Semantic Search Agent

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": "Semantic Search Agent",
  "nodes": [
    {
      "id": "webhook-trigger-id",
      "name": "Search Webhook",
      "type": "n8n-nodes-base.webhook",
      "typeVersion": 2,
      "position": [
        100,
        300
      ],
      "parameters": {
        "httpMethod": "POST",
        "path": "search",
        "responseMode": "responseNode",
        "options": {}
      }
    },
    {
      "id": "validate-input-id",
      "name": "Validate Input",
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [
        280,
        300
      ],
      "parameters": {
        "mode": "runOnceForEachItem",
        "jsCode": "const body = $json.body || $json;\nconst query = (body.query || '').trim();\nconst topK  = Math.min(Math.max(parseInt(body.top_k) || 5, 1), 20);\nconst tagFilter = body.tag_filter || null;\n\nif (!query) {\n  throw new Error('Missing required field: query');\n}\n\nreturn {\n  json: { query, top_k: topK, tag_filter: tagFilter }\n};\n"
      }
    },
    {
      "id": "embed-query-id",
      "name": "Embed Query",
      "type": "n8n-nodes-base.httpRequest",
      "typeVersion": 4,
      "position": [
        460,
        300
      ],
      "parameters": {
        "method": "POST",
        "url": "https://api.openai.com/v1/embeddings",
        "authentication": "genericCredentialType",
        "genericAuthType": "httpHeaderAuth",
        "sendHeaders": true,
        "headerParameters": {
          "parameters": [
            {
              "name": "Content-Type",
              "value": "application/json"
            },
            {
              "name": "Authorization",
              "value": "=Bearer {{ $vars.OPENAI_API_KEY }}"
            }
          ]
        },
        "sendBody": true,
        "contentType": "json",
        "body": {
          "model": "text-embedding-3-small",
          "input": "={{ $json.query }}"
        },
        "options": {}
      }
    },
    {
      "id": "merge-embedding-id",
      "name": "Merge Embedding",
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [
        640,
        300
      ],
      "parameters": {
        "mode": "runOnceForAllItems",
        "jsCode": "// Items: [0] = validate output, [1] = embed output\n// After chained nodes, $input.all() has the embedding result;\n// we reach back to Validate Input for the original params.\nconst embedding = $json.data?.[0]?.embedding;\nif (!embedding) throw new Error('Embedding API returned no vector');\n\nconst query     = $('Validate Input').first().json.query;\nconst topK      = $('Validate Input').first().json.top_k;\nconst tagFilter = $('Validate Input').first().json.tag_filter;\n\nreturn [{ json: { query, top_k: topK, tag_filter: tagFilter, embedding } }];\n"
      }
    },
    {
      "id": "qdrant-search-id",
      "name": "Qdrant Vector Search",
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [
        820,
        300
      ],
      "parameters": {
        "mode": "runOnceForEachItem",
        "jsCode": "const { embedding, top_k, tag_filter } = $json;\n\n// Build optional tag filter\nconst filter = tag_filter ? {\n  must: [{\n    key: 'metadata.tags',\n    match: { value: tag_filter }\n  }]\n} : undefined;\n\nconst body = {\n  vector: embedding,\n  limit: top_k,\n  with_payload: true,\n  score_threshold: 0.35,\n  ...(filter ? { filter } : {})\n};\n\nconst response = await fetch('http://qdrant:6333/collections/arxiv_papers/points/search', {\n  method: 'POST',\n  headers: {\n    'Content-Type': 'application/json',\n    'api-key': $vars.QDRANT_API_KEY || ''\n  },\n  body: JSON.stringify(body)\n});\n\nif (!response.ok) {\n  const err = await response.text();\n  throw new Error(`Qdrant search failed: ${response.status} ${err}`);\n}\n\nconst data = await response.json();\nconst results = (data.result || []).map(point => ({\n  score:         point.score,\n  title:         point.payload?.metadata?.title         || point.payload?.title         || 'Untitled',\n  url:           point.payload?.metadata?.url           || point.payload?.url           || '',\n  tags:          point.payload?.metadata?.tags          || point.payload?.tags          || [],\n  relevance:     point.payload?.metadata?.relevance     || point.payload?.relevance     || '',\n  what_it_does:  point.payload?.metadata?.what_it_does  || point.payload?.what_it_does  || '',\n  how_it_works:  point.payload?.metadata?.how_it_works  || point.payload?.how_it_works  || '',\n  why_it_matters:point.payload?.metadata?.why_it_matters|| point.payload?.why_it_matters|| '',\n  thumbnail_url: point.payload?.metadata?.thumbnail_url || point.payload?.thumbnail_url || '',\n  summary:       point.payload?.document || ''\n}));\n\nreturn [{ json: { ...$json, results } }];\n"
      }
    },
    {
      "id": "gpt-answer-id",
      "name": "GPT-4o Answer",
      "type": "n8n-nodes-base.httpRequest",
      "typeVersion": 4,
      "position": [
        1000,
        300
      ],
      "parameters": {
        "method": "POST",
        "url": "https://api.openai.com/v1/chat/completions",
        "authentication": "genericCredentialType",
        "genericAuthType": "httpHeaderAuth",
        "sendHeaders": true,
        "headerParameters": {
          "parameters": [
            {
              "name": "Content-Type",
              "value": "application/json"
            },
            {
              "name": "Authorization",
              "value": "=Bearer {{ $vars.OPENAI_API_KEY }}"
            }
          ]
        },
        "sendBody": true,
        "contentType": "json",
        "body": {
          "model": "gpt-4o",
          "response_format": {
            "type": "json_object"
          },
          "messages": [
            {
              "role": "system",
              "content": "You are ResearchFlow, an AI research assistant. Answer the user's question using ONLY the papers provided below. Cite every claim with the paper title in square brackets, e.g. [Paper Title]. If none of the papers are relevant to the question, say so clearly. Be concise and precise \u2014 this is for technical readers.\n\nRespond with a JSON object and nothing else:\n{\n  \"answer\": \"Your answer with inline [Paper Title] citations\",\n  \"sources\": [\n    {\n      \"title\": \"Paper title\",\n      \"url\": \"arXiv link\",\n      \"relevance_to_query\": \"One sentence on why this paper is relevant\"\n    }\n  ],\n  \"confidence\": \"High | Medium | Low \u2014 how well the retrieved papers answer the question\"\n}"
            },
            {
              "role": "user",
              "content": "=Question: {{ $json.query }}\n\nPapers:\n{{ $json.results.map((r, i) => `[${i+1}] ${r.title}\\nWhat it does: ${r.what_it_does}\\nHow it works: ${r.how_it_works}\\nWhy it matters: ${r.why_it_matters}\\nLink: ${r.url}`).join('\\n\\n') }}"
            }
          ]
        },
        "options": {}
      }
    },
    {
      "id": "format-response-id",
      "name": "Format Response",
      "type": "n8n-nodes-base.code",
      "typeVersion": 2,
      "position": [
        1180,
        300
      ],
      "parameters": {
        "mode": "runOnceForEachItem",
        "jsCode": "const raw = $json.choices?.[0]?.message?.content || '{}';\n\nlet parsed;\ntry {\n  parsed = JSON.parse(raw);\n} catch (e) {\n  throw new Error(`GPT-4o returned non-JSON: ${raw.slice(0, 200)}`);\n}\n\nconst results = $('Qdrant Vector Search').first().json.results || [];\nconst query   = $('Validate Input').first().json.query;\n\nreturn {\n  json: {\n    query,\n    answer:     parsed.answer     || 'No answer generated.',\n    sources:    parsed.sources    || [],\n    confidence: parsed.confidence || 'Unknown',\n    result_count: results.length,\n    retrieved_papers: results.map(r => ({\n      title:         r.title,\n      url:           r.url,\n      score:         r.score,\n      tags:          r.tags,\n      thumbnail_url: r.thumbnail_url\n    }))\n  }\n};\n"
      }
    },
    {
      "id": "respond-webhook-id",
      "name": "Respond to Webhook",
      "type": "n8n-nodes-base.respondToWebhook",
      "typeVersion": 1,
      "position": [
        1360,
        300
      ],
      "parameters": {
        "respondWith": "json",
        "responseBody": "={{ JSON.stringify($json) }}",
        "options": {
          "responseHeaders": {
            "entries": [
              {
                "name": "Content-Type",
                "value": "application/json"
              }
            ]
          }
        }
      }
    },
    {
      "id": "search-error-id",
      "name": "Search Error Response",
      "type": "n8n-nodes-base.respondToWebhook",
      "typeVersion": 1,
      "position": [
        820,
        500
      ],
      "parameters": {
        "respondWith": "json",
        "responseCode": 500,
        "responseBody": "={{ JSON.stringify({ error: $json.message || 'Internal error', query: $('Validate Input').first()?.json?.query || null }) }}",
        "options": {
          "responseHeaders": {
            "entries": [
              {
                "name": "Content-Type",
                "value": "application/json"
              }
            ]
          }
        }
      }
    }
  ],
  "connections": {
    "Search Webhook": {
      "main": [
        [
          {
            "node": "Validate Input",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Validate Input": {
      "main": [
        [
          {
            "node": "Embed Query",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Embed Query": {
      "main": [
        [
          {
            "node": "Merge Embedding",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Merge Embedding": {
      "main": [
        [
          {
            "node": "Qdrant Vector Search",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Qdrant Vector Search": {
      "main": [
        [
          {
            "node": "GPT-4o Answer",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "Search Error Response",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "GPT-4o Answer": {
      "main": [
        [
          {
            "node": "Format Response",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "Search Error Response",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Format Response": {
      "main": [
        [
          {
            "node": "Respond to Webhook",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  },
  "active": true,
  "settings": {
    "executionOrder": "v1",
    "saveManualExecutions": true,
    "callerPolicy": "workflowsFromSameOwner",
    "errorWorkflow": ""
  },
  "tags": [
    {
      "name": "search"
    },
    {
      "name": "rag"
    }
  ]
}

About this workflow

Semantic Search Agent. Uses httpRequest. Webhook trigger; 9 nodes.

Source: https://github.com/keila-moral/researchflow-ai/blob/main/workflows/Semantic_Search_Agent.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.