{ "edges": [ { "to": "6b55d083-2370-478a-812b-02cc06c7fa3a", "from": "669cde21-0d71-45d8-aae8-1104803a67b6", "toPort": "0", "fromPort": "0" }, { "to": "28e7741a-ef0b-470c-8d71-97fe9013a6d9", "from": "669cde21-0d71-45d8-aae8-1104803a67b6", "toPort": "0", "fromPort": "1" }, { "to": "f3b60314-8e02-4a3a-b08f-78ea2a7becf1", "from": "de95cc00-c6a2-4226-ab10-aa7d1b3d0efa", "toPort": "0", "fromPort": "0" }, { "to": "be657b0c-23e1-4b8c-bd25-9915a5947358", "from": "bab39db3-d0d7-422e-aff9-075909f6dd00", "toPort": "0", "fromPort": "0" }, { "to": "de95cc00-c6a2-4226-ab10-aa7d1b3d0efa", "from": "28e7741a-ef0b-470c-8d71-97fe9013a6d9", "toPort": "0", "fromPort": "1" }, { "to": "bab39db3-d0d7-422e-aff9-075909f6dd00", "from": "b63a97fe-e81f-42b3-bf4c-9a48766ea103", "toPort": "0", "fromPort": "ai_languageModel" }, { "to": "bab39db3-d0d7-422e-aff9-075909f6dd00", "from": "f3b60314-8e02-4a3a-b08f-78ea2a7becf1", "toPort": "0", "fromPort": "0" }, { "to": "669cde21-0d71-45d8-aae8-1104803a67b6", "from": "f26790bb-a333-4e1d-9309-f2e30a6c59fd", "toPort": "0", "fromPort": "0" }, { "to": "f26790bb-a333-4e1d-9309-f2e30a6c59fd", "from": "be657b0c-23e1-4b8c-bd25-9915a5947358", "toPort": "0", "fromPort": "0" }, { "to": "bf870fff-71bd-4b9c-ac28-6b2b1fc64da3", "from": "6b55d083-2370-478a-812b-02cc06c7fa3a", "toPort": "0", "fromPort": "0" }, { "to": "28e7741a-ef0b-470c-8d71-97fe9013a6d9", "from": "7ce9b06a-402b-4ff6-b857-431f01c75f2a", "toPort": "0", "fromPort": "0" }, { "to": "28e7741a-ef0b-470c-8d71-97fe9013a6d9", "from": "bf870fff-71bd-4b9c-ac28-6b2b1fc64da3", "toPort": "0", "fromPort": "0" }, { "to": "7ce9b06a-402b-4ff6-b857-431f01c75f2a", "from": "e9618bd5-5db4-4b2c-9b16-41e757e7c818", "toPort": "0", "fromPort": "0" } ], "nodes": [ { "id": "e9618bd5-5db4-4b2c-9b16-41e757e7c818", "name": "When clicking ‘Execute workflow’", "type": "manualTrigger", "params": {}, "position": [ 464, 576 ] }, { "id": "f3b60314-8e02-4a3a-b08f-78ea2a7becf1", "name": "Code in JavaScript", "type": "code", "params": { "jsCode": "function htmlToText(html) {\n if (!html || typeof html !== \"string\") return \"\";\n\n let text = html;\n\n text = text.replace(//gi, \"\");\n text = text.replace(//gi, \"\");\n\n text = text.replace(/<\\/(p|div|section|article|li|tr|h1|h2|h3|h4|h5|h6)>/gi, \"\\n\");\n text = text.replace(//gi, \"\\n\");\n\n text = text.replace(/<\\/?[^>]+>/g, \"\");\n\n text = text\n .replace(/ /gi, \" \")\n .replace(/&/gi, \"&\")\n .replace(/"/gi, \"\\\"\")\n .replace(/'/gi, \"'\")\n .replace(/</gi, \"<\")\n .replace(/>/gi, \">\");\n\n text = text\n .split(\"\\n\")\n .map(line => line.trim())\n .filter(Boolean)\n .join(\"\\n\");\n\n return text;\n}\n\nfunction getDomain(url) {\n try {\n const u = new URL(url);\n return u.hostname.replace(/^www\\./, \"\");\n } catch (e) {\n return \"\";\n }\n}\n\nfunction extractTag(html, tag) {\n if (!html) return \"\";\n const regex = new RegExp(`<${tag}[^>]*>([\\\\s\\\\S]*?)<\\\\/${tag}>`, \"i\");\n const match = html.match(regex);\n return match ? String(match[1]).trim() : \"\";\n}\n\nfunction cleanEntityName(text) {\n if (!text) return \"\";\n return String(text)\n .replace(/reviews?/gi, \"\")\n .replace(/read customer service/gi, \"\")\n .replace(/customer service/gi, \"\")\n .replace(/\\|.*$/g, \"\")\n .replace(/-.*$/g, \"\")\n .replace(/\\s+/g, \" \")\n .trim();\n}\n\n// NEW: extract slug from Trustpilot-like URL\nfunction extractEntitySlugFromUrl(url) {\n if (!url) return \"\";\n try {\n const u = new URL(url);\n // Example path: /review/maclear.ch\n const parts = u.pathname.split(\"/\").filter(Boolean); // [\"review\",\"maclear.ch\"]\n const reviewIdx = parts.findIndex(p => p.toLowerCase() === \"review\");\n if (reviewIdx >= 0 && parts[reviewIdx + 1]) {\n return parts[reviewIdx + 1].trim();\n }\n // Generic fallback: last path segment\n return (parts[parts.length - 1] || \"\").trim();\n } catch (e) {\n // If it's not a valid URL string\n const m = String(url).match(/\\/review\\/([^/?#]+)/i);\n return m ? m[1].trim() : \"\";\n }\n}\n\nconst newItems = [];\n\nfor (const item of items) {\n const json = item.json || {};\n let result = json;\n\n // If wrapped in { results: [ ... ] }, extract first result\n if (Array.isArray(json.results) && json.results.length > 0) {\n result = json.results[0];\n }\n\n // URL can come from different places depending on your node\n const url = result.url || json.url || \"\";\n const html = result.content || result.html || \"\";\n\n const pageText = htmlToText(html);\n const domain = getDomain(url);\n\n // Trustpilot-like slug: maclear.ch / apify.com\n const entitySlug = extractEntitySlugFromUrl(url) || null;\n\n // Prefer h1/title, but fallback to slug if extraction is messy\n const h1Text = extractTag(html, \"h1\");\n const titleText = extractTag(html, \"title\");\n const extractedName = cleanEntityName(h1Text || titleText);\n\n const entityName = extractedName\n ? extractedName\n : (entitySlug ? entitySlug : null);\n\n newItems.push({\n json: {\n url,\n domain,\n entity_slug: entitySlug,\n entity_name: entityName,\n page_text: pageText,\n raw_html_length: html.length\n }\n });\n}\n\nreturn newItems;\n" }, "position": [ 1504, 352 ] }, { "id": "bab39db3-d0d7-422e-aff9-075909f6dd00", "name": "AI Agent", "type": "llm", "params": { "text": "=You are given the cleaned text of a review page.\n\nMetadata:\n- URL: {{ $json[\"url\"] || \"\" }}\n- Domain: {{ $json[\"domain\"] || \"\" }}\n\nFull page text:\n\"\"\"\n{{ $json[\"page_text\"] || \"\" }}\n\"\"\"\n\nFrom this text, you must infer as many reviews as possible and then return ONLY a single JSON object with the following structure:\n\n{\n \"source_url\": \"\",\n \"source_domain\": \"\",\n \"estimated_total_reviews\": 0,\n \"average_rating\": null,\n \"rating_scale_max\": 5,\n \"rating_distribution\": {\n \"1\": 0,\n \"2\": 0,\n \"3\": 0,\n \"4\": 0,\n \"5\": 0\n },\n \"sentiment_distribution\": {\n \"positive\": 0,\n \"neutral\": 0,\n \"negative\": 0\n },\n \"top_positive_keywords\": [],\n \"top_negative_keywords\": [],\n \"top_praises\": [],\n \"top_complaints\": [],\n \"topic_distribution\": {\n \"pricing\": 0,\n \"usability\": 0,\n \"features\": 0,\n \"support\": 0,\n \"performance\": 0,\n \"other\": 0\n },\n \"recent_trend_summary\": \"\",\n \"data_quality_notes\": \"\"\n}\n\nField definitions and instructions:\n\n- \"source_url\":\n - Copy exactly the URL provided.\n\n- \"source_domain\":\n - Copy exactly the domain provided.\n\n- \"estimated_total_reviews\":\n - Your best estimate of how many individual reviews are present in the text.\n - If unclear, use 0.\n\n- \"average_rating\":\n - The average rating across all detected reviews.\n - Use a numeric value like 4.3 if there is enough information.\n - If no rating information is reliably available, use null.\n\n- \"rating_scale_max\":\n - The maximum rating scale observed or assumed (usually 5).\n - If unclear, default to 5.\n\n- \"rating_distribution\":\n - Number of reviews that fall into each star bucket (1–5).\n - If you detect a different scale (e.g., 1–10), still try to map to 1–5 as best as possible.\n - If information is unclear, keep values as 0.\n\n- \"sentiment_distribution\":\n - Number of reviews that are overall positive, neutral, or negative based on the text.\n - If you have no signal at all, use 0 for all three.\n\n- \"top_positive_keywords\":\n - Array of short keywords or phrases frequently mentioned in positive reviews.\n - Examples: [\"easy to use\", \"great support\", \"fast\", \"intuitive interface\"]\n - Maximum 10 entries.\n\n- \"top_negative_keywords\":\n - Array of short keywords or phrases frequently mentioned in negative reviews.\n - Examples: [\"too expensive\", \"bugs\", \"slow performance\", \"poor support\"]\n - Maximum 10 entries.\n\n- \"top_praises\":\n - Short bullet-style sentences summarizing what customers like the most.\n - Example: \"Users appreciate how easy it is to get started.\"\n - 3 to 7 items.\n\n- \"top_complaints\":\n - Short bullet-style sentences summarizing the main pain points or complaints.\n - Example: \"Several users complain about slow response times from support.\"\n - 3 to 7 items.\n\n- \"topic_distribution\":\n - A rough count of how many reviews mention each topic:\n - \"pricing\"\n - \"usability\"\n - \"features\"\n - \"support\"\n - \"performance\"\n - \"other\"\n - Use integer counts. If something is unclear, leave it as 0.\n\n- \"recent_trend_summary\":\n - A short paragraph (3–5 lines) describing:\n - overall tone (improving, stable, deteriorating)\n - any shifts in sentiment or recurring issues\n - anything noteworthy for a product / CX team.\n\n- \"data_quality_notes\":\n - Explain briefly any limitations:\n - e.g., \"Only a small subset of reviews is visible.\"\n - \"Ratings are not explicitly shown, so rating distribution is approximate.\"\n - \"Some reviews appear duplicated.\"\n\nVERY IMPORTANT:\n- If something is unclear or not available, do NOT invent precise numbers.\n- Use conservative estimates or 0/null/empty lists.\n- Return ONLY the JSON object, with no extra commentary, no markdown and no backticks.\n", "prompt": "=You are given the cleaned text of a review page.\n\nMetadata:\n- URL: {{ $json[\"url\"] || \"\" }}\n- Domain: {{ $json[\"domain\"] || \"\" }}\n\nFull page text:\n\"\"\"\n{{ $json[\"page_text\"] || \"\" }}\n\"\"\"\n\nFrom this text, you must infer as many reviews as possible and then return ONLY a single JSON object with the following structure:\n\n{\n \"source_url\": \"\",\n \"source_domain\": \"\",\n \"estimated_total_reviews\": 0,\n \"average_rating\": null,\n \"rating_scale_max\": 5,\n \"rating_distribution\": {\n \"1\": 0,\n \"2\": 0,\n \"3\": 0,\n \"4\": 0,\n \"5\": 0\n },\n \"sentiment_distribution\": {\n \"positive\": 0,\n \"neutral\": 0,\n \"negative\": 0\n },\n \"top_positive_keywords\": [],\n \"top_negative_keywords\": [],\n \"top_praises\": [],\n \"top_complaints\": [],\n \"topic_distribution\": {\n \"pricing\": 0,\n \"usability\": 0,\n \"features\": 0,\n \"support\": 0,\n \"performance\": 0,\n \"other\": 0\n },\n \"recent_trend_summary\": \"\",\n \"data_quality_notes\": \"\"\n}\n\nField definitions and instructions:\n\n- \"source_url\":\n - Copy exactly the URL provided.\n\n- \"source_domain\":\n - Copy exactly the domain provided.\n\n- \"estimated_total_reviews\":\n - Your best estimate of how many individual reviews are present in the text.\n - If unclear, use 0.\n\n- \"average_rating\":\n - The average rating across all detected reviews.\n - Use a numeric value like 4.3 if there is enough information.\n - If no rating information is reliably available, use null.\n\n- \"rating_scale_max\":\n - The maximum rating scale observed or assumed (usually 5).\n - If unclear, default to 5.\n\n- \"rating_distribution\":\n - Number of reviews that fall into each star bucket (1–5).\n - If you detect a different scale (e.g., 1–10), still try to map to 1–5 as best as possible.\n - If information is unclear, keep values as 0.\n\n- \"sentiment_distribution\":\n - Number of reviews that are overall positive, neutral, or negative based on the text.\n - If you have no signal at all, use 0 for all three.\n\n- \"top_positive_keywords\":\n - Array of short keywords or phrases frequently mentioned in positive reviews.\n - Examples: [\"easy to use\", \"great support\", \"fast\", \"intuitive interface\"]\n - Maximum 10 entries.\n\n- \"top_negative_keywords\":\n - Array of short keywords or phrases frequently mentioned in negative reviews.\n - Examples: [\"too expensive\", \"bugs\", \"slow performance\", \"poor support\"]\n - Maximum 10 entries.\n\n- \"top_praises\":\n - Short bullet-style sentences summarizing what customers like the most.\n - Example: \"Users appreciate how easy it is to get started.\"\n - 3 to 7 items.\n\n- \"top_complaints\":\n - Short bullet-style sentences summarizing the main pain points or complaints.\n - Example: \"Several users complain about slow response times from support.\"\n - 3 to 7 items.\n\n- \"topic_distribution\":\n - A rough count of how many reviews mention each topic:\n - \"pricing\"\n - \"usability\"\n - \"features\"\n - \"support\"\n - \"performance\"\n - \"other\"\n - Use integer counts. If something is unclear, leave it as 0.\n\n- \"recent_trend_summary\":\n - A short paragraph (3–5 lines) describing:\n - overall tone (improving, stable, deteriorating)\n - any shifts in sentiment or recurring issues\n - anything noteworthy for a product / CX team.\n\n- \"data_quality_notes\":\n - Explain briefly any limitations:\n - e.g., \"Only a small subset of reviews is visible.\"\n - \"Ratings are not explicitly shown, so rating distribution is approximate.\"\n - \"Some reviews appear duplicated.\"\n\nVERY IMPORTANT:\n- If something is unclear or not available, do NOT invent precise numbers.\n- Use conservative estimates or 0/null/empty lists.\n- Return ONLY the JSON object, with no extra commentary, no markdown and no backticks.\n", "options": { "systemMessage": "You are an AI assistant specialized in analyzing customer reviews and generating aggregated metrics.\n\nYou will receive the cleaned text content of one or more review pages (from platforms like G2, Capterra, Trustpilot, Amazon, Google Play, etc.). The text may be noisy, but it will contain multiple customer reviews, ratings, dates, and comments.\n\nYour job is to:\n- Detect and infer individual reviews from the text (as best as possible).\n- Aggregate the information.\n- Return ONLY a single JSON object with high-level metrics about these reviews.\n\nYou MUST:\n- Follow the JSON schema provided in the user instructions exactly.\n- Return ONLY a valid JSON object, with no extra text, no explanations, and no markdown.\n- Be conservative with assumptions: if something is unclear, use null, 0, or an empty array.\n- Focus on metrics that are useful for product, marketing, and CX teams (quality, sentiment, themes, trends)." }, "provider": "openai", "promptType": "define", "approximated": true, "systemMessage": "You are an AI assistant specialized in analyzing customer reviews and generating aggregated metrics.\n\nYou will receive the cleaned text content of one or more review pages (from platforms like G2, Capterra, Trustpilot, Amazon, Google Play, etc.). The text may be noisy, but it will contain multiple customer reviews, ratings, dates, and comments.\n\nYour job is to:\n- Detect and infer individual reviews from the text (as best as possible).\n- Aggregate the information.\n- Return ONLY a single JSON object with high-level metrics about these reviews.\n\nYou MUST:\n- Follow the JSON schema provided in the user instructions exactly.\n- Return ONLY a valid JSON object, with no extra text, no explanations, and no markdown.\n- Be conservative with assumptions: if something is unclear, use null, 0, or an empty array.\n- Focus on metrics that are useful for product, marketing, and CX teams (quality, sentiment, themes, trends)." }, "position": [ 1728, 352 ] }, { "id": "be657b0c-23e1-4b8c-bd25-9915a5947358", "name": "Code in JavaScript1", "type": "code", "params": { "jsCode": "// This Function node takes the LLM \"output\" field (JSON string),\n// parses it, and flattens it into a structure ready to be saved\n// into Google Sheets (one row per item).\n\nconst newItems = [];\n\nfor (const item of items) {\n const rawOutput = item.json.output;\n\n let parsed;\n if (typeof rawOutput === 'string') {\n try {\n parsed = JSON.parse(rawOutput);\n } catch (error) {\n // If parsing fails, keep debug info\n newItems.push({\n json: {\n parse_error: true,\n error_message: error.message || 'Failed to parse LLM output',\n original_output: rawOutput\n }\n });\n continue;\n }\n } else if (typeof rawOutput === 'object' && rawOutput !== null) {\n // In case the LLM node already returns an object\n parsed = rawOutput;\n } else {\n newItems.push({\n json: {\n parse_error: true,\n error_message: 'Unexpected output type from LLM',\n original_output: rawOutput\n }\n });\n continue;\n }\n\n const ratingDist = parsed.rating_distribution || {};\n const sentimentDist = parsed.sentiment_distribution || {};\n const topicDist = parsed.topic_distribution || {};\n\n // Build a flat JSON object for Google Sheets\n const flat = {\n // Basic metadata\n source_url: parsed.source_url || \"\",\n source_domain: parsed.source_domain || \"\",\n estimated_total_reviews: parsed.estimated_total_reviews ?? null,\n average_rating: parsed.average_rating ?? null,\n rating_scale_max: parsed.rating_scale_max ?? 5,\n\n // Rating distribution (1–5)\n rating_1: ratingDist[\"1\"] ?? 0,\n rating_2: ratingDist[\"2\"] ?? 0,\n rating_3: ratingDist[\"3\"] ?? 0,\n rating_4: ratingDist[\"4\"] ?? 0,\n rating_5: ratingDist[\"5\"] ?? 0,\n\n // Sentiment distribution\n sentiment_positive: sentimentDist.positive ?? 0,\n sentiment_neutral: sentimentDist.neutral ?? 0,\n sentiment_negative: sentimentDist.negative ?? 0,\n\n // Topic distribution\n topic_pricing: topicDist.pricing ?? 0,\n topic_usability: topicDist.usability ?? 0,\n topic_features: topicDist.features ?? 0,\n topic_support: topicDist.support ?? 0,\n topic_performance: topicDist.performance ?? 0,\n topic_other: topicDist.other ?? 0,\n\n // Arrays → strings\n top_positive_keywords: Array.isArray(parsed.top_positive_keywords)\n ? parsed.top_positive_keywords.join(\", \")\n : \"\",\n\n top_negative_keywords: Array.isArray(parsed.top_negative_keywords)\n ? parsed.top_negative_keywords.join(\", \")\n : \"\",\n\n top_praises: Array.isArray(parsed.top_praises)\n ? parsed.top_praises.join(\"\\n\")\n : \"\",\n\n top_complaints: Array.isArray(parsed.top_complaints)\n ? parsed.top_complaints.join(\"\\n\")\n : \"\",\n\n // Text fields\n recent_trend_summary: parsed.recent_trend_summary || \"\",\n data_quality_notes: parsed.data_quality_notes || \"\"\n };\n\n newItems.push({ json: flat });\n}\n\nreturn newItems;\n" }, "position": [ 2080, 352 ] }, { "id": "7ce9b06a-402b-4ff6-b857-431f01c75f2a", "name": "Get row(s) in sheet", "type": "integration", "params": { "options": {}, "service": "googleSheets", "operation": "read", "sheetName": { "__rl": true, "mode": "list", "value": "gid=0", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo/edit#gid=0", "cachedResultName": "input" }, "documentId": { "__rl": true, "mode": "list", "value": "1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo/edit?usp=drivesdk", "cachedResultName": "url_feedback_truspilot" } }, "position": [ 688, 560 ] }, { "id": "f26790bb-a333-4e1d-9309-f2e30a6c59fd", "name": "Append row in sheet", "type": "integration", "params": { "columns": { "value": { "source_url": "={{ $json.source_url }}", "average_rating": "={{ $json.average_rating }}", "sentiment_neutral": "={{ $json.sentiment_neutral }}", "sentiment_negative": "={{ $json.sentiment_negative }}", "sentiment_positive": "={{ $json.sentiment_positive }}", "recent_trend_summary": "={{ $json.recent_trend_summary }}", "top_negative_keywords": "={{ $json.top_negative_keywords }}", "top_positive_keywords": "={{ $json.top_positive_keywords }}", "estimated_total_reviews": "={{ $json.estimated_total_reviews }}" }, "schema": [ { "id": "source_url", "type": "string", "display": true, "required": false, "displayName": "source_url", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "estimated_total_reviews", "type": "string", "display": true, "required": false, "displayName": "estimated_total_reviews", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "average_rating", "type": "string", "display": true, "required": false, "displayName": "average_rating", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "sentiment_positive", "type": "string", "display": true, "required": false, "displayName": "sentiment_positive", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "sentiment_neutral", "type": "string", "display": true, "required": false, "displayName": "sentiment_neutral", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "sentiment_negative", "type": "string", "display": true, "required": false, "displayName": "sentiment_negative", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "top_positive_keywords", "type": "string", "display": true, "required": false, "displayName": "top_positive_keywords", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "top_negative_keywords", "type": "string", "display": true, "required": false, "displayName": "top_negative_keywords", "defaultMatch": false, "canBeUsedToMatch": true }, { "id": "recent_trend_summary", "type": "string", "display": true, "required": false, "displayName": "recent_trend_summary", "defaultMatch": false, "canBeUsedToMatch": true } ], "mappingMode": "defineBelow", "matchingColumns": [], "attemptToConvertTypes": false, "convertFieldsToString": false }, "options": {}, "service": "googleSheets", "operation": "append", "sheetName": { "__rl": true, "mode": "list", "value": 697979134, "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo/edit#gid=697979134", "cachedResultName": "output" }, "documentId": { "__rl": true, "mode": "list", "value": "1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo", "cachedResultUrl": "https://docs.google.com/spreadsheets/d/1TqXzrR4lq9oijo09pLFLqSUz2OK2uCRKu7CPQz0MxWo/edit?usp=drivesdk", "cachedResultName": "url_feedback_truspilot" } }, "position": [ 2304, 352 ] }, { "id": "de95cc00-c6a2-4226-ab10-aa7d1b3d0efa", "name": "Decodo", "type": "integration", "params": { "url": "={{ $('Get row(s) in sheet').item.json.url }}", "service": "decodo", "operation": "run", "__originalType": "@decodo/n8n-nodes-decodo.decodo" }, "position": [ 1280, 352 ] }, { "id": "bf870fff-71bd-4b9c-ac28-6b2b1fc64da3", "name": "Create or update an account", "type": "integration", "params": { "name": "={{ $('Code in JavaScript').item.json.entity_slug }}", "service": "salesforce", "resource": "account", "operation": "upsert", "externalId": "Id", "externalIdValue": "={{ $('Get row(s) in sheet').item.json.salesforce_account_id }}", "additionalFields": { "customFieldsUi": { "customFieldsValues": [ { "value": "={{ $json.recent_trend_summary }}", "fieldId": "feedback__c" } ] } } }, "position": [ 2976, 476 ] }, { "id": "b63a97fe-e81f-42b3-bf4c-9a48766ea103", "name": "OpenAI Chat Model", "type": "llm", "params": { "model": "gpt-4.1-nano", "options": {}, "provider": "openai", "builtInTools": {} }, "position": [ 1800, 576 ] }, { "id": "6b55d083-2370-478a-812b-02cc06c7fa3a", "name": "Code in JavaScript2", "type": "code", "params": { "jsCode": "function truncateTo255Smart(value) {\n if (!value) return \"\";\n\n const text = String(value);\n\n // If already within limit, return as is\n if (text.length <= 255) {\n return text;\n }\n\n // Cut to max length\n const truncated = text.slice(0, 255);\n\n // Try to avoid cutting words\n const lastSpace = truncated.lastIndexOf(\" \");\n\n return lastSpace > 200\n ? truncated.slice(0, lastSpace)\n : truncated;\n}\n\n// Apply truncation to the Salesforce field\nconst safeTrendSummary = truncateTo255Smart($input.first().json.recent_trend_summary);\n\nreturn [\n {\n json: {\n\n recent_trend_summary: safeTrendSummary\n }\n }\n];\n" }, "position": [ 2752, 352 ] }, { "id": "669cde21-0d71-45d8-aae8-1104803a67b6", "name": "If", "type": "if", "params": { "options": {}, "conditions": { "options": { "version": 3, "leftValue": "", "caseSensitive": true, "typeValidation": "strict" }, "combinator": "and", "conditions": [ { "id": "9790b05b-3c77-40d4-b33c-04f97c22129d", "operator": { "type": "string", "operation": "notEquals" }, "leftValue": "={{ $('Get row(s) in sheet').item.json.salesforce_account_id }}", "rightValue": "" } ] } }, "position": [ 2528, 352 ] }, { "id": "28e7741a-ef0b-470c-8d71-97fe9013a6d9", "name": "Loop Over Items", "type": "passthrough", "params": { "mode": "noOp", "options": {} }, "position": [ 1056, 368 ] } ] }

AI-powered review sentiment analysis to Salesforce & Google Sheets with Decodo