All Products
Search

This Product

    Realtime Compute for Apache Flink:Model settings

    Document Center

    Realtime Compute for Apache Flink:Model settings

    Last Updated:Mar 26, 2026

    Use data definition language (DDL) statements in Flink SQL to register, view, modify, and delete AI models.

    Usage notes

    • Supported model services: Alibaba Cloud Model Studio, Platform for AI (PAI), and other services with OpenAI-compatible interfaces.

    • PAI model services must be in the same region as your Realtime Compute for Apache Flink service.

    • Requires Ververica Runtime (VVR) 11.1 or later.

    CREATE MODEL

    Register an AI model so Flink SQL jobs can call it.

    Create a script and enter the following statement in the SQL editor:

    CREATE [TEMPORARY] MODEL [catalog_name.][db_name.]model_name
INPUT ( { <physical_column_definition> [, ...n] )
OUTPUT ( { <physical_column_definition> [, ...n] )
WITH (key1=val1, key2=val2, ...)

<physical_column_definition>:
  column_name column_type [COMMENT column_comment]

    Clauses

    Clause Description Schema requirements Example
    INPUT Defines the names, types, and order of input columns. Exactly one STRING column is required. INPUT ( ` input_text STRING COMMENT 'User comment')`
    OUTPUT Defines the names, types, and order of output columns. chat/completions: exactly one STRING column. embeddings: exactly one ARRAY<FLOAT> column. OUTPUT ( ` sentiment_label STRING COMMENT 'Sentiment label')`
    WITH Specifies connection and model parameters. See Parameters. None. WITH ('provider'='bailian', 'endpoint'='${ENDPOINT}', 'model'='qwen-turbo', 'apiKey'='${KEY}')

    Examples

    Register a chat model (Alibaba Cloud Model Studio)

    CREATE MODEL model_bailian
INPUT (`input` STRING)
OUTPUT (`content` STRING)
WITH (
  'provider'='bailian',
  'endpoint'='<Endpoint>',
  'api-key'='<bailian-key>',
  'model'='qwen3-235b-a22b'
);

    The endpoint format is <base-url>/compatible-mode/v1/<task>, for example https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions.

    base-url:

    task: The model task type. Valid values: chat/completions, embeddings.

    Register an embedding model (PAI)

    CREATE MODEL model_pai
INPUT (`input` STRING)
OUTPUT (`embedding` ARRAY<FLOAT>)
WITH (
  'provider'='bailian',
  'endpoint'='<vpc_endpoint>',
  'api-key'='<Token>',
  'model'='qwen3-235b-a22b'
);

    To get the VPC endpoint and API key from the PAI console:

    1. Log on to the PAI console.

    2. In the left navigation pane, choose Model Gallery > Job Management > Deployment Jobs, then click the target task name.

    3. Click View Invocation Information.

    Important

    The VPC invocation address uses HTTP — change it to HTTPS and append /v1/<task> to the URL. For example: https://************.vpc.cn-hangzhou.pai-eas.aliyuncs.com/api/predict/quickstart_deploy_20250722_7b22/v1/chat/completions. The Token value is used as the api-key parameter.

    For details, see One-click deployment of DeepSeek-V3 and DeepSeek-R1 models.

    Parameters

    General

    Parameter Required Description
    provider Yes The model service provider. Always set to bailian.
    endpoint Yes The API endpoint of Alibaba Cloud Model Studio or an OpenAI-compatible service. For Alibaba Cloud Model Studio and PAI endpoints, see Examples. For other services, refer to the service's API documentation.
    api-key Yes The API key for the model service. See Obtain API key. In VVR 11.1, use apiKey.
    max-context-size No The maximum context length per request. When the limit is exceeded, the behavior is controlled by context-overflow-action. Requires VVR 11.2 or later.
    context-overflow-action No How to handle requests that exceed max-context-size. Default: truncated-tail. Requires VVR 11.2 or later. See valid values below.

    Valid values for `context-overflow-action`:

    Value Behavior
    truncated-tail Truncates the context from the end to fit max-context-size, without logging.
    truncated-tail-log Truncates the context from the end to fit max-context-size and logs the truncation.
    truncated-head Truncates the context from the beginning to fit max-context-size, without logging.
    truncated-head-log Truncates the context from the beginning to fit max-context-size and logs the truncation.
    skipped Drops the record without logging.
    skipped-log Drops the record and logs it.

    chat/completions

    Parameter Required Description
    model Yes The text generation model to call. See supported models. Billed based on the model and the number of input and output tokens.
    system-prompt Yes The system prompt sent with each request. Default: You are a helpful assistant. In VVR 11.1, use systemPrompt.
    temperature No Controls output randomness. Valid range: [0, 2); 0 is not recommended. Higher values produce more varied output; lower values make output more deterministic.
    top-p No Controls output diversity by limiting token selection to a cumulative probability threshold. Higher values introduce more randomness. In VVR 11.1, use topP.
    stop No A string that stops generation when it appears in the output.
    max-tokens No The maximum number of tokens in the generated output. In VVR 11.1, use maxTokens.

    embeddings

    Parameter Required Description
    model Yes The text embedding model to call. See supported models. Billed based on the model and the number of input and output tokens.
    dimension No The output vector dimension. Default: 1024. Valid values: 1024, 768, 512. Other values cause an error.

    View models

    Create a script and run one of the following statements in the SQL editor.

    List registered models:

    SHOW MODELS [ ( FROM | IN ) [catalog_name.]database_name ];

    Show the DDL statement used to register a model:

    SHOW CREATE MODEL [catalog_name.][db_name.]model_name;

    Show the input and output schema of a model:

    DESCRIBE MODEL [catalog_name.][db_name.]model_name;

    Example

    SHOW MODELS;

-- Result
--+------------+
--| model name |
--+------------+
--|          m |
--+------------+

DESCRIBE MODEL m;

-- Result
-- +---------+--------+------+----------+
-- |    name |   type | null | is input |
-- +---------+--------+------+----------+
-- | content | STRING | TRUE |     TRUE |
-- |   label | BIGINT | TRUE |    FALSE |
-- +---------+--------+------+----------+

    Modify models

    Create a script and enter one of the following statements in the SQL editor.

    ALTER MODEL [IF EXISTS] [catalog_name.][db_name.]model_name {
  RENAME TO new_table_name
  SET (key1=val1, ...)
  RESET (key1, ...)
}

    Examples

    Rename a model:

    ALTER MODEL m RENAME TO m1;

    Update a parameter:

    ALTER MODEL m SET ('endpoint' = '<Your_Endpoint>');

    Reset a parameter to its default value:

    ALTER MODEL m RESET ('endpoint');

    Delete models

    Create a script and enter the following statement in the SQL editor.

    DROP [TEMPORARY] MODEL [IF EXISTS] [catalog_name.][db_name.]model_name

    Example

    DROP MODEL m;