Skip to content

CHANGELOG.md

Latest commit

 

History

History
309 lines (257 loc) · 17.1 KB

CHANGELOG.md

File metadata and controls

309 lines (257 loc) · 17.1 KB

Changelog

All notable changes to EdgeAI for Beginners are documented here. This project uses date-based entries and the Keep a Changelog style (Added, Changed, Fixed, Removed, Docs, Moved).

2025-10-30

Added - Module06 AI Agents Comprehensive Enhancement

  • Microsoft Agent Framework Integration (Module06/01.IntroduceAgent.md):

    • Complete section on Microsoft Agent Framework for production-ready agent development
    • Detailed integration patterns with Foundry Local for edge deployment
    • Multi-agent orchestration examples with specialized SLM models
    • Enterprise deployment patterns with resource management and monitoring
    • Security and compliance features for edge agent systems
    • Real-world implementation examples (retail, healthcare, customer service)

  • Production SLM Agent Deployment Strategies:

    • Foundry Local: Complete enterprise-grade edge AI runtime documentation with installation, configuration, and production patterns
    • Ollama: Enhanced community-focused deployment with comprehensive monitoring and model management
    • VLLM: High-performance inference engine with advanced optimization techniques and enterprise features
    • Production deployment checklists and comparison tables for all three platforms

  • Edge-Optimized SLM Frameworks Enhancement:

    • ONNX Runtime: New comprehensive section for cross-platform SLM agent deployment
    • Universal deployment patterns across Windows, Linux, macOS, iOS, and Android
    • Hardware acceleration options (CPU, GPU, NPU) with automatic detection
    • Production-ready features and agent-specific optimizations
    • Complete implementation examples with Microsoft Agent Framework integration

  • References and Further Reading:

    • Comprehensive resource library with 100+ authoritative sources
    • Core research papers on AI agents and Small Language Models
    • Official documentation for all major frameworks and tools
    • Industry reports, market analysis, and technical benchmarks
    • Educational resources, conferences, and community forums
    • Standards, specifications, and compliance frameworks

Changed - Module06 Content Modernization

  • Enhanced Learning Objectives: Added Microsoft Agent Framework mastery and edge deployment capabilities
  • Production Focus: Shifted from conceptual to implementation-ready guidance with production examples
  • Code Examples: Updated all examples to use modern SDK patterns and best practices
  • Architecture Patterns: Added hierarchical agent architectures and edge-to-cloud coordination
  • Performance Optimization: Enhanced with resource management and auto-scaling recommendations

Docs - Module06 Structure Enhancement

  • Comprehensive Agent Framework Coverage: From basic concepts to enterprise deployment
  • Production Deployment Strategies: Complete guides for Foundry Local, Ollama, and VLLM
  • Cross-Platform Optimization: Added ONNX Runtime for universal deployment
  • Resource Library: Extensive references for continued learning and implementation

Added - Module06 Model Context Protocol (MCP) Documentation Update

  • MCP Introduction Modernization (Module06/03.IntroduceMCP.md):

    • Updated with latest MCP specifications from modelcontextprotocol.io (2025-06-18 version)
    • Added official USB-C analogy for standardized AI application connections
    • Updated architecture section with official two-layer design (Data Layer + Transport Layer)
    • Enhanced core primitives documentation with server primitives (Tools, Resources, Prompts) and client primitives (Sampling, Elicitation, Logging)

  • Comprehensive MCP References and Resources:

    • Added MCP for Beginners link (https://aka.ms/mcp-for-beginners)
    • Official MCP documentation and specifications (modelcontextprotocol.io)
    • Development resources including MCP Inspector and reference implementations
    • Technical standards (JSON-RPC 2.0, JSON Schema, OpenAPI, Server-Sent Events)

Added - Module04 Qualcomm QNN Integration

  • New Section 7: Qualcomm QNN Optimization Suite (Module04/05.QualcommQNN.md):

    • Comprehensive 400+ line guide covering Qualcomm's unified AI inference framework
    • Detailed coverage of heterogeneous computing (Hexagon NPU, Adreno GPU, Kryo CPU)
    • Hardware-aware optimization for Snapdragon platforms with intelligent workload distribution
    • Advanced quantization techniques (INT8, INT16, mixed-precision) for mobile deployment
    • Power-efficient inference optimization for battery-powered devices and real-time applications
    • Complete installation guide with QNN SDK setup and environment configuration
    • Practical examples: PyTorch to QNN conversion, multi-backend optimization, context binary generation
    • Advanced usage patterns: custom backend configuration, dynamic quantization, performance profiling
    • Comprehensive troubleshooting section and community resources

  • Enhanced Module04 structure:

    • Updated README.md to include 7 progressive sections (was 6)
    • Added Qualcomm QNN to performance benchmarks table (5-15x speed improvement, 50-80% memory reduction)
    • Comprehensive learning outcomes for mobile AI deployment and power optimization

Changed - Module04 Documentation Updates

  • Microsoft Olive documentation enhancement (Module04/03.MicrosoftOlive.md):

    • Added comprehensive "Olive Recipes Repository" section covering 100+ pre-built optimization recipes
    • Detailed coverage of supported model families (Phi, Llama, Qwen, Gemma, Mistral, DeepSeek)
    • Practical usage examples for recipe customization and community contributions
    • Enhanced with performance benchmarks and integration guidance

  • Section reordering in Module04:

    • Apple MLX moved to Section 5 (was Section 6)
    • Workflow Synthesis moved to Section 6 (was Section 7)
    • Qualcomm QNN positioned as Section 7 (specialized mobile/edge focus)
    • Updated all file references and navigation links accordingly

Fixed - Workshop Sample Validation

  • chat_bootstrap.py validation and repair:
    • Fixed corrupted import statement (util.util.workshop_utils → util.workshop_utils)
    • Created missing __init__.py in util package for proper Python module resolution
    • Installed required dependencies (openai, foundry-local-sdk) in conda environment
    • Successfully validated sample execution with both default and custom prompts
    • Confirmed integration with Foundry Local service and model loading (phi-4-mini with CUDA optimization)

Docs - Comprehensive Guide Updates

  • Module04 README.md complete restructure:

    • Added Qualcomm QNN as major optimization framework alongside OpenVINO, Olive, MLX
    • Updated chapter learning outcomes to include mobile AI deployment and power optimization
    • Enhanced performance comparison table with QNN metrics and mobile/edge use cases
    • Maintained logical progression from enterprise solutions to platform-specific optimizations

  • Cross-references and navigation:

    • Updated all internal links and file references for new section numbering
    • Enhanced workflow synthesis description to include mobile, desktop, and cloud environments
    • Added comprehensive resource links for Qualcomm developer ecosystem

2025-10-08

Added - Workshop Comprehensive Update

  • Workshop README.md complete rewrite:

    • Added comprehensive introduction explaining Edge AI value proposition (privacy, performance, cost)
    • Created 6 core learning objectives with detailed competencies
    • Added learning outcomes table with deliverables and competency matrix
    • Included career-ready skills section for industry relevance
    • Added quick start guide with prerequisites and 3-step setup
    • Created resource tables for Python samples (8 files with run times)
    • Added Jupyter notebooks table (8 notebooks with difficulty ratings)
    • Created documentation table (7 key docs with "Use When" guidance)
    • Added learning path recommendations for different skill levels

  • Workshop validation and testing infrastructure:

    • Created scripts/validate_samples.py - Comprehensive validation tool for syntax, imports, and best practices
    • Created scripts/test_samples.py - Smoke test runner for all Python samples
    • Added validation documentation to scripts/README.md

  • Comprehensive documentation:

    • Created SAMPLES_UPDATE_SUMMARY.md - 400+ line detailed guide covering all improvements
    • Created UPDATE_COMPLETE.md - Executive summary of update completion
    • Created QUICK_REFERENCE.md - Quick reference card for Workshop

Changed - Workshop Python Sample Modernization

  • All 8 Python samples updated with best practices:

    • Enhanced error handling with try-except blocks around all I/O operations
    • Added type hints and comprehensive docstrings
    • Implemented consistent [INFO]/[ERROR]/[RESULT] logging pattern
    • Protected optional imports with installation hints
    • Improved user feedback throughout all samples

  • session01/chat_bootstrap.py:

    • Enhanced client initialization with comprehensive error messages
    • Improved streaming error handling with chunk validation
    • Added better exception handling for service unavailability

  • session02/rag_pipeline.py:

    • Added import guards for sentence-transformers with installation hints
    • Enhanced error handling for embedding and generation operations
    • Improved output formatting with structured results

  • session02/rag_eval_ragas.py:

    • Protected optional imports (ragas, datasets) with user-friendly error messages
    • Added error handling for evaluation metrics
    • Enhanced output formatting for evaluation results

  • session03/benchmark_oss_models.py:

    • Implemented graceful degradation (continues on model failures)
    • Added detailed progress reporting and per-model error handling
    • Enhanced statistics calculation with comprehensive error recovery

  • session04/model_compare.py:

    • Added type hints (Tuple return types)
    • Enhanced output formatting with structured JSON results
    • Implemented per-model error handling with recovery

  • session05/agents_orchestrator.py:

    • Enhanced Agent.act() with comprehensive docstrings
    • Added pipeline error handling with stage-by-stage logging
    • Improved memory management and state tracking

  • session06/models_router.py:

    • Enhanced function documentation for all routing components
    • Added detailed logging in route() function
    • Improved test output with structured results

  • session06/models_pipeline.py:

    • Added error handling to chat() helper function
    • Enhanced pipeline() with stage logging and progress reporting
    • Improved main() with comprehensive error recovery

Docs - Workshop Documentation Enhancement

  • Updated main README.md with Workshop section highlighting hands-on learning path
  • Enhanced STUDY_GUIDE.md with comprehensive Workshop section including:
    • Learning objectives and study focus areas
    • Self-assessment questions
    • Hands-on exercises with time estimates
    • Time allocation for concentrated and part-time study
    • Added Workshop to progress tracking template
  • Updated time allocation guide from 20 hours to 30 hours (including Workshop)
  • Added Workshop sample descriptions and learning outcomes to README

Fixed

  • Resolved inconsistent error handling patterns across Workshop samples
  • Fixed optional dependency import errors with proper guards
  • Corrected missing type hints in critical functions
  • Addressed insufficient user feedback in error scenarios
  • Fixed validation issues with comprehensive testing infrastructure

2025-09-23

Changed - Major Module 08 Modernization

  • Comprehensive alignment with Microsoft Foundry-Local repository patterns

    • Updated all code examples to use modern FoundryLocalManager and OpenAI SDK integration
    • Replaced deprecated manual requests calls with proper SDK usage
    • Aligned implementation patterns with official Microsoft documentation and samples

  • 05.AIPoweredAgents.md modernization:

    • Updated multi-agent orchestration to use modern SDK patterns
    • Enhanced coordinator implementation with advanced features (feedback loops, performance monitoring)
    • Added comprehensive error handling and service health checking
    • Integrated proper references to local samples (samples/05/multi_agent_orchestration.ipynb)
    • Updated function calling examples to use modern tools parameter instead of deprecated functions
    • Added production-ready patterns with monitoring and statistics tracking

  • 06.ModelsAsTools.md complete rewrite:

    • Replaced basic tool registry with intelligent model router implementation
    • Added keyword-based model selection for different task types (general, reasoning, code, creative)
    • Integrated environment-based configuration with flexible model assignment
    • Enhanced with comprehensive service health monitoring and error handling
    • Added production deployment patterns with request monitoring and performance tracking
    • Aligned with local implementation in samples/06/router.py and samples/06/model_router.ipynb

  • Documentation structure improvements:

    • Added overview sections highlighting modernization and SDK alignment
    • Enhanced with emojis and better formatting for improved readability
    • Added proper references to local sample files throughout documentation
    • Included production-ready implementation guidance and best practices

Added

  • Comprehensive overview sections in Module 08 files highlighting modern SDK integration
  • Architecture highlights showcasing advanced features (multi-agent systems, intelligent routing)
  • Direct references to local sample implementations for hands-on experience
  • Production deployment guidance with monitoring and error handling patterns
  • Interactive Jupyter notebook examples with advanced features and benchmarks

Fixed

  • Alignment discrepancies between documentation and actual sample implementations
  • Outdated SDK usage patterns throughout Module 08
  • Missing references to comprehensive local sample library
  • Inconsistent implementation approaches across different sections

2025-09-18

Added

  • Module 08: Microsoft Foundry Local – Complete Developer Toolkit
    • Six sessions: setup, Azure AI Foundry integration, open-source models, cutting-edge demos, agents, and models-as-tools
    • Runnable samples under Module08/samples/01–06 with Windows cmd instructions
      • 01 REST quick chat (chat_quickstart.py)
      • 02 SDK quickstart with OpenAI/Foundry Local and Azure OpenAI support (sdk_quickstart.py)
      • 03 CLI list-and-bench (list_and_bench.cmd)
      • 04 Chainlit demo (app.py)
      • 05 Multi-agent orchestration (python -m samples.05.agents.coordinator)
      • 06 Models-as-Tools router (router.py)
  • Azure OpenAI support in Session 2 SDK sample with environment variable configuration
  • .vscode/settings.json to point to Module08/.venv and improve Python analysis resolution
  • .env with PYTHONPATH hint for VS Code/Pylance awareness

Changed

  • Default model updated to phi-4-mini across Module 08 docs and samples; removed remaining phi-3.5 mentions within Module 08
  • Router (Module08/samples/06/router.py) improvements:
    • Endpoint discovery via foundry service status with regex parsing
    • /v1/models health check on startup
    • Env-configurable model registry (GENERAL_MODEL, REASONING_MODEL, CODE_MODEL, TOOL_REGISTRY JSON)
  • Requirements updated: Module08/requirements.txt now includes openai (alongside requests, chainlit)
  • Chainlit sample guidance clarified and troubleshooting added; import resolution via workspace settings

Fixed

  • Resolved import issues:
    • Router no longer depends on a non-existent utils module; functions are inlined
    • Coordinator uses relative import (from .specialists import ...) and is invoked via module path
    • VS Code/Pylance configuration to resolve chainlit and package imports
  • Corrected minor typo in STUDY_GUIDE.md and added Module 08 coverage

Removed

  • Deleted unused Module08/infra/obs.py and removed the empty infra/ directory; observability patterns retained as optional in docs

Moved

  • Consolidated Module 08 demos under Module08/samples with session-numbered folders
    • Moved Chainlit app to samples/04
    • Moved agents to samples/05 and added __init__.py files for package resolution

Docs

  • Module 08 session docs and all sample READMEs enriched with Microsoft Learn and trusted vendor references
  • Module08/README.md updated with Samples Overview, router configuration, and validation tips
  • Module07/README.md Windows Foundry Local section validated against Learn docs
  • STUDY_GUIDE.md updated:
    • Added Module 08 to overview, schedules, progress tracker
    • Added comprehensive References section (Foundry Local, Azure AI, Olive, ONNX Runtime, OpenVINO, MLX, Llama.cpp, vLLM, Ollama, AI Toolkit, Windows ML)

Historical (summary)

  • Course architecture and modules established (Modules 01–07)
  • Iterative content modernization, formatting standardization, and added case studies
  • Expanded optimization frameworks coverage (Llama.cpp, Olive, OpenVINO, Apple MLX)

Unreleased / Backlog (proposals)

  • Optional per-sample smoke tests to validate Foundry Local availability
  • Review translations to align model references (e.g., phi-4-mini) where appropriate
  • Add minimal pyright config if teams prefer workspace-wide strictness
⚡