Awesome AI Agents › AI Workflow Orchestrators

Azure-Samples/miyagi

⭐ 752 Jupyter Notebook repository created 2023-02-16

Project Miyagi is a comprehensive sample project designed to showcase Microsoft's Copilot Stack for building intelligent, AI-infused enterprise applications. It serves as an envisioning workshop that guides developers through designing, developing, and deploying advanced AI-powered apps that enhance productivity and enable hyper-personalization. The project explores a variety of use cases involving both generative AI and traditional machine learning, providing an experiential approach to integrating AI capabilities into products. It introduces software engineers to emerging AI design patterns such as prompt engineering techniques including chain-of-thought and retrieval-augmentation, vectorization for long-term memory, fine-tuning of open-source models, agent-like orchestration, and the use of plugins or tools to augment and ground large language models (LLMs). Miyagi includes practical examples and implementations using a rich set of AI and cloud-native technologies such as Semantic Kernel, Promptflow, LlamaIndex, LangChain, Azure AI Search, CosmosDB Postgres with pgvector, and generative image utilities like DreamFusion and ControlNet. It also features fine-tuned foundation models from AzureML, including Llama2 and Phi-2, enabling users to modernize and transform their applications with AI and build custom Copilots tailored to private data. The project architecture is based on a cloud-native, event-driven microservices design that ensures enterprise-grade qualities like availability, scalability, and maintainability. It supports various AI application scenarios such as personalized financial coaching, summarization, and agent orchestration. Miyagi also provides a VSCode extension for GitHub Copilot Agent, a ChatGPT plugin, and multiple experimental implementations demonstrating vector stores, knowledge graph memory, and integration with Microsoft Graph API. The project is a work in progress and continuously evolving with incremental implementations of new use cases. It offers a self-guided workshop experience to help developers explore the art of the possible with AI and build intelligent systems using the Copilot stack. The repository includes detailed documentation, architectural diagrams, and links to related resources and demos, making it a valuable resource for developers interested in AI-driven application development on Azure. Keywords: AI, Copilot Stack, Microsoft, Semantic Kernel, Promptflow, LlamaIndex, LangChain, Azure AI Search, CosmosDB, generative AI, machine learning, vector stores, fine-tuning, foundation models, AzureML, microservices, event-driven architecture, personalized coaching, summarization, agent orchestration, VSCode extension, ChatGPT plugin, knowledge graph, Microsoft Graph API, cloud-native, scalable, maintainable, AI workshop, prompt engineering, chain-of-thought, retrieval-augmentation, vectorization, plugins, tools, DreamFusion, ControlNet, Llama2, Phi-2, GitHub Copilot Agent, AI-infused applications, enterprise-grade, intelligent apps, AI transformation, AI capabilities, AI design patterns, generative text, generative images, reinforcement learning, DeepSpeed Chat, AI studio, AI speech, AI search, AutoGen, TaskWeaver, Azure Functions, APIM, Service Bus, Event Grid, Logic Apps, AKS. Summary: Project Miyagi is a sample and workshop project that demonstrates how to build enterprise-grade intelligent applications using Microsoft's Copilot Stack, integrating generative AI, machine learning, and cloud-native microservices to enable AI-infused product experiences with hyper-personalization and advanced AI design patterns.

https://github.com/Azure-Samples/miyagi

agent-orchestrationagentsaiai-capabilitiesai-design-patternsai-infused-applicationsai-searchai-speechai-studioai-transformationai-workshopaksapimassistantsautogenazureazure-ai-searchazure-functionsazure-openaiazureaiazuremlchain-of-thoughtchatgpt-plugincloud-nativecontrolnetcopilotcopilot-stackcosmosdbdeepspeed-chatdreamfusionenterprise-gradeevent-driven-architectureevent-gridfine-tuningfoundation-modelsgenerative-aigenerative-imagesgenerative-textgithub-copilot-agentgpt-4guidanceintelligent-appsknowledge-graphlangchainllama-indexllama2llamaindexlogic-appsmachine-learningmaintainablemicroservicesmicrosoftmicrosoft-graph-apiopenaipersonalized-coachingphi-2pluginsprompt-engineeringpromptflowreinforcement-learningretrieval-augmentationscalablesemantic-kernelservice-bussummarizationtaskweavertoolstypechatvector-storesvectorizationvscode-extension