MaaS_Ge

Model Description

The MaaS_Ge_omni_flash_preview_20260630 model currently only supports InteractionAPI

Practice Methods

This interface supports both synchronous task mode and asynchronous task mode. When using background= true in the create interface, it corresponds to an asynchronous task. According to current actual tests, the ID of the asynchronous task can be queried via the Get interface to obtain the task result.

However, for the MaaS_Ge_omni_flash_preview_20260630 model, using the asynchronous task ID as the previous_interaction_id in the create interface is temporarily not effective (the supplier side is confirming the issue).

Parameter Support

For the MaaS_Ge_omni_flash_preview_20260630 model

response_modalities only supports textand video
It is recommended to use a step-form input to transmit video content
generation_config. max_output_tokensIt is recommended to set a relatively large value for this parameter; our tests show that if it is set too small, it may result in the following error being returned: Request contains an invalid argument.
Tool calling is not supported for the time being
Streaming calls are not supported for the time being

Functional Interface Details

Create Interaction (Generate)

Request URL

POST https://genaiapi-m2.cloudsway.net/v1/ai/{YOUR_ENDPOINT}/omni/interactions

Request Header

Header	Required or not	Description
`Authorization`	Yes	`Bearer {YOUR_AK}`
`Content-Type`	Yes	`application/json`

Request Body Parameters

Parameter Name	Field Type	Required or not	default value	Description
`input`	string or Content array or Step array	Yes	-	Interactive input (see the description of the input field).
`system_instruction`	string	No	-	Interactive system instructions.
`tools`	array(Tool)	No	-	The list of tool declarations that the model can invoke during interaction (see the description of the `tools` field).
`response_format`	ResponseFormat or ResponseFormatList	No	-	The forcibly generated response is a JSON object that conforms to the JSON Schema specified by this field.
`response_mime_type`	string	No	-	The MIME type of the response. If `response_format`is set, this field is required.
`stream`	boolean	No	`false`	Input only. Specifies whether to return interactions in streaming mode. The current model does not support streaming calls for the time being.
`store`	boolean	No	-	Input only. Determines whether to store responses and requests for subsequent retrieval.
`background`	boolean	No	-	Input only. Determines whether model interactions run in the background.
`generation_config`	GenerationConfig	No	-	Model configuration parameters (see the description of the generation_config field).
`previous_interaction_id`	string	No	-	The ID of the previous interaction, if any.
`response_modalities`	array(string)	No	-	The response modality of the request. Optional values: `text`, `image`, `audio`, `video`, `document`.

input field description

inputis a required field and supports one of the following three formats (consistent with the Google Interactions API):

1. String

Directly pass in the user text, for example:

"input": "Introduce artificial intelligence in one sentence."

2. Content Array (array of Content)

An array of MultiModal Machine Learning content blocks. Each Content is categorized by its type; see the Content type for details.

! Note: When passing base64 data via data, do not include the base64 header

Example:

"input": [
  { "type": "text", "text": "Please describe this image." },
  { "type": "image", "data": "BASE64_ENCODED_IMAGE", "mime_type": "image/png" }
]

3. Step Array (array of Step)

A structured step array used to explicitly distinguish steps such as user input. For details, see the Step type (input).

Example:

"input": [
  {
    "type": "user_input",
    "content": [
      { "type": "text", "text": "Hello, please introduce yourself." }
    ]
  }
]

Content Type

Content is a polymorphic object, and its specific type is distinguished via the typefield.

TextContent (Text Content Block)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"text"`.
`text`	string	Yes	Text content.
`annotations`	array(Annotation)	No	Citation information for model-generated content (see Annotation type).

Annotation Type

annotationsis a polymorphic array, and the specific reference type is distinguished via type.

UrlCitation (URL Citation)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"url_citation"`.
`url`	string	No	URL address.
`title`	string	No	URL Title.
`start_index`	integer	No	The start position (in ByteDance) of the response fragment attributed to this source.
`end_index`	integer	No	The end position of the attributed segment (exclusive).

FileCitation (File Reference)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"file_citation"`.
`document_uri`	string	No	The URI of the file.
`file_name`	string	No	File name.
`source`	string	No	The source to which the text snippet is attributed.
`custom_metadata`	object	No	Retrieval context metadata provided by the user.
`page_number`	integer	No	Page number of the cited document (if applicable).
`media_id`	string	No	The media ID for image references (if applicable).
`start_index`	integer	No	The start position of the response fragment attributed to this source, measured in ByteDance.
`end_index`	integer	No	The end position of the attributed segment (exclusive).

PlaceCitation (Location Citation)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"place_citation"`.
`place_id`	string	No	Location ID, in the format of `places/{place_id}`.
`name`	string	No	Location Title.
`url`	string	No	URI reference of the location.
`review_snippets`	array(ReviewSnippet)	No	User review snippets used to generate location-related responses.

ReviewSnippet (Review Snippet)

Parameter Name	Field Type	Required or not	Description
`title`	string	No	Review Title.
`url`	string	No	This is the link corresponding to the reviews on Google Maps.
`review_id`	string	No	Evaluation snippet ID.
`start_index`	integer	No	The start position of the response fragment attributed to this source, measured in ByteDance.
`end_index`	integer	No	The end position of the attributed segment (exclusive).

ImageContent (Image Content Block)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"image"`.
`data`	string	No	Image content (Base64 encoded).
`uri`	string	No	Image URI.
`mime_type`	string	No	Image MIME type. Supported values: `image/png`, `image/jpeg`, `image/webp`, `image/heic`, `image/heif`, `image/gif`, `image/bmp`, `image/tiff`.
`resolution`	string	No	Media resolution. Optional values: `low`, `medium`, `high`, `ultra_high`.

AudioContent (Audio Content Block)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"audio"`.
`data`	string	No	Audio content (Base64 encoded).
`uri`	string	No	Audio URI.
`mime_type`	string	No	Audio MIME type. Supported values: `audio/wav`, `audio/mp3`, `audio/aiff`, `audio/aac`, `audio/ogg`, `audio/flac`, `audio/mpeg`, `audio/m4a`, `audio/l16`, `audio/opus`, `audio/alaw`, `audio/mulaw`.
`channels`	integer	No	Number of audio channels.
`sample_rate`	integer	No	Audio sampling rate.

VideoContent (Video Content Block)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"video"`.
`data`	string	No	Video content (Base64 encoded).
`uri`	string	No	Video URI.
`mime_type`	string	No	Video MIME type. Supported values: `video/mp4`, `video/mpeg`, `video/mpg`, `video/mov`, `video/avi`, `video/x-flv`, `video/webm`, `video/wmv`, `video/3gpp`.
`resolution`	string	No	Media resolution. Supported values: `low`, `medium`, `high`, `ultra_high`.

DocumentContent (Document Content Block)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed to `"document"`.
`data`	string	No	Document content (Base64 encoded).
`uri`	string	No	Document URI.
`mime_type`	string	No	Document MIME type. Supported values: `application/pdf`.

Step Type (input)

Step is a polymorphic object, and its specific type is distinguished via the typefield. The following are common input step types:

UserInputStep (User Input Step)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"user_input"`.
`content`	array(Content)	No	The input content provided by the user, whose elements are of the Content type mentioned above.

ModelOutputStep (Model Output Step)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"model_output"`.
`content`	array(Content)	No	The output content generated by the model consists of elements of the Content type mentioned above.

ThoughtStep (Thinking Step)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"thought"`.
`signature`	string	No	A signature hash used for backend verification.
`summary`	ThoughtSummaryContent	No	Thought summary, with the same structure as TextContent (`type`is `"text"`, containing the `text`field).

tools Field Description

toolsis an array for tool declarations, and each Tool is distinguished by typeto specify its concrete type.

Function (function tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"function"`.
`name`	string	No	Function name.
`description`	string	No	Function description.
`parameters`	object	No	The JSON Schema for function parameters.

CodeExecution (Code Execution Tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"code_execution"`.

Google Search (Google Search Tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"google_search"`.
`search_types`	array(string)	No	Enabled search enhancement types. Optional values: `web_search`, `image_search`, `enterprise_web_search`.

UrlContext (URL Context Tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"url_context"`.

Google Maps (Google map tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed to `"google_maps"`.
`enable_widget`	boolean	No	Whether to return the widget context token in the response of the tool call result.
`latitude`	number	No	The latitude of the user's location.
`longitude`	number	No	The longitude of the user's location.

Retrieval (File Retrieval Tool)

Parameter Name	Field Type	Required or not	Description
`type`	string	Yes	Fixed as `"retrieval"`.
`retrieval_types`	array(string)	No	Enabled file search types. Optional values: `vertex_ai_search`.
`vertex_ai_search_config`	VertexAISearchConfig	No	Vertex AI Search configuration (refer to the nested field vertex_ai_search_config).

vertex_ai_search_config nested field

Parameter Name	Field Type	Required or not	Description
`engine`	string	No	Agent Platform Search Engine.
`datastores`	array(string)	No	Data storage for Agent Platform Search.

Description of the generation_config Field

Parameter Name	Field Type	Required or not	Description
`temperature`	number	No	Control the randomness of the output.
`top_p`	number	No	The maximum cumulative token probability considered during sampling.
`seed`	integer	No	A reproducibility seed for decoding.
`stop_sequences`	array(string)	No	Stop outputting interactions when these character sequences are encountered.
`thinking_level`	string	No	The level of reasoning tokens that the model should generate. Available values: `minimal`, `low`, `medium`, `high`.
`thinking_summaries`	string	No	Whether to include a reasoning summary in the response. Optional values: `auto`, `none`.
`max_output_tokens`	integer	No	The maximum number of tokens contained in the response.
`speech_config`	SpeechConfig	No	Voice interaction configuration (refer to the nested field of speech_config).
`image_config`	ImageConfig	No	Image interaction configuration (refer to the nested field of image_config).
`tool_choice`	ToolChoiceConfig or ToolChoiceType	No	Tool selection and configuration.

Nested fields of speech_config

Parameter Name	Field Type	Required or not	Description
`voice`	string	No	The speaker's voice.
`language`	string	No	the language of speech.
`speaker`	string	No	Speaker name, which shall be consistent with the speaker name provided in the prompt.

image_config nested field

Parameter Name	Field Type	Required or not	Description
`aspect_ratio`	string	No	Aspect ratio. Optional values: `1:1`, `2:3`, `3:2`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`, `1:8`, `8:1`, `1:4`, `4:1`.
`image_size`	string	No	Image size. Available values: `512`, `1K`, `2K`, `4K`.

Response Parameters (synchronous, `stream= false`)

Returns Interaction resource.

Parameter Name	Field Type	Description
`id`	string	Output field. Unique Device Identifier for the interaction.
`status`	string	Output field. Interaction status. Valid values: `in_progress`, `requires_action`, `completed`, `failed`, `cancelled`, `incomplete`.
`created`	string	Output field. Response creation time, in ISO 8601 format (`YYYY-MM-DDThh: mm: ssZ`).
`updated`	string	Output field. Last update time of the response, in ISO 8601 format.
`role`	string	Output fields. Interactive roles.
`model`	string	The name of the model used to generate interactions.
`system_instruction`	string	Interactive system instructions.
`tools`	array(Tool)	A list of tool declarations that the model can invoke during the interaction process.
`usage`	Usage	Output field. Token usage statistics of the interaction request (see the description of the `usage` field), which is used for billing.
`response_modalities`	array(string)	The response modality of the request (TEXT, IMAGE, AUDIO, etc.).
`response_mime_type`	string	The MIME type of the response.
`previous_interaction_id`	string	The ID of the previous interaction, if any.
`steps`	array(Step)	Output field. Constitutes the step list of the interaction (see the description of the steps field).
`object`	string	Fixed as `"interaction"`.

Description of the usage Field

Parameter Name	Field Type	Description
`total_input_tokens`	integer	The number of tokens in the prompt (context).
`input_tokens_by_modality`	array(ModalityTokens)	Input token usage split by modality.
`total_cached_tokens`	integer	The number of tokens in the cached portion of the prompt.
`cached_tokens_by_modality`	array(ModalityTokens)	Cache token usage split by modality.
`total_output_tokens`	integer	The total number of tokens in all generated responses.
`output_tokens_by_modality`	array(ModalityTokens)	Output token usage split by modality.
`total_tool_use_tokens`	integer	The number of tokens in the tool call prompt.
`tool_use_tokens_by_modality`	array(ModalityTokens)	Tool call token usage broken down by modality.
`total_thought_tokens`	integer	The number of thinking tokens for thinking-type models.
`total_tokens`	integer	Total number of tokens for the interaction request (prompt + response + other internal tokens).

ModalityTokens Object

Parameter Name	Field Type	Description
`modality`	string	Modalities associated with token counting. Valid values: `text`, `image`, `audio`, `video`, `document`.
`tokens`	integer	The number of tokens for this modality.

Description of the steps field

stepsis an output field that represents each step in the interaction. The step type is consistent with the Step type (input) in the request input, and the common types are as follows:

UserInputStep (User Input Step)

Parameter Name	Field Type	Description
`type`	string	Fixed as `"user_input"`.
`content`	array(Content)	The input content provided by the user.

ModelOutputStep (Model Output Step)

Parameter Name	Field Type	Description
`type`	string	Fixed as `"model_output"`.
`content`	array(Content)	The output content generated by the model.

ThoughtStep (Thinking Step)

Parameter Name	Field Type	Description
`type`	string	Fixed as `"thought"`.
`signature`	string	A signature hash used for backend verification.
`summary`	ThoughtSummaryContent	Thinking Summary.

contenthas the same Content block type as the Content type mentioned above (text, image, audio, video, document, etc.).

Streaming Response (`stream= true`)

When stream= true, the response is an SSE (Content-Type: text/event-stream) stream, and each event consists of an eventline and a dataline. datais a JSON object, in which event_typeidentifies the event type.

The main SSE events involved in this interface are as follows:

Event Name	event_type	Description
`interaction.created`	`interaction.created`	The interaction has been created. `interaction`contains `id`, `status`and other initial information.
`step.delta`	`step.delta`	Incremental content. `index`is the step index, `delta`refers to the incremental data (e. g., the `type: "text"`scenario's `text`field).
`interaction.completed`	`interaction.completed`	The interaction is complete. `interaction`which includes the final status and usage statistics.
`done`	-	Stream end marker, `data`is`[DONE]`.

Request Example (Synchronous)

curl --location 'https://genaiapi-m2.cloudsway.net/v1/ai/{YOUR_ENDPOINT}/omni/interactions' \
--header 'Authorization: Bearer {YOUR_AK}' \
--header 'Content-Type: application/json' \
--data '{
    "input": [
        {
            "type": "user_input",
            "content": [
                {
                    "type": "text",
                    "text": "Generate a cute cat, playing with kids."
                }
            ]
        }
    ],
    "response_modalities": ["text", "video"]
}'

Response Example (Synchronous)

{
    "id": "rnVLaoO1KuKF694Pr4GXgQI",
    "status": "completed",
    "role": "model",
    "created": "2026-07-06T09:30:56Z",
    "updated": "2026-07-06T09:30:56Z",
    "object": "interaction",
    "model": "gemini-omni-flash-preview",
    "usage": {
        "total_input_tokens": 9,
        "input_tokens_by_modality": [
            {
                "modality": "text",
                "tokens": 9
            }
        ],
        "total_output_tokens": 57920,
        "output_tokens_by_modality": [
            {
                "modality": "video",
                "tokens": 57920
            }
        ],
        "total_tool_use_tokens": 0,
        "total_thought_tokens": 331,
        "total_tokens": 58260
    },
    "steps": [
        {
            "type": "thought",
            "signature": "signature",
            "summary": [
                {
                    "text": "**Creating a Charming Scene**\n\nI'm developing a prompt for a video model, envisioning a playful scene. It starts in a sunny living room with a boy and girl interacting with a cat. Close-ups will highlight the cat's joy and focus, as well as the children's reactions. The video will culminate with the cat resting on the boy's lap. The entire scene will have cheerful background music and sound effects.\n\n\n",
                    "type": "text"
                }
            ]
        },
        {
            "type": "model_output",
            "content": [
                {
                    "type": "video",
                    "data": "video data encoded by base64"
                    "mime_type": "video/mp4"
                }
            ]
        }
    ]
}

Query GET Interface

Request URL

GET https://genaiapi-m2.cloudsway.net/v1/ai/{YOUR_ENDPOINT}/omni/interactions/{interactionId}

Request Header

Header	Required or not	Description
`Authorization`	Yes	`Bearer {YOUR_AK}`
`Content-Type`	Yes	`application/json`

Request Query Parameters

Parameter Name	Field Type	Required or not	default value	Description
`stream`	boolean	No	-	Whether to use SSE streaming for pulling; defaults to `false`
`last_event_id`	string	No	-	Previous event ID for streaming resumption (only applicable when `stream= true`)

Response Parameters

The response body is the same as that of a synchronously created response.

Request Example

curl --location 'https://genaiapi-m2.cloudsway.net/v1/ai/{YOUR_ENDPOINT}/omni/interactions/{interactionId}' \
--header 'Authorization: Bearer {YOUR_AK}'

MaaS_Ge

Model Description

Practice Methods

Parameter Support

Functional Interface Details

Create Interaction (Generate)

Request URL

Request Header

Request Body Parameters

input field description

Content Type

Step Type (input)

tools Field Description

Description of the generation_config Field

Response Parameters (synchronous, stream= false)

Description of the usage Field

Description of the steps field

Streaming Response (stream= true)

Request Example (Synchronous)

Response Example (Synchronous)

Query GET Interface

Request URL

Request Header

Request Query Parameters

Response Parameters

Request Example

Response Parameters (synchronous, `stream= false`)

Streaming Response (`stream= true`)