' (default: 'date').	None
split_date	Optional[str]	Split date for train/test in ISO format (e.g., '2018-01-01T06:00:00.000Z').	None
split_percent	Optional[int]	Percentage of data for test set (default: 20).	None
precache_count	Optional[int]	Number of pre-shuffled dataset variants to create (default: 1).	None
seed	Optional[int]	Random seed for shuffle reproducibility (default: 0).	None
tags	Optional[list[str]]	List of tags for organization.	None
project	Optional[str]	Project name for grouping related schemas.	None

Returns:

Name	Type	Description
Schema	Schema	The newly created schema object.

Raises:

Type	Description
LitDataError	If schema creation fails.

Examples:

>>> # Clone and modify an existing schema
>>> schema = Schema.create(
...     team='contoso',
...     name='schema_sanity_test',
...     from_schema='schema_hourly_volatility',
...     features=['100_1_minute_bars', '100_1_minute_atr', 'new_label', 'per_minute'],
...     tags=['sanity-test', 'volatility']
... )

>>> # Create from scratch
>>> schema = Schema.create(
...     team='contoso',
...     name='my_new_schema',
...     dataset='aapl',
...     features=['100_1_minute_bars', '100_1_minute_atr', 'atr_up_next_hour', 'per_minute'],
...     split_date='2020-01-01T00:00:00.000Z',
...     num_chunks=40,
...     tags=['experiment', 'v1'],
...     project='volatility'
... )

remove()

Permanently removes a schema.

get_schemas(team)

Get the saved schemas for a given team.

Parameters:

Name	Type	Description	Default
team	str	The name of the team.	required

Returns:

Type	Description
list[Schema]	A list of schema objects.

Examples:

>>> schemas = get_schemas('contoso')
>>> schemas[0].path
/data/contoso/schemas/demo.json