RAG - Cargar Secciones (api)

The workflow JSON

Copy or download the full n8n JSON below. Paste it into a new n8n workflow, add your credentials, activate. Full import guide →

{
  "name": "RAG - Cargar Secciones (API)",
  "nodes": [
    {
      "parameters": {
        "httpMethod": "POST",
        "path": "rag-cargar-seccion",
        "authentication": "headerAuth",
        "responseMode": "lastNode",
        "responseData": "noData",
        "options": {}
      },
      "id": "3f1c9a20-0000-4000-8000-000000000001",
      "name": "Recibir Seccion",
      "type": "n8n-nodes-base.webhook",
      "typeVersion": 2.1,
      "position": [
        240,
        300
      ],
      "notes": ">>> REPLACE: attach a Header Auth credential. The script sends the token in X-Carga-Token. responseMode is 'lastNode' on purpose: the webhook answers when indexing FINISHES, so the loading script becomes genuinely sequential and never has two sections vectorising at once."
    },
    {
      "parameters": {
        "model": "bge-m3:latest"
      },
      "id": "3f1c9a20-0000-4000-8000-000000000002",
      "name": "Embeddings Locales",
      "type": "@n8n/n8n-nodes-langchain.embeddingsOllama",
      "typeVersion": 1,
      "position": [
        480,
        520
      ],
      "notes": ">>> REPLACE: attach an Ollama credential pointing at your instance. This model must be IDENTICAL to the one used at query time \u2014 changing it invalidates every vector already stored."
    },
    {
      "parameters": {
        "chunkOverlap": 150,
        "options": {}
      },
      "id": "3f1c9a20-0000-4000-8000-000000000003",
      "name": "Partir en Fragmentos",
      "type": "@n8n/n8n-nodes-langchain.textSplitterRecursiveCharacterTextSplitter",
      "typeVersion": 1,
      "position": [
        700,
        700
      ],
      "notes": "chunkSize stays at the default 1000. Research said size is the wrong lever \u2014 where you cut is what matters, and that is already solved upstream: each call carries exactly one section, so no fragment can cross a section boundary."
    },
    {
      "parameters": {
        "dataType": "json",
        "jsonMode": "expressionData",
        "jsonData": "={{ $json.body.texto }}",
        "textSplittingMode": "custom",
        "options": {
          "metadata": {
            "metadataValues": [
              {
                "name": "seccion",
                "value": "={{ $json.body.seccion }}"
              },
              {
                "name": "citation",
                "value": "={{ $json.body.citation }}"
              },
              {
                "name": "source",
                "value": "={{ $json.body.source }}"
              },
              {
                "name": "retrieved",
                "value": "={{ $json.body.retrieved }}"
              }
            ]
          }
        }
      },
      "id": "3f1c9a20-0000-4000-8000-000000000004",
      "name": "Leer Seccion",
      "type": "@n8n/n8n-nodes-langchain.documentDefaultDataLoader",
      "typeVersion": 1.1,
      "position": [
        700,
        520
      ],
      "notes": "This metadata list is FIXED for everything that enters in one call \u2014 and that is exactly why the design works. One call carries one section, so all its fragments legitimately share the same \u00a7. Trying to tag fragment-by-fragment inside a single large document is what this node cannot do."
    },
    {
      "parameters": {
        "mode": "insert",
        "tableName": "documentos_v2",
        "options": {}
      },
      "id": "3f1c9a20-0000-4000-8000-000000000005",
      "name": "Guardar en PGVector",
      "type": "@n8n/n8n-nodes-langchain.vectorStorePGVector",
      "typeVersion": 1.3,
      "position": [
        560,
        300
      ],
      "notes": ">>> REPLACE: attach a Postgres credential. The table is created if it does not exist. Note it is a NEW table rather than the original one \u2014 a corpus rebuilt from a different source does not belong in the same table as the old one, or the comparison between them stops being possible."
    },
    {
      "parameters": {
        "content": "## Load by API, one section per call\n\nEach POST carries ONE section with its \u00a7 already attached. That is why the loader's fixed metadata is correct here: every fragment from this call belongs to the same section.\n\nThe webhook requires the X-Carga-Token header. Without it, 403 and nothing runs.",
        "height": 220,
        "width": 560,
        "color": 4
      },
      "id": "3f1c9a20-0000-4000-8000-000000000006",
      "name": "Nota",
      "type": "n8n-nodes-base.stickyNote",
      "typeVersion": 1,
      "position": [
        200,
        40
      ]
    }
  ],
  "connections": {
    "Recibir Seccion": {
      "main": [
        [
          {
            "node": "Guardar en PGVector",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Embeddings Locales": {
      "ai_embedding": [
        [
          {
            "node": "Guardar en PGVector",
            "type": "ai_embedding",
            "index": 0
          }
        ]
      ]
    },
    "Leer Seccion": {
      "ai_document": [
        [
          {
            "node": "Guardar en PGVector",
            "type": "ai_document",
            "index": 0
          }
        ]
      ]
    },
    "Partir en Fragmentos": {
      "ai_textSplitter": [
        [
          {
            "node": "Leer Seccion",
            "type": "ai_textSplitter",
            "index": 0
          }
        ]
      ]
    }
  },
  "settings": {
    "executionOrder": "v1"
  }
}

About this workflow

RAG - Cargar Secciones (API). Uses embeddingsOllama, textSplitterRecursiveCharacterTextSplitter, documentDefaultDataLoader, vectorStorePGVector. Webhook trigger; 6 nodes.

Source: https://github.com/andresjmnz92-jpg/rag-privado/blob/main/workflows/cargar-secciones.json — original creator credit. Request a take-down →

More AI & RAG workflows → · Browse all categories →

Related workflows

Workflows that share integrations, category, or trigger type with this one. All free to copy and import.