OPTIONAL. Define custom structured output using Gemini's JSON mode. NOT REQUIRED - by default, descriptions are stored as plain text. When provided, Gemini will extract structured data matching this schema. Two modes supported: 1. Natural language prompt (string): Describe desired output in plain English - Gemini automatically infers JSON schema from your description - Example: 'Extract product names, colors, and aesthetic labels' 2. Explicit JSON schema (dict): Provide complete JSON schema for output structure - Full control over output structure, types, and constraints - Use response_mime_type='application/json' in generation_config - Example: {'type': 'object', 'properties': {'products': {'type': 'array', ...}}} Use when: - Need to extract structured product/entity information from videos - Want consistent, parseable output format (not free-form text) - Require specific fields like visibility_percentage, product categories, etc. - Building e-commerce, fashion, or product discovery applications Output fields are automatically added to collection schema and stored in document metadata. Note: When using response_shape, set description_prompt to describe the extraction task.
| Name | Type | Description | Notes |
|---|
from mixpeek.models.response_shape1 import ResponseShape1
# TODO update the JSON string below
json = "{}"
# create an instance of ResponseShape1 from a JSON string
response_shape1_instance = ResponseShape1.from_json(json)
# print the JSON string representation of the object
print(ResponseShape1.to_json())
# convert the object into a dict
response_shape1_dict = response_shape1_instance.to_dict()
# create an instance of ResponseShape1 from a dict
response_shape1_from_dict = ResponseShape1.from_dict(response_shape1_dict)