Microsoft Fabric Blog details the latest generally available enhancements to Fabric AI functions, including deeper configurability and new features for developers, data scientists, and analysts.

Microsoft Fabric AI Functions: Enhanced Features Now Generally Available

Fabric AI functions have received significant updates, offering users improved flexibility and power when transforming data with AI. Key highlights from this release include:

New and Enhanced Functions

  • ai.analyze_sentiment() – Detect emotional state, now with configurable labels (‘positive’, ‘negative’, ‘neutral’, ‘mixed’) or custom options.
  • ai.classify() – Categorize data based on user-defined labels.
  • ai.embed() (New!) – Generate vector embeddings from text, enabling semantic comparison, grouping, and search.
  • ai.extract() – Extract specific data types using advanced parameters:
    • label: Custom column names
    • description: Extra context/instructions for extraction
    • max_items: Limit extraction quantity
    • type: Data type (string, number, integer, boolean, object, array)
    • properties: JSON schema elements for complex types
  • ai.fix_grammar() – Automated correction of spelling, grammar, and punctuation.
  • ai.generate_response() – Custom prompt-based generation, now supporting output formatting (text, JSON, pydantic model schema).
  • ai.similarity() – Compare semantic meaning across text values.
  • ai.summarize() – Summarize content, now with instructions for controlled output length.
  • ai.translate() – Translate text to other languages.

Major Enhancements

  • New Optional Parameters for deeper control across functions
  • Support for Advanced Models:
    • Use GPT-5 with configurable reasoning_effort and verbosity
    • Choose from Fabric-supported models, Azure OpenAI resources, or AI Foundry (access models beyond OpenAI, e.g. Claude, LLaMA)
  • Faster, Parallel Execution:
    • Default concurrency raised to 200 for improved processing speed with asynchronous requests

Developer Experience

Example Use Cases

  • Applying sentiment analysis with domain-specific labels
  • Extracting structured information from text with custom JSON schema
  • Using embeddings for intelligent search and grouping
  • Summarizing data and customizing output length
  • Accelerating processing with increased concurrency
  • Expanding AI and ML modeling beyond OpenAI models

Availability

These enhancements enable more powerful, flexible, and scalable AI-driven data science and ML workflows within Microsoft Fabric, leveraging both built-in and bring-your-own-model approaches.

This post appeared first on “Microsoft Fabric Blog”. Read the entire article here