ms-agent-framework-raglisted

# Microsoft Agent Framework - Agentic RAG System This skill provides scaffolding and guidance for building production-ready Agentic RAG (Retrieval-Augmented Generation) systems using Microsoft Agent Framework with C#. ## Quick Start Use the scaffolding script to create a new RAG system: ```bash scripts/create_rag_system.sh <project-name> [--output-dir <path>] ``` Example: ```bash scripts/create_rag_system.sh MyKnowledgeBase --output-dir ./my-rag-project ``` ## Architecture Overview An Agentic RAG system consists of: 1. **Ingestion Layer**: Document parsing, chunking, and embedding generation 2. **Vector Store**: Semantic search index (Azure AI Search, Qdrant, or Pinecone) 3. **Agent Framework**: Multi-agent orchestration with Microsoft AutoGen 4. **LLM Integration**: Azure OpenAI or OpenAI API for generation 5. **API Layer**: RESTful endpoints for querying ## Core Components ### 1. Semantic Search - Use Azure AI Search for integrated vector + keyword search - Store embeddings with metadata (source, timestamp, tags) - Implement hybrid search (vector + BM25) for best results See `references/semantic_search.md` for implementation details. ### 2. Multi-Agent System Build specialized agents: - **Research Agent**: Finds relevant documents - **Synthesis Agent**: Combines information from multiple sources - **Validation Agent**: Checks accuracy and citations See `references/agent_patterns.md` for agent design patterns. ### 3. Document Processing - Supported formats: