|
1 | 1 | # What's New in AI Toolkit for VS Code Preview |
2 | 2 |
|
| 3 | +## Version 0.20.0 |
| 4 | +This major milestone release introduces significant enhancements and powerful new capabilities: |
| 5 | + |
| 6 | +- **Reimagined Agent Builder**: Experience a completely redesigned Agent Builder with enhanced usability and functionality: |
| 7 | + |
| 8 | + |
| 9 | + - **Conversation-Centric Interface**: Navigate effortlessly through a redesigned UI that puts conversations at the center, streamlining agent interactions and management. |
| 10 | + - **Flexible Modular Architecture**: Create your AI agent using a modular interface for prompt editing, model selection, and tool configuration — all seamlessly integrated for maximum flexibility. |
| 11 | + - **Unified Agent Management**: Access all your agents from a single, centralized location in **My Resources**, making it simple to create, edit, and switch between different agent configurations. |
| 12 | + |
| 13 | +- **Local Model Support from Foundry**: Run AI models directly on your machine with our new Foundry Local Models integration: |
| 14 | + |
| 15 | + |
| 16 | + - **Seamless Model Discovery**: Browse and select from an extensive catalog of Foundry Local Models with just a few clicks. |
| 17 | + - **Integrated Local Execution**: Run models locally through the Model Playground without external dependencies. |
| 18 | + - **GitHub Copilot Integration**: Leverage Foundry Local Models directly within GitHub Copilot for enhanced development experiences. |
| 19 | + |
| 20 | +- **Enhanced MCP Experience**: Enjoy a completely overhauled Model Context Protocol (MCP) experience with improved functionality: |
| 21 | + |
| 22 | + - **Latest Specification Support**: Full compatibility with the newest MCP specification, including streamable HTTP transport and OAuth authentication capabilities. |
| 23 | + - **Intuitive Configuration Interface**: Configure MCP servers effortlessly through our redesigned quick-pick UI, featuring instant access to popular servers and direct configuration file editing. |
| 24 | + - **Streamlined Server Management**: Manage all your MCP servers from a unified location in **My Resources**, with simplified lifecycle management and configuration options. |
| 25 | + |
| 26 | +- **GitHub Copilot Tool Integration**: Accelerate AI-powered development with new tools designed specifically for GitHub Copilot: |
| 27 | + |
| 28 | + - **AI Model Guidance Tool**: Get intelligent recommendations for selecting and implementing AI models in your applications. |
| 29 | + - **Advanced Tracing Tool**: Implement comprehensive tracing instrumentation with automatic deeplink generation for seamless visualization in AI Toolkit. |
| 30 | + |
| 31 | +## Version 0.18.3 |
| 32 | +This incremental release introduces the latest [GPT-5 family of models](https://openai.com/index/introducing-gpt-5/) and OpenAI's open source models [GPT OSS](https://azure.microsoft.com/en-us/blog/openais-open%E2%80%91source-model-gpt%E2%80%91oss-on-azure-ai-foundry-and-windows-ai-foundry/) to AI Toolkit. |
| 33 | + |
| 34 | +- **GPT-5 Family of Models:** Now available in AI Toolkit through GitHub, Azure AI Foundry, or direct OpenAI access |
| 35 | + |
| 36 | + - **[gpt-5](https://github.com/marketplace/models/azure-openai/gpt-5):** Delivers advanced reasoning for analytics and code generation. |
| 37 | + - **[gpt-5-mini](https://github.com/marketplace/models/azure-openai/gpt-5-mini):** Optimized for low-cost, fast experiences such as real-time agents and tool orchestration for customer support. |
| 38 | + - **[gpt-5-nano](https://github.com/marketplace/models/azure-openai/gpt-5-nano):** Designed for speed and efficiency, offering strong reasoning for concise question answering and responses. |
| 39 | + - **[gpt-5-chat (preview)](https://github.com/marketplace/models/azure-openai/gpt-5-chat):** Enables natural, multimodal, multi-turn conversations with persistent context for agentic workflows. |
| 40 | + |
| 41 | + |
| 42 | +- **GPT OSS Models:** Via Azure AI Foundry and Local ONNX Runtime |
| 43 | + |
| 44 | + - **[gpt-oss-120b](https://ai.azure.com/catalog/models/gpt-oss-120b) via Azure AI Foundry:** gpt-oss-120b is an open-weight model with the following key characteristics: Text-in/text-out only Reasoning with chain-of-thought (with ability to change reasoning effort to low, medium, or high) Support for structured outputs API compatible with Responses |
| 45 | + - **[gpt-oss-20b-cuda-gpu](https://ai.azure.com/catalog/models/gpt-oss-20b-cuda-gpu) via Local ONNX Runtime:** This model is an optimized version of gpt-oss-20b to enable local inference on CUDA GPUs. This model uses RTN quantization. |
| 46 | + |
| 47 | +- **Experience GPT-5 and GPT OSS Capabilities** |
| 48 | + - Compare GPT-5 models with others in the catalog to evaluate their capabilities. |
| 49 | +  |
| 50 | + - Integrate GPT-5 models into your applications using the `View Code` feature in Model Playground, with support for Azure AI Inference SDK, OpenAI SDK, and multiple programming languages. |
| 51 | +  |
| 52 | + |
| 53 | +## Version 0.18.0 |
| 54 | +This is a major milestone release with significant updates and new features: |
| 55 | +- **Local Observability for AI applications and Agents**: You can trace the execution of your AI applications, including interactions with generative AI models and use AI Toolkit tracing visualization tool to gain insights into their behavior and performance. |
| 56 | + |
| 57 | + |
| 58 | +- **AI-Powered Prompt Improver**: The new AI-powered prompt improver helps you refine your prompts for better results. It analyzes your prompts and suggests improvements to enhance their effectiveness. |
| 59 | + |
| 60 | + |
| 61 | +- **View Code Language Expansion**: The "View Code" feature in the Agent Builder and Playground now supports multiple programming languages including Python, JavaScript, C# and Java. This allows you to see how your prompts can be implemented in different programming languages, making it easier to integrate AI capabilities into your applications. |
| 62 | + |
| 63 | + |
| 64 | +- **Enhancements and Bug Fixes**: |
| 65 | + - Model sections will be automatically expanded when you apply a filter, making it easier to navigate through models. |
| 66 | + - Added CUDA models to the model catalog, providing more options for GPU-accelerated AI applications. |
| 67 | + - Fixed an issue where MCP initialization in generated code does not carry headers. [#243](https://github.com/microsoft/vscode-ai-toolkit/issues/243) |
| 68 | + |
3 | 69 | ## Version 0.16.0 |
4 | 70 | This is a major milestone release featuring significant enhancements and exciting new capabilities: |
5 | 71 |
|
|
0 commit comments