{
  "id": "knowledge-fabric-impl:haystack-rag-fabric",
  "_kind": "KnowledgeFabricImpl",
  "_file": "agent-stack/knowledge-fabric-impls/rag-framework-fabrics.yaml",
  "_cluster": "agent-stack",
  "attributes": {
    "displayName": "Haystack as RAG Knowledge Fabric",
    "description": "Haystack (by deepset) as a pipeline-based knowledge fabric. Provides\ncomposable pipeline architecture with typed components for document\npreprocessing, embedding, retrieval (dense, sparse, hybrid), re-ranking,\nand generation. Pipeline-as-code design with YAML serialization enables\nreproducible knowledge fabric configurations. Supports multiple document\nstore backends (Elasticsearch, OpenSearch, Weaviate, Qdrant, Chroma,\nPinecone). As a knowledge fabric, Haystack's pipeline abstraction makes\nretrieval configurations explicit, testable, and version-controllable.\n",
    "knowledgeFileFormats": [
      "any (via converters)"
    ],
    "retrievalStrategy": "hybrid",
    "knowledgePersistence": "delegated (via document store backends)",
    "knowledgeScopes": [
      "project",
      "organization",
      "enterprise"
    ],
    "autoExtractionSupport": false,
    "notes": "Haystack's pipeline YAML serialization is its distinguishing knowledge\nfabric feature — retrieval configurations become version-controlled\nartifacts. This supports knowledge fabric reproducibility and evolution\ntracking. The typed component system (v2) enforces interface contracts\nbetween pipeline stages, reducing configuration errors compared to\nLangChain's more dynamic composition model.\n"
  },
  "outgoingEdges": [
    {
      "from": "knowledge-fabric-impl:haystack-rag-fabric",
      "to": "layer:12-knowledge-fabric",
      "kind": "realizes",
      "attributes": {}
    },
    {
      "from": "knowledge-fabric-impl:haystack-rag-fabric",
      "to": "tool:haystack",
      "kind": "integrates_with",
      "attributes": {}
    }
  ],
  "incomingEdges": []
}