Customize AI functions with pandas

AI functions are designed to work out of the box, with the underlying model and settings configured by default. Users who want more flexible configurations, however, can customize their solutions with a few extra lines of code.

Configurations

By default, AI functions are powered by the built-in AI endpoint in Fabric. The large language model (LLM) settings are globally configured in the aifunc.Conf class. If you work with AI functions in pandas, you can use the aifunc.Conf class to modify some or all of these settings:

The following code sample shows how to override aifunc.Conf settings globally, so that they apply to all AI function calls in a session:

# This code uses AI. Always review output for mistakes.

aifunc.default_conf.temperature = 0.5 # Default: 0.0
aifunc.default_conf.concurrency = 300 # Default: 200

df = pd.DataFrame([
        "Hello! How are you doing today?", 
        "Tell me what you'd like to know, and I'll do my best to help.", 
        "The only thing we have to fear is fear itself.",
    ], columns=["text"])

df["translations"] = df["text"].ai.translate("spanish")
df["sentiment"] = df["text"].ai.analyze_sentiment()
display(df)

You can also customize these settings for each individual function call. Each AI function accepts an optional conf parameter. The following code sample modifies the default aifunc settings for only the ai.translate function call, via a custom temperature value. (The ai.analyze_sentiment call still uses the default values, because no custom values are set).

# This code uses AI. Always review output for mistakes. 

from synapse.ml.aifunc import Conf

df = pd.DataFrame([
        "Hello! How are you doing today?", 
        "Tell me what you'd like to know, and I'll do my best to help.", 
        "The only thing we have to fear is fear itself.",
    ], columns=["text"])

df["translations"] = df["text"].ai.translate("spanish", conf=Conf(temperature=0.5))
df["sentiment"] = df["text"].ai.analyze_sentiment()
display(df)

The following code sample shows how to configure the gpt-5 and other reasoning models for all functions.

aifunc.default_conf.model_deployment_name = "gpt-5"
aifunc.default_conf.temperature = 1  # gpt-5 only accepts default value of temperature
aifunc.default_conf.top_p = 1  # gpt-5 only accepts default value of top_p
aifunc.default_conf.verbosity = "low"
aifunc.default_conf.reasoning_effort = "low"

Custom Models

To use an AI model other than the default, you can choose another model supported by Fabric or configure a custom model endpoint.

Choose another supported large language model

Select one of the models supported by Fabric and configure it using the model_deployment_name parameter. You can do this configuration in one of two ways:

• Globally in the aifunc.Conf class. Example:

  aifunc.default_conf.model_deployment_name = "<model deployment name>"
  

• Individually in each AI function call:

  df["translations"] = df["text"].ai.translate(
      "spanish",
      conf=Conf(model_deployment_name="<model deployment name>"),
  )
  

Choose another supported embedding model

Select one of the models supported by Fabric and configure it using the embedding_deployment_name parameter. You can do this configuration in one of two ways:

• Globally in the aifunc.Conf class. Example:

  aifunc.default_conf.embedding_deployment_name = "<embedding deployment name>"
  

• Individually in each AI function call. Example:

  df["embedding"] = df["text"].ai.embed(
      conf=Conf(embedding_deployment_name="<embbedding deployment name>"),
  )
  

Configure a custom model endpoint

By default, AI functions use the Fabric LLM endpoint API for unified billing and easy setup. You may choose to use your own model endpoint by setting up an Azure OpenAI or OpenAI-compatible client with your endpoint and key. The following example shows how to bring your own Microsoft AI Foundry (formerly Azure OpenAI) resource using aifunc.setup:

from openai import AzureOpenAI

# Example to create client for Microsoft AI Foundry OpenAI models
client = AzureOpenAI(
    azure_endpoint="https://<ai-foundry-resource>.openai.azure.com/",
    api_key="<API_KEY>",
    api_version=aifunc.session.api_version,  # Default "2025-04-01-preview"
    max_retries=aifunc.session.max_retries,  # Default: sys.maxsize ~= 9e18
)
aifunc.setup(client)  # Set the client for all functions.

The following code sample uses placeholder values to show you how to override the built-in Fabric AI endpoint with a custom Microsoft AI Foundry resource to use models beyond OpenAI:

from openai import OpenAI

# Example to create client for Azure AI Foundry models
client = OpenAI(
    base_url="https://<ai-foundry-resource>.services.ai.azure.com/openai/v1/",
    api_key="<API_KEY>",
    max_retries=aifunc.session.max_retries,  # Default: sys.maxsize ~= 9e18
)
aifunc.setup(client)  # Set the client for all functions.

aifunc.default_conf.model_deployment_name = "grok-4-fast-non-reasoning"

Related content

• Customize AI functions configurations with PySpark.

• Detect sentiment with ai.analyze_sentiment.

• Categorize text with ai.classify.

• Extract entities with ai_extract.

• Fix grammar with ai.fix_grammar.

• Answer custom user prompts with ai.generate_response.

• Calculate similarity with ai.similarity.

• Summarize text with ai.summarize.

• Translate text with ai.translate.

• Learn more about the full set of AI functions.

• Did we miss a feature you need? Suggest it on the Fabric Ideas forum.