Published on

Microsoft announced that it will deeply integrate Azure AI with GitHub to greatly simplify the AI ​​application development process

Microsoft announced a partnership with GitHub to seamlessly integrate Codespaces and Visual Studio Code through GitHub.com, enabling more than 100 million developers to build AI applications directly from GitHub. This collaboration makes AI development more convenient and scalable, providing developers with the opportunity to access Azure AI's leading models through GitHub Models and simple APIs.

  • The new feature allows developers to access and use Azure AI services directly in GitHub.
  • The integration includes features such as code generation, AI-assisted programming, and automated testing.
  • Developers can use Azure AI's latest AI model library directly in GitHub Models
  • Developers can use the Azure AI Inference API to easily experiment and compare different AI models without changing the underlying code.

Main content:

  1. Seamless integration and developer empowerment:
    • Developers can use GitHub Codespaces for rapid prototyping and Microsoft Visual Studio Code for code generation and debugging.
    • GitHub Models provides developers with access to leading Azure AI models, simplifying the process of developing AI applications from experimentation to production-ready models.
  2. Model Diversity and Choice:
    • Azure AI provides the most complete model library on the market, including the latest models from OpenAI, Meta, Mistral, etc., as well as Microsoft's own Phi-3 series of small language models.
    • Developers can explore and leverage the latest AI models in GitHub Models, choosing the unique combination of capabilities, performance metrics, and cost-effectiveness that best suits their application needs.
  3. Safety and Security:
    • GitHub Models comes with Azure AI content security built in, providing real-time protection against the generation of harmful content, copyrighted material, hallucinations, and new AI-specific attacks like jailbreaking and prompt injection attacks.
    • Azure AI works with model providers and partners like HiddenLayer to reduce the risk of cybersecurity breaches, malware, and other signs of tampering.
  4. Simplified model experimentation and selection:
    • Developers can easily experiment and compare different models through the Azure AI Inference API, using a unified set of functions to switch between multiple base models without changing the underlying code.
    • The Azure AI Inference SDK provides client libraries for Python and JavaScript, and will soon support C# and .NET, making it easier to integrate AI into applications.
  5. Enterprise-level integration and access:
    • Organizations can more easily access GitHub Enterprise through an Azure subscription, combining GitHub's cloud-native platform with Azure's enterprise-grade security and scalability.
    • Integration with GitHub through Microsoft Entra ID simplifies user management and access control.

GitHub introduces GitHub Models

GitHub has launched GitHub Models, which allows developers to test and compare different models for free through a built-in interactive model playground, and seamlessly integrate these models into Codespaces and Visual Studio Code. GitHub Models provides a simplified path from model experimentation to production deployment.

  • Model playground:
    • Developers can use the interactive model playground on GitHub to test different prompts and model parameters and explore models from Meta, Mistral, Azure OpenAI Service, Microsoft, and more.
    • Privacy and Security Commitment: GitHub Models does not share hints or outputs with model providers, nor is it used to train or improve models.
  • Model selection and comparison:
    • Developers can test and compare different models in GitHub Models, such as Llama 3.1, GPT-4o, Phi 3, etc., to find the model that best suits their application needs.
    • Provides support for various scenarios, including low latency requirements and multimodal applications.
  • Seamless Integration:
    • Using Codespaces, developers can quickly introduce model inference code into their own projects.
    • You can run prompt evaluations in GitHub Actions or build GitHub Copilot extensions to simplify AI application development.
  • Enterprise-level deployment:
    • Developers can deploy applications to production environments with Azure AI and enjoy built-in responsible AI, enterprise-grade security, and global availability.
    • Replace your GitHub personal access token with your Azure subscription and credentials for a seamless migration.

Further reading:

Official introduction: https://azure.microsoft.com/en-us/blog/accelerating-ai-app-development-with-azure-ai-and-github/

Introduction document: https://docs.github.com/zh/github-models/prototyping-with-ai-models