This documentation is for an unreleased version of Apache Flink. We recommend you use the latest stable version.

OpenAI #

The OpenAI Model Function allows Flink SQL to call OpenAI API for inference tasks.

Overview #

The function supports calling remote OpenAI model services via Flink SQL for prediction/inference tasks. Currently, the following tasks are supported:

• Chat Completions: generate a model response from a list of messages comprising a conversation.
• Embeddings: get a vector representation of a given input that can be easily consumed by machine learning models and algorithms.

Usage examples #

The following example creates a chat completions model and uses it to predict sentiment labels for movie reviews.

First, create the chat completions model with the following SQL statement:

CREATE MODEL ai_analyze_sentiment
INPUT (`input` STRING)
OUTPUT (`content` STRING)
WITH (
    'provider'='openai',
    'endpoint'='https://api.openai.com/v1/chat/completions',
    'api-key' = '<YOUR KEY>',
    'model'='gpt-3.5-turbo',
    'system-prompt' = 'Classify the text below into one of the following labels: [positive, negative, neutral, mixed]. Output only the label.'
);

Suppose the following data is stored in a table named movie_comment, and the prediction result is to be stored in a table named print_sink:

CREATE TEMPORARY VIEW movie_comment(id, movie_name,  user_comment, actual_label)
AS VALUES
  (1, 'Good Stuff', 'The part where children guess the sounds is my favorite. It's a very romantic narrative compared to other movies I've seen. Very gentle and full of love.', 'positive');

CREATE TEMPORARY TABLE print_sink(
  id BIGINT,
  movie_name VARCHAR,
  predicit_label VARCHAR,
  actual_label VARCHAR
) WITH (
  'connector' = 'print'
);

Then the following SQL statement can be used to predict sentiment labels for movie reviews:

INSERT INTO print_sink
SELECT id, movie_name, content as predicit_label, actual_label
FROM ML_PREDICT(
  TABLE movie_comment,
  MODEL ai_analyze_sentiment,
  DESCRIPTOR(user_comment));

Model Options #

Common #

Chat Completions #

Embeddings #

Schema Requirement #