{ "name": "Email Classification Filter v1", "nodes": [ { "parameters": { "pollTimes": { "item": [ { "mode": "everyMinute", "minute": 5 } ] }, "filters": { "readStatus": "unread", "receivedAfter": "" }, "options": { "simple": false } }, "id": "gmail-trigger", "name": "Watch for Emails", "type": "n8n-nodes-base.gmailTrigger", "typeVersion": 1, "position": [ 240, 300 ], "credentials": { "gmailOAuth2": { "name": "" } }, "notes": "CRITICAL SETTINGS:\n1. Simplify = OFF (needed for threadId and labelIds)\n2. Poll every 5 minutes (adjust based on responsiveness needs)\n3. Read Status = Unread only\n\nThis triggers on new unread emails and passes full metadata." }, { "parameters": { "mode": "runOnceForAllItems", "jsCode": "// Email Pre-Filter v1.0\n// 4-layer filtering: Domain, Headers, Patterns, Gmail Categories\n// Catches 60-80% of non-customer emails BEFORE AI classification\n\nconst items = $input.all();\nconst results = [];\n\n// ============ CONFIGURATION ============\n// EDIT THIS: Add your company's email domains\nconst COMPANY_DOMAINS = [\n 'yourcompany.com',\n 'yoursupport.com'\n // Add more domains that your workflow sends from\n];\n\n// Known auto-reply sender patterns\nconst AUTO_REPLY_SENDERS = [\n /noreply@/i,\n /no-reply@/i,\n /notifications?@/i,\n /alerts?@/i,\n /mailer-daemon@/i,\n /postmaster@/i,\n /donotreply@/i\n];\n\n// Auto-reply subject patterns\nconst AUTO_REPLY_SUBJECTS = [\n /out of office/i,\n /automatic reply/i,\n /auto-reply/i,\n /auto reply/i,\n /vacation.*reply/i,\n /away from (my )?office/i,\n /delivery status notification/i,\n /undeliverable/i,\n /mail delivery failed/i\n];\n\n// Gmail categories to skip (these are FREE to filter)\nconst SKIP_GMAIL_CATEGORIES = [\n 'CATEGORY_PROMOTIONS',\n 'CATEGORY_SOCIAL',\n 'SPAM'\n];\n// =========================================\n\nfor (const item of items) {\n const email = item.json;\n \n // Extract email fields safely\n const from = (email.from || email.From || '').toLowerCase();\n const subject = email.subject || email.Subject || '';\n const labelIds = email.labelIds || [];\n \n // Parse headers (Gmail returns as array or object)\n let headers = {};\n if (email.payload && email.payload.headers) {\n // Full Gmail format\n email.payload.headers.forEach(h => {\n headers[h.name.toLowerCase()] = h.value;\n });\n } else if (email.headers) {\n // Simplified format\n headers = email.headers;\n }\n \n // ============ LAYER 1: OWN DOMAIN (Loop Prevention) ============\n // CRITICAL: Prevents infinite reply loops\n for (const domain of COMPANY_DOMAINS) {\n if (from.includes('@' + domain.toLowerCase())) {\n results.push({\n json: {\n skip: true,\n reason: 'own_domain',\n classification_override: 'internal',\n confidence: 1.0,\n message: `Email from own domain (${domain}) - loop prevention`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n continue;\n }\n }\n if (results.length > 0 && results[results.length - 1].json.skip) continue;\n \n // ============ LAYER 2: AUTO-REPLY HEADERS ============\n const autoReplyHeaders = [\n 'x-auto-response-suppress',\n 'auto-submitted',\n 'x-autoreply',\n 'x-autorespond'\n ];\n \n for (const headerName of autoReplyHeaders) {\n const headerValue = headers[headerName];\n if (headerValue && headerValue.toLowerCase() !== 'no') {\n results.push({\n json: {\n skip: true,\n reason: 'auto_reply_header',\n classification_override: 'automated',\n confidence: 1.0,\n message: `Auto-reply header detected: ${headerName}`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n break;\n }\n }\n if (results.length > 0 && results[results.length - 1].json.skip) continue;\n \n // Check Precedence header (bulk, junk, list)\n const precedence = headers['precedence'];\n if (precedence && /bulk|junk|list/i.test(precedence)) {\n results.push({\n json: {\n skip: true,\n reason: 'bulk_mail',\n classification_override: 'automated',\n confidence: 0.95,\n message: `Bulk mail header: Precedence: ${precedence}`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n continue;\n }\n \n // ============ LAYER 3: PATTERN MATCHING ============\n // Check sender patterns\n for (const pattern of AUTO_REPLY_SENDERS) {\n if (pattern.test(from)) {\n results.push({\n json: {\n skip: true,\n reason: 'auto_reply_sender',\n classification_override: 'automated',\n confidence: 0.95,\n message: `Known auto-reply sender pattern: ${from}`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n break;\n }\n }\n if (results.length > 0 && results[results.length - 1].json.skip) continue;\n \n // Check subject patterns\n for (const pattern of AUTO_REPLY_SUBJECTS) {\n if (pattern.test(subject)) {\n results.push({\n json: {\n skip: true,\n reason: 'auto_reply_subject',\n classification_override: 'automated',\n confidence: 0.90,\n message: `Subject matches auto-reply pattern`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n break;\n }\n }\n if (results.length > 0 && results[results.length - 1].json.skip) continue;\n \n // ============ LAYER 4: GMAIL CATEGORIES ============\n for (const category of SKIP_GMAIL_CATEGORIES) {\n if (labelIds.includes(category)) {\n results.push({\n json: {\n skip: true,\n reason: 'gmail_category',\n classification_override: category === 'SPAM' ? 'spam' : 'automated',\n confidence: 0.90,\n message: `Gmail category: ${category}`,\n email_id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject\n }\n });\n break;\n }\n }\n if (results.length > 0 && results[results.length - 1].json.skip) continue;\n \n // ============ PASSED ALL FILTERS ============\n // Email continues to AI classification\n results.push({\n json: {\n skip: false,\n reason: null,\n email_data: {\n id: email.id || email.messageId,\n thread_id: email.threadId,\n from: from,\n subject: subject,\n snippet: (email.snippet || '').substring(0, 500),\n body: email.textPlain || email.body || email.snippet || '',\n labelIds: labelIds,\n date: email.internalDate || email.date\n }\n }\n });\n}\n\nreturn results;" }, "id": "code-prefilter", "name": "Free Filter (Pre-Check)", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ 460, 300 ], "notes": "4-layer pre-filter catches 60-80% of emails for FREE:\n\n1. Own Domain - Prevents infinite reply loops (CRITICAL)\n2. Auto-Reply Headers - X-Auto-Response-Suppress, etc.\n3. Pattern Matching - noreply@, Out of Office, etc.\n4. Gmail Categories - Promotions, Social, Spam\n\nEDIT THE COMPANY_DOMAINS ARRAY with your domain(s)." }, { "parameters": { "conditions": { "options": { "caseSensitive": true, "leftValue": "", "typeValidation": "strict" }, "conditions": [ { "id": "condition-skip", "leftValue": "={{ $json.skip }}", "rightValue": true, "operator": { "type": "boolean", "operation": "equals" } } ] } }, "id": "if-skip", "name": "Pre-Filtered?", "type": "n8n-nodes-base.if", "typeVersion": 2, "position": [ 680, 300 ], "notes": "Routes emails based on pre-filter result:\n- TRUE (skip=true): Email was filtered, log and end\n- FALSE (skip=false): Continue to AI classification" }, { "parameters": { "operation": "append", "documentId": { "__rl": true, "mode": "id", "value": "YOUR_GOOGLE_SHEET_ID" }, "sheetName": { "__rl": true, "mode": "list", "value": "Pre-Filter Log" }, "columns": { "mappingMode": "defineBelow", "value": { "timestamp": "={{ $now.toISO() }}", "email_id": "={{ $json.email_id }}", "from": "={{ $json.from }}", "subject": "={{ $json.subject }}", "reason": "={{ $json.reason }}", "classification": "={{ $json.classification_override }}", "confidence": "={{ $json.confidence }}", "message": "={{ $json.message }}" } }, "options": {} }, "id": "sheets-log-skipped", "name": "Log Skipped Email", "type": "n8n-nodes-base.googleSheets", "typeVersion": 4.5, "position": [ 900, 200 ], "credentials": { "googleSheetsOAuth2Api": { "name": "" } }, "notes": "Logs pre-filtered emails to Google Sheets.\n\nCreate a sheet named 'Pre-Filter Log' with columns:\ntimestamp, email_id, from, subject, reason, classification, confidence, message" }, { "parameters": { "inputText": "={{ 'Subject: ' + $json.email_data.subject + '\\n\\nFrom: ' + $json.email_data.from + '\\n\\nBody: ' + ($json.email_data.snippet || $json.email_data.body).substring(0, 1000) }}", "categories": { "categories": [ { "category": "customer_service", "description": "A real customer or potential customer needing help, information, or support. Includes: product questions, order issues, account problems, complaints, feature requests, or general inquiries. WHEN IN DOUBT, USE THIS CATEGORY." }, { "category": "spam", "description": "Promotional content, marketing emails, unsolicited offers, or junk mail. Includes: sales pitches, newsletters, advertisements, cold outreach, special offers, and prize notifications." }, { "category": "internal", "description": "Communications from within the company or related to internal operations. Includes: team updates, HR announcements, internal tool notifications." }, { "category": "automated", "description": "System-generated emails that don't require human response. Includes: order confirmations, shipping notifications, password resets, calendar invites." }, { "category": "other", "description": "Content that doesn't clearly fit the above categories. Use sparingly." } ] }, "options": { "allowMultipleClassesPerItem": false, "enableAutoFixOutput": true } }, "id": "text-classifier", "name": "AI Classification", "type": "@n8n/n8n-nodes-langchain.textClassifier", "typeVersion": 1, "position": [ 900, 400 ], "notes": "Classifies email into 5 categories using AI.\n\nCategories are ordered by priority - customer_service is emphasized as the default when uncertain.\n\nModel: Configure with Gemini Flash 2.5-Lite for lowest cost." }, { "parameters": { "model": "models/gemini-2.0-flash-lite", "options": { "temperature": 0.1 } }, "id": "model-classifier", "name": "Gemini Flash Lite", "type": "@n8n/n8n-nodes-langchain.lmChatGoogleGemini", "typeVersion": 1, "position": [ 900, 600 ], "credentials": { "googlePalmApi": { "name": "" } }, "notes": "Gemini Flash 2.5-Lite for cost-effective classification.\nTemperature: 0.1 for deterministic results.\n\nCost: ~$0.04 per 1M tokens = ~$0.00001 per email" }, { "parameters": { "mode": "manual", "duplicateItem": false, "assignments": { "assignments": [ { "id": "class-result", "name": "classification", "value": "={{ $json.categories[0] || 'customer_service' }}", "type": "string" }, { "id": "class-confidence", "name": "confidence", "value": "={{ $json.categoriesWithScores ? $json.categoriesWithScores[0]?.score : 0.8 }}", "type": "number" }, { "id": "original-email", "name": "email_data", "value": "={{ $('Free Filter (Pre-Check)').item.json.email_data }}", "type": "object" } ] } }, "id": "set-classification", "name": "Format Classification", "type": "n8n-nodes-base.set", "typeVersion": 3.4, "position": [ 1120, 400 ], "notes": "Formats classification result for logging and routing.\n\nDefaults to 'customer_service' if classification fails (conservative)." }, { "parameters": { "operation": "append", "documentId": { "__rl": true, "mode": "id", "value": "YOUR_GOOGLE_SHEET_ID" }, "sheetName": { "__rl": true, "mode": "list", "value": "Classification Log" }, "columns": { "mappingMode": "defineBelow", "value": { "timestamp": "={{ $now.toISO() }}", "email_id": "={{ $json.email_data.id }}", "thread_id": "={{ $json.email_data.thread_id }}", "from": "={{ $json.email_data.from }}", "subject": "={{ $json.email_data.subject }}", "classification": "={{ $json.classification }}", "confidence": "={{ $json.confidence }}", "model": "gemini-2.5-flash-lite", "pre_filter": "passed" } }, "options": {} }, "id": "sheets-log-classified", "name": "Log Classification", "type": "n8n-nodes-base.googleSheets", "typeVersion": 4.5, "position": [ 1340, 400 ], "credentials": { "googleSheetsOAuth2Api": { "name": "" } }, "notes": "Logs AI classification results to Google Sheets.\n\nCreate a sheet named 'Classification Log' with columns:\ntimestamp, email_id, thread_id, from, subject, classification, confidence, model, pre_filter" }, { "parameters": { "conditions": { "options": { "caseSensitive": true, "leftValue": "", "typeValidation": "strict" }, "conditions": [ { "id": "condition-customer", "leftValue": "={{ $json.classification }}", "rightValue": "customer_service", "operator": { "type": "string", "operation": "equals" } } ] } }, "id": "if-customer", "name": "Customer Email?", "type": "n8n-nodes-base.if", "typeVersion": 2, "position": [ 1560, 400 ], "notes": "Routes based on classification:\n- TRUE (customer_service): Continue to CinnaMon agent\n- FALSE (spam/internal/automated/other): End workflow, archive" }, { "parameters": { "mode": "manual", "duplicateItem": false, "assignments": { "assignments": [ { "id": "output-status", "name": "status", "value": "CUSTOMER_EMAIL", "type": "string" }, { "id": "output-next", "name": "next_step", "value": "CinnaMon sentiment analysis", "type": "string" }, { "id": "output-email", "name": "email_data", "value": "={{ $json.email_data }}", "type": "object" }, { "id": "output-classification", "name": "filter_result", "value": "={{ { classification: $json.classification, confidence: $json.confidence } }}", "type": "object" } ] } }, "id": "set-output-customer", "name": "Continue to CinnaMon", "type": "n8n-nodes-base.set", "typeVersion": 3.4, "position": [ 1780, 300 ], "notes": "Customer email identified. Output ready for CinnaMon agent.\n\nConnect this node's output to your CinnaMon workflow." }, { "parameters": { "mode": "manual", "duplicateItem": false, "assignments": { "assignments": [ { "id": "archive-status", "name": "status", "value": "FILTERED", "type": "string" }, { "id": "archive-reason", "name": "reason", "value": "={{ 'Classified as: ' + $json.classification }}", "type": "string" }, { "id": "archive-email", "name": "email_id", "value": "={{ $json.email_data.id }}", "type": "string" } ] } }, "id": "set-output-filtered", "name": "End (Not Customer)", "type": "n8n-nodes-base.set", "typeVersion": 3.4, "position": [ 1780, 500 ], "notes": "Non-customer email. Workflow ends here.\n\nOptionally: Add a Gmail node to archive/label these emails." } ], "connections": { "Watch for Emails": { "main": [ [ { "node": "Free Filter (Pre-Check)", "type": "main", "index": 0 } ] ] }, "Free Filter (Pre-Check)": { "main": [ [ { "node": "Pre-Filtered?", "type": "main", "index": 0 } ] ] }, "Pre-Filtered?": { "main": [ [ { "node": "Log Skipped Email", "type": "main", "index": 0 } ], [ { "node": "AI Classification", "type": "main", "index": 0 } ] ] }, "AI Classification": { "main": [ [ { "node": "Format Classification", "type": "main", "index": 0 } ] ] }, "Gemini Flash Lite": { "ai_languageModel": [ [ { "node": "AI Classification", "type": "ai_languageModel", "index": 0 } ] ] }, "Format Classification": { "main": [ [ { "node": "Log Classification", "type": "main", "index": 0 } ] ] }, "Log Classification": { "main": [ [ { "node": "Customer Email?", "type": "main", "index": 0 } ] ] }, "Customer Email?": { "main": [ [ { "node": "Continue to CinnaMon", "type": "main", "index": 0 } ], [ { "node": "End (Not Customer)", "type": "main", "index": 0 } ] ] } }, "settings": { "executionOrder": "v1" }, "staticData": null, "meta": { "templateId": "email-filter-v1" }, "versionId": "v1-2025-11-27", "tags": [ { "name": "Email", "createdAt": "2025-11-27T00:00:00.000Z", "updatedAt": "2025-11-27T00:00:00.000Z" }, { "name": "Classification", "createdAt": "2025-11-27T00:00:00.000Z", "updatedAt": "2025-11-27T00:00:00.000Z" }, { "name": "Week-4", "createdAt": "2025-11-27T00:00:00.000Z", "updatedAt": "2025-11-27T00:00:00.000Z" }, { "name": "Customer-Service", "createdAt": "2025-11-27T00:00:00.000Z", "updatedAt": "2025-11-27T00:00:00.000Z" } ] }