{
  "id": "knowledge-fabric-impl:chroma-fabric",
  "_kind": "KnowledgeFabricImpl",
  "_file": "agent-stack/knowledge-fabric-impls/vector-store-fabrics.yaml",
  "_cluster": "agent-stack",
  "attributes": {
    "displayName": "ChromaDB as Knowledge Fabric",
    "description": "ChromaDB as a knowledge fabric storage backend. Local-first, embeddable\nin Python and JavaScript, with collections supporting metadata filtering.\nChroma stores documents alongside their embeddings and metadata, returning\nnearest-neighbor results via cosine similarity or other distance metrics.\nRuns in-memory, as a local persistent store, or as a client-server\ndeployment. First-class integrations with LangChain, LlamaIndex, and\nOpenAI make it the default prototyping backend for RAG pipelines. As a\nknowledge fabric, Chroma provides the vector storage layer for semantic\nretrieval of organizational knowledge.\n",
    "knowledgeFileFormats": [
      "embeddings",
      "document-chunks"
    ],
    "retrievalStrategy": "semantic-search",
    "knowledgePersistence": "vector-store",
    "knowledgeScopes": [
      "project",
      "organization"
    ],
    "autoExtractionSupport": false,
    "notes": "Chroma is often the first vector store developers reach for due to its\nsimplicity and zero-config local mode. As a knowledge fabric backend,\nit excels for small-to-medium knowledge bases (< 1M documents) where\noperational simplicity matters. For larger deployments, distributed\nstores like Milvus or managed services like Pinecone are preferred.\n"
  },
  "outgoingEdges": [
    {
      "from": "knowledge-fabric-impl:chroma-fabric",
      "to": "layer:12-knowledge-fabric",
      "kind": "realizes",
      "attributes": {}
    },
    {
      "from": "knowledge-fabric-impl:chroma-fabric",
      "to": "tool:chromadb",
      "kind": "integrates_with",
      "attributes": {}
    }
  ],
  "incomingEdges": []
}