Custom AI fields

Declare a field whose value the AI suggests while the user types.

Two serializer fields do this: PredictiveCharField for a single line and PredictiveTextField for longer text. Both emit an ai format, which the frontend renders as SchemaPredictiveTextInput.

Declaration

from habarico_enterprise.ai.fields import PredictiveCharField, PredictiveTextField
from habarico.serializers import AdminModelSerializer

from .models import Product


class ProductSerializer(AdminModelSerializer):
    name = PredictiveCharField()
    short_description = PredictiveTextField(source_fields=["name", "category"])

    class Meta:
        model = Product
        fields = ["id", "name", "sku", "description", "short_description", "status", "category"]

Options

Option	Type	Default	Description
source_fields	list[str]	every other filled field	Which form values are sent as context for the suggestion

Everything else is inherited from DRF's CharField — required, allow_blank, max_length, …

Generated schema

Field class	type	format	Options
PredictiveCharField	string	ai	ai.source_fields
PredictiveTextField	string	ai-text	ai.source_fields

The frontend renders both with SchemaPredictiveTextInput; ai-text gets the multiline variant and asks for a longer completion.

caution

An empty source_fields means all other non-empty form values travel to the AI provider. On a resource holding personal data, list the fields explicitly.

See also

• Usage — enabling the prediction endpoint on the admin
• SchemaPredictiveTextInput — the input this produces
• FieldSchema — where options.ai sits in the schema