{
  "entities": {
    "text_c-ec4611a61b53": {
      "id": "text_c-ec4611a61b53",
      "name": "Retrieval-Augmented Generation (RAG) is a techniqu…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "Retrieval-Augmented Generation (RAG) is a technique that enhances Large Language Models by retrieving relevant documents from a knowledge base before generating answers. RAG was introduced by Lewis et al. in 2020 and has become a standard approach for grounding LLM outputs in external knowledge. The key advantage of RAG is that it significantly reduces hallucinations by providing the model with verified source material.",
      "_item_text": "Retrieval-Augmented Generation (RAG) is a technique that enhances Large Language Models by retrieving relevant documents from a knowledge base before generating answers. RAG was introduced by Lewis et al. in 2020 and has become a standard approach for grounding LLM outputs in external knowledge. The key advantage of RAG is that it significantly reduces hallucinations by providing the model with verified source material.",
      "_item_page_idx": 0
    },
    "doc-4c5673cd0cc9": {
      "id": "doc-4c5673cd0cc9",
      "name": "rag_intro.txt",
      "type": "document",
      "properties": {
        "doc_id": "b72c4759-f6b6-45fb-a01b-cb086ba472dd"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-55153b8f9577": {
      "id": "text_c-55153b8f9577",
      "name": "The RAG pipeline consists of five core components:…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "The RAG pipeline consists of five core components: 1.Document Ingestion — parsing and chunking source documents into manageable pieces.2.Embedding — converting text chunks into dense vector representations using models like text-embedding-3-small.3.Vector Storage — storing embeddings in a vector database such as Milvus, Pinecone, or ChromaDB.4.Retrieval — finding the most semantically similar documents to a user query using cosine similarity search.",
      "_item_text": "The RAG pipeline consists of five core components: 1.Document Ingestion — parsing and chunking source documents into manageable pieces.2.Embedding — converting text chunks into dense vector representations using models like text-embedding-3-small.3.Vector Storage — storing embeddings in a vector database such as Milvus, Pinecone, or ChromaDB.4.Retrieval — finding the most semantically similar documents to a user query using cosine similarity search.",
      "_item_page_idx": 0
    },
    "text_c-c3c3cde6071f": {
      "id": "text_c-c3c3cde6071f",
      "name": "5.Generation — passing retrieved context along wit…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 1
      },
      "content_type": "text",
      "content_text": "5.Generation — passing retrieved context along with the query to an LLM to produce an accurate, cited response.",
      "_item_text": "5.Generation — passing retrieved context along with the query to an LLM to produce an accurate, cited response.",
      "_item_page_idx": 0
    },
    "doc-f4b87e7b51cb": {
      "id": "doc-f4b87e7b51cb",
      "name": "rag_components.txt",
      "type": "document",
      "properties": {
        "doc_id": "cb8cc41d-4fb5-4760-902a-0a1b2f7b8ecf"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-ef467a62127a": {
      "id": "text_c-ef467a62127a",
      "name": "Milvus is an open-source vector database designed …",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "Milvus is an open-source vector database designed for similarity search at scale. It supports multiple index types including IVF_FLAT, HNSW, and DiskANN. Vector databases store embeddings as high-dimensional vectors and enable efficient approximate nearest neighbor (ANN) search. Key metrics include cosine similarity, Euclidean distance, and inner product.",
      "_item_text": "Milvus is an open-source vector database designed for similarity search at scale. It supports multiple index types including IVF_FLAT, HNSW, and DiskANN. Vector databases store embeddings as high-dimensional vectors and enable efficient approximate nearest neighbor (ANN) search. Key metrics include cosine similarity, Euclidean distance, and inner product.",
      "_item_page_idx": 0
    },
    "doc-1b2e2373a460": {
      "id": "doc-1b2e2373a460",
      "name": "vector_databases.txt",
      "type": "document",
      "properties": {
        "doc_id": "1d971110-0404-499a-a3fd-902221b95293"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-7bd94cd4393c": {
      "id": "text_c-7bd94cd4393c",
      "name": "Text embedding models convert natural language int…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "Text embedding models convert natural language into fixed-length vector representations. OpenAI's text-embedding-3-small produces 1536-dimensional vectors optimized for semantic search. Alternative models include BGE from BAAI, E5 from Microsoft, and multilingual models for cross-lingual retrieval. Embedding quality directly impacts RAG retrieval accuracy.",
      "_item_text": "Text embedding models convert natural language into fixed-length vector representations. OpenAI's text-embedding-3-small produces 1536-dimensional vectors optimized for semantic search. Alternative models include BGE from BAAI, E5 from Microsoft, and multilingual models for cross-lingual retrieval. Embedding quality directly impacts RAG retrieval accuracy.",
      "_item_page_idx": 0
    },
    "doc-a97956d2e621": {
      "id": "doc-a97956d2e621",
      "name": "embedding_models.txt",
      "type": "document",
      "properties": {
        "doc_id": "898fbb98-1b40-4229-99c2-542a8f400d47"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-676838941010": {
      "id": "text_c-676838941010",
      "name": "The ReAct (Reasoning + Acting) framework enables A…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "The ReAct (Reasoning + Acting) framework enables AI agents to interleave reasoning steps with tool-calling actions. ReAct agents follow a Think → Act → Observe loop, where each iteration involves the agent reasoning about what to do next, selecting and invoking a tool, and observing the result. This pattern is fundamental to building reliable AI agents.",
      "_item_text": "The ReAct (Reasoning + Acting) framework enables AI agents to interleave reasoning steps with tool-calling actions. ReAct agents follow a Think → Act → Observe loop, where each iteration involves the agent reasoning about what to do next, selecting and invoking a tool, and observing the result. This pattern is fundamental to building reliable AI agents.",
      "_item_page_idx": 0
    },
    "doc-19e2c95f250b": {
      "id": "doc-19e2c95f250b",
      "name": "react_agent.txt",
      "type": "document",
      "properties": {
        "doc_id": "bbbe6cb6-d79b-4d8f-9629-d3579d4c4239"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-1926476d68ac": {
      "id": "text_c-1926476d68ac",
      "name": "RAG System Architecture Diagram — showing the flow…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "RAG System Architecture Diagram — showing the flow from user query to final answer.",
      "_item_text": "RAG System Architecture Diagram — showing the flow from user query to final answer.",
      "_item_page_idx": 0
    },
    "image-d107b040969f": {
      "id": "image-d107b040969f",
      "name": "image_1",
      "type": "image",
      "properties": {
        "page_idx": 0,
        "content_type": "image",
        "has_caption": false,
        "index": 1
      },
      "content_type": "image",
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "doc-621766721b8c": {
      "id": "doc-621766721b8c",
      "name": "rag_architecture_diagram",
      "type": "document",
      "properties": {
        "doc_id": "86e6838c-1570-45d1-9531-60f8048e5c13"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-723213cefe2f": {
      "id": "text_c-723213cefe2f",
      "name": "RAG performance benchmarks across different config…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "RAG performance benchmarks across different configurations.",
      "_item_text": "RAG performance benchmarks across different configurations.",
      "_item_page_idx": 0
    },
    "table-31e9e6015b3d": {
      "id": "table-31e9e6015b3d",
      "name": "| Configuration | Accuracy | Latency |\n| Naive RAG",
      "type": "table",
      "properties": {
        "page_idx": 0,
        "content_type": "table",
        "has_caption": true,
        "index": 1
      },
      "content_type": "table",
      "content_text": "| Configuration | Accuracy | Latency |\n| Naive RAG | 78% | 200ms |\n| Hybrid RAG | 94% | 350ms |\n| Graph RAG | 96% | 600ms |\n\n[Table]: RAG Performance Comparison\n| Configuration | Accuracy | Latency |\n| Naive RAG | 78% | 200ms |\n| Hybrid RAG | 94% | 350ms |\n| Graph RAG | 96% | 600ms |",
      "_item_text": "| Configuration | Accuracy | Latency |\n| Naive RAG | 78% | 200ms |\n| Hybrid RAG | 94% | 350ms |\n| Graph RAG | 96% | 600ms |",
      "_item_page_idx": 0
    },
    "doc-63f5b9638e1c": {
      "id": "doc-63f5b9638e1c",
      "name": "performance_benchmark",
      "type": "document",
      "properties": {
        "doc_id": "b6908504-9061-41b7-bda6-da3819cdf256"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-c531165a9bda": {
      "id": "text_c-c531165a9bda",
      "name": "HyDE (Hypothetical Document Embeddings) improves r…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "HyDE (Hypothetical Document Embeddings) improves retrieval by generating a hypothetical answer first.",
      "_item_text": "HyDE (Hypothetical Document Embeddings) improves retrieval by generating a hypothetical answer first.",
      "_item_page_idx": 0
    },
    "doc-16aa639a33f8": {
      "id": "doc-16aa639a33f8",
      "name": "test_hyde",
      "type": "document",
      "properties": {
        "doc_id": "be3bc86f-6a10-431c-9c31-71666b2feb23"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "text_c-9a5155983e92": {
      "id": "text_c-9a5155983e92",
      "name": "CLIP model bridges vision and language for cross-m…",
      "type": "text_chunk",
      "properties": {
        "page_idx": 0,
        "content_type": "text",
        "has_caption": false,
        "index": 0
      },
      "content_type": "text",
      "content_text": "CLIP model bridges vision and language for cross-modal retrieval.",
      "_item_text": "CLIP model bridges vision and language for cross-modal retrieval.",
      "_item_page_idx": 0
    },
    "image-dfc8f45882f2": {
      "id": "image-dfc8f45882f2",
      "name": "image_1",
      "type": "image",
      "properties": {
        "page_idx": 0,
        "content_type": "image",
        "has_caption": false,
        "index": 1
      },
      "content_type": "image",
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    },
    "doc-4d768d8694fc": {
      "id": "doc-4d768d8694fc",
      "name": "api_multimodal",
      "type": "document",
      "properties": {
        "doc_id": "24101de9-b27a-4b1a-a63f-96aa92329107"
      },
      "content_type": null,
      "content_text": "",
      "_item_text": "",
      "_item_page_idx": 0
    }
  },
  "relations": [
    {
      "source": "text_c-ec4611a61b53",
      "target": "doc-4c5673cd0cc9",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-55153b8f9577",
      "target": "doc-f4b87e7b51cb",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-c3c3cde6071f",
      "target": "doc-f4b87e7b51cb",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-55153b8f9577",
      "target": "text_c-c3c3cde6071f",
      "type": "nearby",
      "weight": 0.8
    },
    {
      "source": "text_c-ef467a62127a",
      "target": "doc-1b2e2373a460",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-7bd94cd4393c",
      "target": "doc-a97956d2e621",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-676838941010",
      "target": "doc-19e2c95f250b",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-1926476d68ac",
      "target": "doc-621766721b8c",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "image-d107b040969f",
      "target": "doc-621766721b8c",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-1926476d68ac",
      "target": "image-d107b040969f",
      "type": "nearby",
      "weight": 0.8
    },
    {
      "source": "image-d107b040969f",
      "target": "text_c-1926476d68ac",
      "type": "describes",
      "weight": 0.6
    },
    {
      "source": "text_c-723213cefe2f",
      "target": "doc-63f5b9638e1c",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "table-31e9e6015b3d",
      "target": "doc-63f5b9638e1c",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-723213cefe2f",
      "target": "table-31e9e6015b3d",
      "type": "nearby",
      "weight": 0.8
    },
    {
      "source": "table-31e9e6015b3d",
      "target": "text_c-723213cefe2f",
      "type": "describes",
      "weight": 0.6
    },
    {
      "source": "text_c-c531165a9bda",
      "target": "doc-16aa639a33f8",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-9a5155983e92",
      "target": "doc-4d768d8694fc",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "image-dfc8f45882f2",
      "target": "doc-4d768d8694fc",
      "type": "belongs_to",
      "weight": 1.0
    },
    {
      "source": "text_c-9a5155983e92",
      "target": "image-dfc8f45882f2",
      "type": "nearby",
      "weight": 0.8
    },
    {
      "source": "image-dfc8f45882f2",
      "target": "text_c-9a5155983e92",
      "type": "describes",
      "weight": 0.6
    }
  ]
}