Dynamischer KI-Netzwerkforscher: Von reinem Text zu benutzerdefiniertem CSV

• •Möglicherweise sind Ziel-API-Anmeldedaten erforderlich
• •OpenAI API Key

{
  "nodes": [
    {
      "id": "7df78b4d-8c66-4353-a87e-7b151913f856",
      "name": "Bei Formularabgabe",
      "type": "n8n-nodes-base.formTrigger",
      "position": [
        0,
        16
      ],
      "webhookId": "0d4374b6-84a6-4a71-834b-40dd3d3e3adf",
      "parameters": {
        "options": {},
        "formTitle": "New research",
        "formFields": {
          "values": [
            {
              "fieldType": "textarea",
              "fieldLabel": "Describe your research",
              "requiredField": true
            }
          ]
        }
      },
      "typeVersion": 2.3
    },
    {
      "id": "3c8ad682-a662-4266-bdb5-732f6bb8614c",
      "name": "OpenAI Chat Model",
      "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
      "position": [
        304,
        240
      ],
      "parameters": {
        "model": {
          "__rl": true,
          "mode": "list",
          "value": "gpt-5-chat-latest",
          "cachedResultName": "gpt-5-chat-latest"
        },
        "options": {
          "responseFormat": "json_object"
        }
      },
      "credentials": {
        "openAiApi": {
          "id": "dMiSy27YCK6c6rra",
          "name": "Duv's OpenAI"
        }
      },
      "typeVersion": 1.2
    },
    {
      "id": "074eea9f-b07e-4677-8b58-75b92155df44",
      "name": "Aufteilen",
      "type": "n8n-nodes-base.splitOut",
      "position": [
        1008,
        16
      ],
      "parameters": {
        "options": {},
        "fieldToSplitOut": "List"
      },
      "typeVersion": 1
    },
    {
      "id": "e9d5daa7-aa07-4873-888f-b704cf0d6d7c",
      "name": "Über Elemente iterieren",
      "type": "n8n-nodes-base.splitInBatches",
      "position": [
        1344,
        16
      ],
      "parameters": {
        "options": {}
      },
      "typeVersion": 3
    },
    {
      "id": "baf2d169-9593-4ec2-9415-1ed45cb303d0",
      "name": "Objektname und Wert abrufen",
      "type": "n8n-nodes-base.set",
      "position": [
        1584,
        256
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "3de0459c-a3bf-4db0-94cc-ce007fa5db55",
              "name": "={{ $('Prepare prompts and schema').item.json.ObjectName }}",
              "type": "string",
              "value": "={{ $json.PropertyValue }}"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "84f376a1-3ab5-4bab-b962-357e4f57854c",
      "name": "Kurznotiz",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        640,
        -176
      ],
      "parameters": {
        "color": 5,
        "width": 288,
        "height": 384,
        "content": "## AI web-search to find all items that match the conditions\n\nDon't forget to connect your Linkup credentials."
      },
      "typeVersion": 1
    },
    {
      "id": "53b2a08a-f61f-4c60-8c25-9edc8731a9d4",
      "name": "Linkup abfragen, um Liste zu finden",
      "type": "n8n-nodes-base.httpRequest",
      "position": [
        736,
        16
      ],
      "parameters": {
        "url": "https://api.linkup.so/v1/search",
        "method": "POST",
        "options": {},
        "sendBody": true,
        "authentication": "genericCredentialType",
        "bodyParameters": {
          "parameters": [
            {
              "name": "q",
              "value": "={{ $json.discoveryQuery }}"
            },
            {
              "name": "depth",
              "value": "deep"
            },
            {
              "name": "outputType",
              "value": "structured"
            },
            {
              "name": "structuredOutputSchema",
              "value": "={{ JSON.stringify($json.discoverySchema) }}"
            },
            {
              "name": "includeImages",
              "value": "false"
            }
          ]
        },
        "genericAuthType": "httpBearerAuth"
      },
      "credentials": {
        "httpBearerAuth": {
          "id": "W7AgeoVOv60DlvyS",
          "name": "Linkup - web search AI"
        }
      },
      "typeVersion": 4.2
    },
    {
      "id": "74957f1e-01ba-4cfd-a5e7-2fec11e0a7ad",
      "name": "Finale JSON für dieses Element vorbereiten",
      "type": "n8n-nodes-base.code",
      "position": [
        2128,
        448
      ],
      "parameters": {
        "jsCode": "// Get the first key-value from the \"Get object name and value\" node\n// Replace 'Get_object_name_and_value' with the actual name of your node\nconst firstNodeData = $node[\"Get object name and value\"].json;\n\n// Get the previous node data (the one with multiple keys)\nconst previousNodeData = items[0].json;\n\n// Create the output object\nconst output = {};\n\n// Add the first key-value pair from the first node\n// Assuming the node only has one key-value pair\nconst firstKey = Object.keys(firstNodeData)[0];\noutput[firstKey] = String(firstNodeData[firstKey]);\n\n// Add all key-value pairs from the previous node, stringified\nfor (const [key, value] of Object.entries(previousNodeData)) {\n    if (Array.isArray(value)) {\n        output[key] = value.join(', ');\n    } else if (typeof value === 'object' && value !== null) {\n        output[key] = JSON.stringify(value);\n    } else {\n        output[key] = String(value);\n    }\n}\n\n// Return the new item\nreturn [{ json: output }];\n"
      },
      "typeVersion": 2
    },
    {
      "id": "d170bc2a-68a3-497c-803b-78f904ec9351",
      "name": "Linkup abfragen, um alle Eigenschaften zu finden",
      "type": "n8n-nodes-base.httpRequest",
      "position": [
        1840,
        256
      ],
      "parameters": {
        "url": "https://api.linkup.so/v1/search",
        "method": "POST",
        "options": {},
        "sendBody": true,
        "authentication": "genericCredentialType",
        "bodyParameters": {
          "parameters": [
            {
              "name": "q",
              "value": "=For item: {{ $('Loop Over Items').item.json.PropertyValue }}\n{{ $('Prepare prompts and schema').item.json.enrichmentQuery }}"
            },
            {
              "name": "depth",
              "value": "standard"
            },
            {
              "name": "outputType",
              "value": "structured"
            },
            {
              "name": "structuredOutputSchema",
              "value": "={{ JSON.stringify($('Prepare prompts and schema').item.json.enrichmentSchema) }}"
            },
            {
              "name": "includeImages",
              "value": "false"
            }
          ]
        },
        "genericAuthType": "httpBearerAuth"
      },
      "credentials": {
        "httpBearerAuth": {
          "id": "W7AgeoVOv60DlvyS",
          "name": "Linkup - web search AI"
        }
      },
      "typeVersion": 4.2
    },
    {
      "id": "c4b50e76-3ff0-4d07-830e-7141a6d90d4d",
      "name": "Prompts und Schema vorbereiten",
      "type": "@n8n/n8n-nodes-langchain.chainLlm",
      "position": [
        256,
        16
      ],
      "parameters": {
        "text": "={{ $json['Describe your research'] }}",
        "batching": {},
        "messages": {
          "messageValues": [
            {
              "message": "=# Role\n\nYou are an AI research strategy generator for an automated research workflow. Your first task is to identify the primary **object** of the user's research (e.g., Company, Person, City, Product).\n\nYou will always return a valid JSON object with **five** keys, starting with `ObjectName`:\n\n{\n  \"ObjectName\": \"string\",\n  \"discoveryQuery\": \"string\",\n  \"discoverySchema\": { ... },\n  \"enrichmentQuery\": \"string\",\n  \"enrichmentSchema\": { ... }\n}\n\n\n## Step 0: Identify the ObjectName\n\n**Goal**: First, identify the core subject of the user's request and create a simple label for it.\n\n  * Based on the user's prompt, determine the primary object being researched.\n\n  * The **ObjectName** must be a simple, **singular noun** (e.g., \"Company\", not \"Companies\").\n\n  * This name should be a general category, not overly specific.\n\n  * **Examples**:\n\n      * If the request is `\"List 50 German fashion companies...\"`, the object is a company. So, **ObjectName** should be `\"Company\"`.\n      * If the request is `\"Find 25 CEOs of technology companies...\"`, the object is a person. So, **ObjectName** should be `\"Person\"`.\n      * If the request is `\"Provide 25 HEX color codes...\"`, the object is a color. So, **ObjectName** should be `\"Color\"`.\n\n\n## Step 1: Object Discovery\n\n**Goal**: Define **what to search for** and how to structure the list of results into a predictable array.\n\n  * **`discoveryQuery`**:\n\n      * This is a single sentence prompt that describes EXACTLY what list of items needs to be found.\n      * It must be explicit, concise, and human-readable.\n      * It must include a mention \"if not enough relevant items have been found, don't force adding the requested number, quality prevails\"\n\n  * **`discoverySchema`**:\n\n      * A JSON Schema that forces the output into a specific, non-changing structure.\n\n      * The structure is **fixed** and must be used exactly as shown below.\n\n      * **CRUCIAL**: You must set the `\"description\"` for `\"PropertyValue\"` to tell the AI what content to find. This description should correspond to the unique identifier of the `ObjectName` you identified in Step 0 (e.g., 'The official name of the company').\n\n      * **Fixed Structure Example**:\n\n        {\n          \"type\": \"object\",\n          \"properties\": {\n            \"List\": {\n              \"type\": \"array\",\n              \"items\": {\n                \"type\": \"object\",\n                \"properties\": {\n                  \"PropertyValue\": {\n                    \"type\": \"string\",\n                    \"description\": \"A clear description of what this value should be. E.g., 'The official name of the company'.\"\n                  }\n                },\n                \"required\": [\"PropertyValue\"]\n              }\n            }\n          },\n          \"required\": [\"List\"]\n        }\n\n\n### Step 2: Object Enrichment\n\n**Goal**: For each item discovered in Step 1, define **what extra information to research**. If not specified by the user, aim for 5-8 properties that make the most sense. Mention them in the enrichmentQuery, and give them easily readable names in the enrichmentSchema (e.g., prefer \"Company Name\" over \"companyName\").\n\n  * **`enrichmentQuery`**:\n\n      * A single sentence prompt for Linkup. Refer to the object generically based on your identified `ObjectName` (e.g., \"For that company...\", \"For that person...\"). The actual name of the item will be provided contextually, so do not use a placeholder.\n\n  * **`enrichmentSchema`**:\n\n      * A detailed JSON Schema object that defines all the fields to capture for each object, based on the user's goal.\n      * The root must be `\"type\": \"object\"`.\n      * All properties must be required (see example later)\n\n\n\n## Output Format\n\n  * Return a single, valid **JSON object** with five keys: `\"ObjectName\"`, `\"discoveryQuery\"`, `\"discoverySchema\"`, `\"enrichmentQuery\"`, and `\"enrichmentSchema\"`.\n  * Do not stringify the schemas; they must remain as JSON objects.\n  * Do not include explanations or any text outside the final JSON.\n\n\n### Example JSON Output\n\nHere is a complete example for the request: \"I need 50 german companies in fashion industry between 40 and 130 employees...\"\n\n{\n  \"ObjectName\": \"Company\",\n  \"discoveryQuery\": \"List 50 German fashion industry companies with 40 to 130 employees. If not enough relevant companies have been found, don't force adding the requested number, quality prevails\",\n  \"discoverySchema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"List\": {\n        \"type\": \"array\",\n        \"items\": {\n          \"type\": \"object\",\n          \"properties\": {\n            \"PropertyValue\": {\n              \"type\": \"string\",\n              \"description\": \"The official name of the German fashion company.\"\n            }\n          },\n          \"required\": [\"PropertyValue\"]\n        }\n      }\n    },\n    \"required\": [\"List\"]\n  },\n  \"enrichmentQuery\": \"For that company, provide detailed business information useful for personalized outreach including its Website, employee count, headquarters, key contacts, and recent news.\",\n  \"enrichmentSchema\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"Company Name\": {\n        \"type\": \"string\",\n        \"description\": \"The official company name\"\n      },\n      \"Website\": {\n        \"type\": \"string\",\n        \"description\": \"Official website URL\"\n      },\n      \"Employee Count\": {\n        \"type\": \"integer\",\n        \"description\": \"Number of employees\"\n      },\n      \"Headquarters\": {\n        \"type\": \"string\",\n        \"description\": \"Headquarters location\"\n      },\n      \"Key Contacts\": {\n        \"type\": \"array\",\n        \"items\": {\n          \"type\": \"string\"\n        },\n        \"description\": \"List of key contact persons (e.g. CEO, Marketing Head)\"\n      },\n      \"Recent News\": {\n        \"type\": \"string\",\n        \"description\": \"Any recent news or updates about the company for outreach personalization\"\n      }\n    },\n    \"required\": [\n  \"Company Name\",\n  \"Website\",\n  \"Employee Count\",\n  \"Headquarters\",\n  \"Key Contacts\",\n  \"Recent News\"\n    ]\n  }\n}"
            }
          ]
        },
        "promptType": "define"
      },
      "typeVersion": 1.7
    },
    {
      "id": "ef93d5c6-48bc-4afb-80e9-5511c4143d7f",
      "name": "In CSV konvertieren",
      "type": "n8n-nodes-base.convertToFile",
      "position": [
        1664,
        -400
      ],
      "parameters": {
        "options": {}
      },
      "typeVersion": 1.1
    },
    {
      "id": "fa8be6ed-db4f-43a6-b5cf-b33ff400f266",
      "name": "Kurznotiz2",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        -512,
        -416
      ],
      "parameters": {
        "width": 448,
        "height": 752,
        "content": "# **Dynamic AI Web Researcher**\n\nThis workflow turns any plain-text research request into a custom-built, structured spreadsheet (CSV). It uses a \"thinker\" AI to plan the research and a \"doer\" AI to execute it.\n\n## **How it works**\n1.  **Plan:** An AI \"thinker\" analyzes your request and dynamically creates a custom plan, deciding on the perfect spreadsheet columns for your goal.\n2.  **Discover:** An AI \"researcher\" (Linkup) performs a deep web search to find an initial list of items based on the plan.\n3.  **Enrich:** It then loops through each item, performing fast web searches to fill in all the detailed columns defined by the planner.\n4.  **Output:** The final, structured data is converted into a CSV file.\n\n## **How to use**\n1.  **Connect your AI provider** to the **OpenAI Chat Model** node.\n2.  **Connect Linkup:** Add your **Linkup API Key** to the two HTTP Request nodes that query Linkup.\n3.  **Run:** Activate the workflow and use the form to submit a research request.\n\n\n\n\n*This template was created by Guillaume Duvernay*"
      },
      "typeVersion": 1
    },
    {
      "id": "04ecb528-4934-4565-8677-e69917a56f58",
      "name": "Kurznotiz3",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        192,
        -176
      ],
      "parameters": {
        "color": 6,
        "width": 352,
        "height": 576,
        "content": "## The architect brain\n\nThis AI step defines how the output of the research will look like (CSV schema), and how to get there (prompts).\n\nMake sure to connect a powerful AI model"
      },
      "typeVersion": 1
    },
    {
      "id": "48f2305c-5ef5-41c0-8705-37a8582502e5",
      "name": "Kurznotiz1",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        1760,
        64
      ],
      "parameters": {
        "color": 7,
        "width": 288,
        "height": 384,
        "content": "## AI web-search to find all information about one item\n\nDon't forget to connect your Linkup credentials."
      },
      "typeVersion": 1
    },
    {
      "id": "29343905-2b7c-4d34-8455-066039945a6a",
      "name": "Kurznotiz4",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        1216,
        -112
      ],
      "parameters": {
        "color": 5,
        "width": 1088,
        "height": 816,
        "content": "## Running a loop on each item\n\nWithin this loop, each item from the list will go through an AI web search that will find all the values for the properties that the initial AI step had identified. These values will be cleaned and prepared before the outcome gets converted into a CSV."
      },
      "typeVersion": 1
    },
    {
      "id": "2cc1bca9-29ba-4d79-9e3f-6217c39d495a",
      "name": "Kurznotiz5",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        1520,
        -592
      ],
      "parameters": {
        "color": 4,
        "width": 384,
        "height": 352,
        "content": "## The output: A custom CSV\n\nThis CSV contains one row per item that was found, and each column is a property that got enriched thanks to the Web AI search. "
      },
      "typeVersion": 1
    }
  ],
  "connections": {
    "074eea9f-b07e-4677-8b58-75b92155df44": {
      "main": [
        [
          {
            "node": "e9d5daa7-aa07-4873-888f-b704cf0d6d7c",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "e9d5daa7-aa07-4873-888f-b704cf0d6d7c": {
      "main": [
        [
          {
            "node": "ef93d5c6-48bc-4afb-80e9-5511c4143d7f",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "baf2d169-9593-4ec2-9415-1ed45cb303d0",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "3c8ad682-a662-4266-bdb5-732f6bb8614c": {
      "ai_languageModel": [
        [
          {
            "node": "c4b50e76-3ff0-4d07-830e-7141a6d90d4d",
            "type": "ai_languageModel",
            "index": 0
          }
        ]
      ]
    },
    "7df78b4d-8c66-4353-a87e-7b151913f856": {
      "main": [
        [
          {
            "node": "c4b50e76-3ff0-4d07-830e-7141a6d90d4d",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "baf2d169-9593-4ec2-9415-1ed45cb303d0": {
      "main": [
        [
          {
            "node": "d170bc2a-68a3-497c-803b-78f904ec9351",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "c4b50e76-3ff0-4d07-830e-7141a6d90d4d": {
      "main": [
        [
          {
            "node": "53b2a08a-f61f-4c60-8c25-9edc8731a9d4",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "53b2a08a-f61f-4c60-8c25-9edc8731a9d4": {
      "main": [
        [
          {
            "node": "074eea9f-b07e-4677-8b58-75b92155df44",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "74957f1e-01ba-4cfd-a5e7-2fec11e0a7ad": {
      "main": [
        [
          {
            "node": "e9d5daa7-aa07-4873-888f-b704cf0d6d7c",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "d170bc2a-68a3-497c-803b-78f904ec9351": {
      "main": [
        [
          {
            "node": "74957f1e-01ba-4cfd-a5e7-2fec11e0a7ad",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  }
}