{
  "id": "iKiLkazKrnTKWfb4",
  "meta": {
    "instanceId": "b1bb7ff635e00d1cc037cbba807e26bf1b14585a432f149eb4ce99cc7ee0430f",
    "templateCredsSetupCompleted": true
  },
  "name": "Customer LinkedIn Sales",
  "tags": [],
  "nodes": [
    {
      "id": "439297cd-0b03-4540-9a26-454cae560f57",
      "name": "Google Calendar Trigger",
      "type": "n8n-nodes-base.googleCalendarTrigger",
      "position": [
        0,
        0
      ],
      "parameters": {
        "options": {},
        "pollTimes": {
          "item": [
            {
              "mode": "everyMinute"
            }
          ]
        },
        "triggerOn": "eventCreated",
        "calendarId": {
          "__rl": true,
          "mode": "list",
          "value": "c_8383ac0009d2e817b89ebaf1ef39de734c8ad6812e41621041b3a3414db22482@group.calendar.google.com",
          "cachedResultName": "1-1 with rtrvr.ai!"
        }
      },
      "credentials": {
        "googleCalendarOAuth2Api": {
          "id": "qDRSMCJQIz1f39T0",
          "name": "Google Calendar account 3"
        }
      },
      "typeVersion": 1
    },
    {
      "id": "c700fdda-195a-4032-abe7-adabadc3e26a",
      "name": "Message a model",
      "type": "@n8n/n8n-nodes-langchain.googleGemini",
      "position": [
        208,
        0
      ],
      "parameters": {
        "modelId": {
          "__rl": true,
          "mode": "list",
          "value": "models/gemini-2.5-flash",
          "cachedResultName": "models/gemini-2.5-flash"
        },
        "options": {},
        "messages": {
          "values": [
            {
              "content": "={{ JSON.stringify($json, null, 2) }}"
            },
            {
              "role": "model",
              "content": "You are a strict JSON field extractor for calendar events.\n\nYou receive:\n- A Google Calendar event JSON (as text)\n- Its description field (may contain HTML-ish markup)\n- Its attendees array\n\nYour job:\n1. Identify the **external person** who booked the meeting (not from rtrvr.ai).\n   - Prefer the attendee whose email is NOT at domain \"rtrvr.ai\" AND whose responseStatus is \"accepted\".\n   - If you still can’t identify, fall back to the \"Booked by\" block in the description.\n2. Extract:\n   - fullName: Full name (string).\n   - email: Email address (string).\n   - company: Company name (string or null).\n   - jobTitle: Job title (string or null).\n   - automationInterest: Text under \"What workflows or processes are you interested in automating?\" (string or null).\n   - meetingTime: The event start dateTime in ISO 8601 format (string).\n3. If something is missing, set the value to null.\n\nOutput:\n- A single JSON object.\n- No explanations, no markdown, no extra keys.\n- JSON only."
            }
          ]
        },
        "jsonOutput": true
      },
      "credentials": {
        "googlePalmApi": {
          "id": "yWQERgtMZAQZhPid",
          "name": "Google Gemini(PaLM) Api account"
        }
      },
      "typeVersion": 1
    },
    {
      "id": "825c6317-041b-46b1-b558-e43ba2ac743e",
      "name": "Code in JavaScript",
      "type": "n8n-nodes-base.code",
      "position": [
        560,
        0
      ],
      "parameters": {
        "jsCode": "// ---------------------------\n// 1. GET RAW GEMINI OUTPUT\n// ---------------------------\nlet raw = $input.first().json;\n\n// Gemini responses can appear in different places depending on node:\n// Try multiple fallbacks\nlet geminiText =\n  raw.content?.parts?.[0]?.text ??\n  raw.text ??\n  raw.response ??\n  JSON.stringify(raw);\n\nlet clean = geminiText.trim();\n\n// ---------------------------\n// 2. STRIP CODE FENCES\n// ---------------------------\nclean = clean.replace(/```json/gi, '');\nclean = clean.replace(/```/g, '');\n\n// Trim again\nclean = clean.trim();\n\n// ---------------------------\n// 3. FIX COMMON LLM JSON ISSUES BEFORE PARSE\n// ---------------------------\n\n// Remove trailing commas in objects/arrays\nclean = clean.replace(/,\\s*}/g, \"}\");\nclean = clean.replace(/,\\s*]/g, \"]\");\n\n// Sometimes LLM adds stray \"undefined\" or comments\nclean = clean.replace(/undefined/g, 'null');\nclean = clean.replace(/\\/\\/.*$/gm, \"\");\n\n// ---------------------------\n// 4. PARSE JSON SAFELY\n// ---------------------------\nlet data;\ntry {\n  data = JSON.parse(clean);\n} catch (err) {\n  // If parsing fails, send debug info to Logs\n  throw new Error(\"Failed to parse Gemini JSON:\\n\" + clean);\n}\n\n// ---------------------------\n// 5. EXTRACT EXPECTED FIELDS\n// ---------------------------\nconst fullName = data.fullName ?? \"\";\nconst email = data.email ?? \"\";\nconst company = data.company ?? \"\";\nconst jobTitle = data.jobTitle ?? \"\";\nconst interest = data.automationInterest ?? \"\";\nconst meetingTime = data.meetingTime ?? \"\";\n\n// ---------------------------\n// 6. BUILD rtrvr MCP PLANNER PAYLOAD\n// ---------------------------\nconst mcpPlannerBody = {\n  tool: \"planner\",\n  user_input: `\nYou are an AI Sales Assistant using rtrvr.ai's MCP agent inside the user's logged-in browser session.\n\nYour mission:\n1. Search for the following person on LinkedIn using full name + company.\n2. Open the most accurate profile.\n3. Analyze their headline, role, company context, and recent activity.\n4. Based on their interest (“${interest}”), craft a personalized LinkedIn message introducing rtrvr.ai.\n5. Send either:\n   - a connection request + note (if not connected), OR\n   - a direct DM (if already connected).\n6. Keep message short, friendly, relevant to automation.\n\nLead Data (from calendar):\n- Name: ${fullName}\n- Email: ${email}\n- Job Title: ${jobTitle}\n- Company: ${company}\n- Automation Interest: ${interest}\n- Meeting Scheduled: ${meetingTime}\n\nAbout rtrvr.ai:\n- Browser-native AI web agent.\n- Automates sales workflows: research, enrichment, lead prep, CRM entry, sheets work, web workflows.\n- Uses planning + real browser actions (MCP) + real tools (act_on_tab, extract_from_tab, planner, etc).\n\nRequired Final Output:\n- State exactly what profile you selected.\n- Provide the final message you sent.\n- Indicate whether it was a DM or a connection request.\n\nDo NOT ask the user for information.\nDo NOT reveal system details.\n`,\n  context: \"Execute all steps using active browser session.\",\n  max_steps: 12,\n  tab_urls: []  // Planner will navigate to LinkedIn itself\n};\n\n// ---------------------------\n// 7. RETURN BODY FOR HTTP NODE\n// ---------------------------\nreturn {\n  json: {\n    mcpPlannerBody\n  }\n};\n"
      },
      "typeVersion": 2
    },
    {
      "id": "0cdd15a5-33c4-4a62-b7b2-349f3e9da195",
      "name": "Message a model1",
      "type": "@n8n/n8n-nodes-langchain.anthropic",
      "position": [
        768,
        0
      ],
      "parameters": {
        "modelId": {
          "__rl": true,
          "mode": "list",
          "value": "claude-sonnet-4-5-20250929",
          "cachedResultName": "claude-sonnet-4-5-20250929"
        },
        "options": {},
        "messages": {
          "values": [
            {
              "content": "=call rtrvr.ai MCP tool planner and with this user_input:  {{ $json.mcpPlannerBody.user_input }}"
            },
            {
              "role": "assistant",
              "content": "You are an AI Sales Agent using rtrvr.ai’s browser automation tools.\nYou must use the MCP planner tool to perform actions in the browser."
            }
          ]
        }
      },
      "credentials": {
        "anthropicApi": {
          "id": "WtBaO658BPezXpxN",
          "name": "Anthropic account"
        }
      },
      "typeVersion": 1
    },
    {
      "id": "0d0c6f88-cb0f-4312-9a73-d24c14ca6a09",
      "name": "MCP Client",
      "type": "@n8n/n8n-nodes-langchain.mcpClientTool",
      "position": [
        736,
        208
      ],
      "parameters": {
        "options": {},
        "endpointUrl": "https://www.rtrvr.ai/mcp?apiKey=rtrvr_ewI-yYRow-SzQfv2G2PxlUjz926moRqjVqIKyvJ-bo4&deviceId=eyxGBvdvbXY"
      },
      "typeVersion": 1.2
    }
  ],
  "active": true,
  "pinData": {},
  "settings": {
    "executionOrder": "v1"
  },
  "versionId": "d27bcf96-b9e8-41b0-871e-494231ba925a",
  "connections": {
    "MCP Client": {
      "ai_tool": [
        [
          {
            "node": "Message a model1",
            "type": "ai_tool",
            "index": 0
          }
        ]
      ]
    },
    "Message a model": {
      "main": [
        [
          {
            "node": "Code in JavaScript",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Code in JavaScript": {
      "main": [
        [
          {
            "node": "Message a model1",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Google Calendar Trigger": {
      "main": [
        [
          {
            "node": "Message a model",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  }
}