{
  "code": "RAG",
  "collection": "architecture",
  "href": "https://banes-lab.com/ontology#architecture-retrieval-augmented-generation",
  "id": "retrieval-augmented-generation",
  "kind": "pattern",
  "layer": {
    "href": "https://banes-lab.com/ontology/schema#layer-correctness-core",
    "json": "https://banes-lab.com/json/records/layer/correctness-core",
    "label": "Correctness Core",
    "markdown": "https://banes-lab.com/records/layer/correctness-core.md",
    "ref": "layer:correctness-core"
  },
  "name": "Retrieval-Augmented Generation",
  "ref": "architecture:retrieval-augmented-generation",
  "relations": [
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-retriever",
          "json": "https://banes-lab.com/json/records/lexicon/retriever",
          "label": "Retriever",
          "markdown": "https://banes-lab.com/records/lexicon/retriever.md",
          "ref": "lexicon:retriever"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-document-store",
          "json": "https://banes-lab.com/json/records/lexicon/document-store",
          "label": "Document Store",
          "markdown": "https://banes-lab.com/records/lexicon/document-store.md",
          "ref": "lexicon:document-store"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-grounding-strategy",
          "json": "https://banes-lab.com/json/records/lexicon/grounding-strategy",
          "label": "Grounding Strategy",
          "markdown": "https://banes-lab.com/records/lexicon/grounding-strategy.md",
          "ref": "lexicon:grounding-strategy"
        }
      ],
      "relation": "requires"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-explainability",
          "json": "https://banes-lab.com/json/records/architecture/explainability",
          "label": "Explainability",
          "markdown": "https://banes-lab.com/records/architecture/explainability.md",
          "ref": "architecture:explainability"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-knowledge-freshness",
          "json": "https://banes-lab.com/json/records/lexicon/knowledge-freshness",
          "label": "Knowledge Freshness",
          "markdown": "https://banes-lab.com/records/lexicon/knowledge-freshness.md",
          "ref": "lexicon:knowledge-freshness"
        }
      ],
      "relation": "reinforces"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-contextual-generation",
          "json": "https://banes-lab.com/json/records/lexicon/contextual-generation",
          "label": "Contextual Generation",
          "markdown": "https://banes-lab.com/records/lexicon/contextual-generation.md",
          "ref": "lexicon:contextual-generation"
        }
      ],
      "relation": "enables"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-ungrounded-generation",
          "json": "https://banes-lab.com/json/records/lexicon/ungrounded-generation",
          "label": "Ungrounded Generation",
          "markdown": "https://banes-lab.com/records/lexicon/ungrounded-generation.md",
          "ref": "lexicon:ungrounded-generation"
        }
      ],
      "relation": "conflicts-with"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-retrieval-quality-latency",
          "json": "https://banes-lab.com/json/records/lexicon/retrieval-quality-latency",
          "label": "Retrieval Quality/Latency",
          "markdown": "https://banes-lab.com/records/lexicon/retrieval-quality-latency.md",
          "ref": "lexicon:retrieval-quality-latency"
        }
      ],
      "relation": "tensions-with"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/schema#tension-retrieval-augmented-generation-rag-retrieval-quality-latency",
          "json": "https://banes-lab.com/json/records/tension/retrieval-augmented-generation-rag-retrieval-quality-latency",
          "label": "Retrieval-Augmented Generation (RAG) / Retrieval Quality/Latency",
          "markdown": "https://banes-lab.com/records/tension/retrieval-augmented-generation-rag-retrieval-quality-latency.md",
          "ref": "tension:retrieval-augmented-generation-rag-retrieval-quality-latency"
        }
      ],
      "relation": "tensions"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-ungrounded-generation",
          "json": "https://banes-lab.com/json/records/lexicon/ungrounded-generation",
          "label": "Ungrounded Generation",
          "markdown": "https://banes-lab.com/records/lexicon/ungrounded-generation.md",
          "ref": "lexicon:ungrounded-generation"
        }
      ],
      "relation": "violated-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-retriever",
          "json": "https://banes-lab.com/json/records/lexicon/retriever",
          "label": "Retriever",
          "markdown": "https://banes-lab.com/records/lexicon/retriever.md",
          "ref": "lexicon:retriever"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-evidence-citation",
          "json": "https://banes-lab.com/json/records/lexicon/evidence-citation",
          "label": "Evidence Citation",
          "markdown": "https://banes-lab.com/records/lexicon/evidence-citation.md",
          "ref": "lexicon:evidence-citation"
        }
      ],
      "relation": "refactored-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology/schema#vocabulary-severity-contextual",
          "json": "https://banes-lab.com/json/records/vocabulary/severity-contextual",
          "label": "contextual",
          "markdown": "https://banes-lab.com/records/vocabulary/severity-contextual.md",
          "ref": "vocabulary:severity-contextual"
        }
      ],
      "relation": "severity"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-category-model-architecture",
          "json": "https://banes-lab.com/json/records/architecture-category/model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/records/architecture-category/model-architecture.md",
          "ref": "architecture-category:model-architecture"
        }
      ],
      "relation": "category"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-ungrounded-content",
          "json": "https://banes-lab.com/json/records/architecture/ungrounded-content",
          "label": "Ungrounded Content",
          "markdown": "https://banes-lab.com/records/architecture/ungrounded-content.md",
          "ref": "architecture:ungrounded-content"
        }
      ],
      "relation": "refactors"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/ontology#architecture-vector-search",
          "json": "https://banes-lab.com/json/records/architecture/vector-search",
          "label": "Vector Search",
          "markdown": "https://banes-lab.com/records/architecture/vector-search.md",
          "ref": "architecture:vector-search"
        }
      ],
      "relation": "reinforced-by"
    },
    {
      "links": [
        {
          "href": "https://banes-lab.com/software-architecture/scale#systems-built-around-a-model",
          "json": "https://banes-lab.com/json/software-architecture/scale/systems-built-around-a-model",
          "label": "Systems built around a model",
          "markdown": "https://banes-lab.com/software-architecture/scale/systems-built-around-a-model.md",
          "ref": "chapter:/software-architecture/scale#systems-built-around-a-model"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-category-anti-patterns",
          "json": "https://banes-lab.com/json/ontology/principles/architecture-category-anti-patterns",
          "label": "Anti-patterns",
          "markdown": "https://banes-lab.com/ontology/principles/architecture-category-anti-patterns.md",
          "ref": "chapter:/ontology#architecture-category-anti-patterns"
        },
        {
          "href": "https://banes-lab.com/ontology#architecture-category-model-architecture",
          "json": "https://banes-lab.com/json/ontology/principles/architecture-category-model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/ontology/principles/architecture-category-model-architecture.md",
          "ref": "chapter:/ontology#architecture-category-model-architecture"
        },
        {
          "href": "https://banes-lab.com/ontology/lexicon#lexicon-category-model-architecture",
          "json": "https://banes-lab.com/json/ontology/lexicon/lexicon-category-model-architecture",
          "label": "Model Architecture",
          "markdown": "https://banes-lab.com/ontology/lexicon/lexicon-category-model-architecture.md",
          "ref": "chapter:/ontology/lexicon#lexicon-category-model-architecture"
        },
        {
          "href": "https://banes-lab.com/ontology/schema#the-vocabulary-severity",
          "json": "https://banes-lab.com/json/ontology/schema/the-vocabulary-severity",
          "label": "Severity levels",
          "markdown": "https://banes-lab.com/ontology/schema/the-vocabulary-severity.md",
          "ref": "chapter:/ontology/schema#the-vocabulary-severity"
        },
        {
          "href": "https://banes-lab.com/ontology/schema#the-resolutions",
          "json": "https://banes-lab.com/json/ontology/schema/the-resolutions",
          "label": "The resolutions",
          "markdown": "https://banes-lab.com/ontology/schema/the-resolutions.md",
          "ref": "chapter:/ontology/schema#the-resolutions"
        }
      ],
      "relation": "linked-from"
    }
  ],
  "summary": "A design pattern that retrieves relevant documents for a question and passes them to the model, so the answer can cite its sources.",
  "siblings": {
    "next": {
      "href": "https://banes-lab.com/ontology#architecture-vector-search",
      "json": "https://banes-lab.com/json/records/architecture/vector-search",
      "label": "Vector Search",
      "markdown": "https://banes-lab.com/records/architecture/vector-search.md",
      "ref": "architecture:vector-search"
    },
    "previous": {
      "href": "https://banes-lab.com/ontology#architecture-model-inference",
      "json": "https://banes-lab.com/json/records/architecture/model-inference",
      "label": "Model Inference",
      "markdown": "https://banes-lab.com/records/architecture/model-inference.md",
      "ref": "architecture:model-inference"
    }
  },
  "up": {
    "href": null,
    "json": "https://banes-lab.com/json/api/records/architecture",
    "label": "Architecture principles",
    "markdown": "https://banes-lab.com/api/records/architecture.md",
    "ref": "api:records/architecture"
  },
  "aliases": [
    "RAG"
  ],
  "exemplar": {
    "after": "const query = normalizeFooQuery(question);\nconst documents = await fooRetriever.search(query, { topK: 8 });\nconst groundedPrompt = buildGroundedFooPrompt(question, documents);\nconst answer = await model.generate(groundedPrompt);\nreturn attachCitations(answer, documents);",
    "before": "const answer = await model.generate(`Answer: ${question}`);",
    "lang": "ts",
    "medium": "code"
  },
  "formedBy": null,
  "scope": [
    "language-model system",
    "knowledge retrieval"
  ],
  "severity": "contextual"
}
