{ "name": "Company Sector Classification - Ollama AI", "nodes": [ { "parameters": { "httpMethod": "POST", "path": "classify-company", "responseMode": "responseNode", "options": { "rawBody": true } }, "id": "70efe005-7bd9-40f4-a795-dcda79b978f3", "name": "Webhook", "type": "n8n-nodes-base.webhook", "typeVersion": 2, "position": [ -1440, 336 ], "webhookId": "ea6c3c6a-8fa0-4940-bb55-a4580c37f350" }, { "parameters": { "jsCode": "// Validate URL(s) - handles both single URL and array of URLs\nconst item = $input.first();\nconst urlInput = item.json.body?.url || item.json.url || item.json.body?.urls || item.json.urls || '';\n\n// Normalize to array\nlet urls = [];\nif (Array.isArray(urlInput)) {\n urls = urlInput;\n} else if (typeof urlInput === 'string') {\n // Check if it's a JSON string array\n try {\n const parsed = JSON.parse(urlInput);\n if (Array.isArray(parsed)) {\n urls = parsed;\n } else {\n urls = [urlInput];\n }\n } catch(e) {\n urls = [urlInput];\n }\n} else {\n urls = [String(urlInput || '')];\n}\n\n// Validate each URL and create output items\nconst outputItems = [];\nfor (let i = 0; i < urls.length; i++) {\n const urlString = typeof urls[i] === 'string' ? urls[i].trim() : String(urls[i] || '').trim();\n if (!urlString) continue;\n \n if (!/^https?:\\/\\//i.test(urlString)) {\n throw new Error(`Invalid URL at index ${i}: \"${urlString}\". Must start with http:// or https://`);\n }\n \n outputItems.push({ json: { url: urlString, urlIndex: i, totalUrls: urls.length } });\n}\n\nif (outputItems.length === 0) {\n throw new Error('No valid URLs provided');\n}\n\nreturn outputItems;" }, "id": "0c0f4f65-4ee5-4cf3-9875-a9926505cafb", "name": "Validate URL", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ -1200, 336 ] }, { "parameters": { "url": "={{ $json.url }}", "sendHeaders": true, "headerParameters": { "parameters": [ { "name": "User-Agent", "value": "Mozilla/5.0" } ] }, "options": { "redirect": { "redirect": { "maxRedirects": 3 } }, "response": { "response": { "neverError": true, "responseFormat": "text" } }, "timeout": 30000 } }, "id": "db0051b3-d4e5-48c0-88ed-91266bd17be5", "name": "Fetch Homepage", "type": "n8n-nodes-base.httpRequest", "typeVersion": 4.2, "position": [ -960, 224 ] }, { "parameters": { "jsCode": "// Ensure body is always string and preserve metadata - process ALL items\nconst items = $input.all();\nconst outputItems = [];\n\nfor (const item of items) {\n let htmlString = '';\n if (typeof item.json === 'string') htmlString = item.json;\n else if (item.json && typeof item.json.body === 'string') htmlString = item.json.body;\n else htmlString = JSON.stringify(item.json || '');\n \n // Preserve URL and metadata\n const output = { body: String(htmlString), url: String(item.json.url || '') };\n if (item.json.urlIndex !== undefined) output.urlIndex = item.json.urlIndex;\n if (item.json.totalUrls !== undefined) output.totalUrls = item.json.totalUrls;\n \n outputItems.push({ json: output });\n}\n\nreturn outputItems;" }, "id": "e381d8e7-1497-44d9-afb7-171d6d832541", "name": "Prepare HTML", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ -784, 224 ] }, { "parameters": { "operation": "extractHtmlContent", "dataPropertyName": "body", "extractionValues": { "values": [ { "key": "pageTitle", "cssSelector": "title" }, { "key": "metaDescription", "cssSelector": "meta[name='description']", "returnValue": "attribute" }, { "key": "mainContent", "cssSelector": "main, article, .content, #content, body" } ] }, "options": { "trimValues": true, "cleanUpText": true } }, "id": "66045587-fdbd-4eec-953b-1d328e72168f", "name": "Extract Text", "type": "n8n-nodes-base.html", "typeVersion": 1.2, "position": [ -592, 224 ] }, { "parameters": { "jsCode": "// Preserve metadata after HTML extraction - process ALL items\nconst items = $input.all();\nconst outputItems = [];\n\nfor (const item of items) {\n const output = { ...item.json };\n \n // Ensure URL and metadata are preserved\n if (item.json.url) output.url = item.json.url;\n if (item.json.urlIndex !== undefined) output.urlIndex = item.json.urlIndex;\n if (item.json.totalUrls !== undefined) output.totalUrls = item.json.totalUrls;\n \n outputItems.push({ json: output });\n}\n\nreturn outputItems;" }, "id": "6f8b6234-d046-4844-97bd-e7b51c2a8055", "name": "Preserve Metadata", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ -432, 224 ] }, { "parameters": { "jsCode": "// Build LLM prompt with full glossary - process ALL items\nconst items = $input.all();\nconst outputItems = [];\n\n// Full glossary (defined once, reused for all items)\nconst glossary = `SECTOR CLASSIFICATION GLOSSARY:\\n\\n1. Fintech & Insurtech\\n - Insurtech - Use of technology, digital solutions, and data analytics to disrupt and innovate the traditional insurance industry\\n - Lending & Financing - Providing financial assistance through loans or credit via traditional institutions or fintech platforms\\n - Money Transfer & Management - Online banking, secure movement of funds between parties, digital payment methods\\n - Trading & Investing - Buying, selling, or holding financial assets like stocks, bonds, cryptocurrencies, or commodities\\n - Personal Finance Management - Managing individual's financial resources, budgeting, expense tracking, financial goals\\n\\n2. Retail Platforms\\n - Digital Products - Intangible items sold digitally (software, e-books, online courses, music, digital art)\\n - Services - Intangible activities provided in exchange for payment (consulting, coaching, freelance skills, SaaS subscriptions)\\n - Goods - Physical items sold through retail platforms (electronics, clothing, home items, beauty products, food)\\n\\n3. Business Software - IT, R&D, & Data Solutions\\n - IT & Network Management - Software for administration and management of IT resources and network infrastructure\\n - Data Architecture - Tools for designing and defining data structures, databases, and data models\\n - Data Analysis & Decision Support - Software for analyzing data to provide insights and support decision-making\\n - Project Management & Collaboration - Platforms for planning, executing, and monitoring project work\\n - Product Management Tools - Software for planning, developing, and launching products\\n - Software Development Tools - Tools assisting in software development processes\\n\\n4. Business Software - Operations Solutions\\n - Human Resources Management - Systems for managing employee information, recruitment, onboarding, performance, payroll\\n - Legal and Compliance Solutions - Software ensuring compliance with legal requirements\\n - Financial Operations Software - Tools for accounting, budgeting, invoicing, payroll, financial planning, risk management\\n\\n5. Business Software - Sales & Marketing Solutions\\n - Advertising Platforms - Software for refining and improving online advertising campaigns\\n - eCommerce Tools - Specialized software for online business transactions, storefronts, shopping carts, payment gateways\\n\\n6. Life Sciences & HealthTech - Digital Healthcare\\n - Medical Decision Support - Companies helping physicians make accurate, personalized, data-driven decisions\\n - Patient Remote Monitoring - Technological tools that continually monitor and collect patient data remotely\\n - Digital Accessibility Solutions - Devices or software aiding consumers by improving quality of life\\n - Digital Therapeutics - Digital tools empowering consumers to track, treat and manage medical conditions\\n - Digital Medical Diagnostics - Digital data analytics tools providing automatic, fast, affordable medical diagnostics\\n - Patient Engagement - Technologies increasing patient adherence and compliance with healthcare guidelines\\n - Clinical Workflow - Solutions digitalizing healthcare administrative workflow\\n\\n7. Life Sciences & HealthTech - Medical Devices\\n - Medical Treatment & Therapeutics - Products aimed at remediation of health problems\\n - Assistive Devices - Products helping maintain or improve individual's functioning\\n - Medical Monitors - Remote patient monitoring devices\\n - Diagnostics Devices - Equipment for discovering what is wrong with people who are ill\\n - Aesthetic Devices - Instruments providing desired change in visual appearance\\n - Life Support - Medical procedures keeping patients alive until body functions recover\\n\\n8. Pharma & Medical Biotechnology\\n - Medical Cannabis - Cannabis and cannabinoids prescribed by physicians\\n - Drug Delivery - Approaches for transporting pharmaceutical compounds to target sites\\n - Cosmetic Materials - Constituted mixtures of chemical compounds from natural or synthetic sources\\n - Drugs Discovery & Development - Process of identifying and developing new pharmaceutical drugs\\n - Test Diagnostics & Screening - Tests for early detection of disease or risk factors\\n - Biomaterials & Tissue Engineering - Man-made materials for medical treatments and methods to produce living functional tissue\\n\\n9. Agriculture & Food Technologies - AgriTech\\n - Agriculture Genetics & Inputs - Genetic improvement of crops, genetic engineering, biotechnology, seed breeding\\n - Farm Equipment & Treatment - Equipment used on farms, treatment processes for crops, soil, and livestock\\n - Precision Agriculture - Approach utilizing technology, data, and analytics to optimize agricultural production\\n\\n10. Agriculture & Food Technologies - FoodTech\\n - Produce Preservation - Technologies for preserving and packaging agricultural and food products\\n - Novel Food & Ingredients - Alternative sources-based foods, cultivated foods, fermentation\\n - Food Processing - Update, enrichment, and preparation of existing food with robotics and equipment\\n - Food Trade & Services - Food delivery platforms, meal kit services, restaurant management systems\\n\\n11. Industrial Technologies\\n - Product Fabrication Technologies - Methods and machinery for manufacturing goods (3D printing, CNC machining)\\n - Material Innovations & Applications - Development of new materials or novel application of existing materials\\n - Process Automation & Intelligence Systems - Technologies automating and controlling industrial processes\\n - Operational Optimization Solutions - Software solutions enhancing efficiency and productivity of industrial operations\\n - Microelectronics and Photonics Solutions - Electronic components, semiconductor devices, optical systems, MEMS, quantum computing\\n\\n12. Automotive & Mobility Technologies\\n - Autonomous and Connected Technology - Self-operating vehicles with AI, sensors, V2V, V2I, V2X communication\\n - Electric Mobility & Alternative Fuels - Electric vehicles, charging infrastructure, batteries, hydrogen fuel cells, biofuels\\n - Urban Mobility - Public transit, micro mobility (bikes, scooters), ride-hailing services\\n - Logistics & Transportation Management - Technology optimizing movement of goods, fleet management, route optimization\\n\\n13. Energy Tech\\n - Energy Production & Sources - Technologies for generating energy from fossil fuels, renewable sources, nuclear power\\n - Energy Management & Grid Integration - Infrastructure for distributing energy, smart grids, energy storage\\n - Energy Consumption & Efficiency - Technologies improving energy use in transportation, industry, buildings\\n\\n14. Aerospace, Defense & HLS\\n - Defense & HLS Solutions - Defense tech for navy, air, ground forces; Homeland Security measures\\n - Unmanned Systems - UAVs, UGVs, USVs, robotics with AI and communication modules\\n - Aircrafts & Avionics - Aircraft technologies, aerostructures, sensor integration, targeting\\n - C4I - Command, Control, Communications, Computers, and Intelligence solutions\\n - Defense Electronics - Cybersecurity, communications security, encryption, sensors, signal processing\\n - Space Tech - Space Exploration, Space Components and Infrastructure, Communication & Navigation, Earth Observation\\n\\n15. Cybersecurity\\n - Cloud Security - Securing cloud computing platforms, services, and data\\n - Security Operations - Combining information security and IT operations practices\\n - Data Security - Protecting sensitive data from unauthorized access\\n - Application & Website Security - Securing software applications and web systems\\n - Security Intelligence - Gathering, analysis, and utilization of security threat information\\n - Network & Infrastructure Security - Securing computer networks, firewalls, intrusion detection\\n - GRC & Vulnerability - Governance, Risk Management, Compliance, and security vulnerability management\\n - Industrial Security - Protection of critical infrastructure and industrial control systems\\n - Identity Security - Managing user identities, authentication, and access control\\n - Endpoint Security - Securing endpoints like laptops, desktops, mobile devices\\n\\n16. Media & Entertainment Technologies\\n - Digital Content Production - Technologies facilitating content production (audio, video, graphic design, web design)\\n - Digital Content Distribution - Platforms for content distribution, CMS, DAM, CDNs, social media platforms\\n - Digital Content Engagement & Entertainment - Streaming services, broadcasting, gaming, AR/VR, gadgets\\n\\n17. Education & Knowledge Technologies\\n - Learning/Self Learning Applications - Tools for individual and corporate training, language learning, MOOCs\\n - Teaching & Instruction Tools - Technologies assisting educators in content creation and course management\\n - Assessment & Verification - Tools for evaluating educational programs, automated grading, performance analytics\\n`;\n\nfor (const item of items) {\n let text = item.json.mainContent || '';\n const url = item.json.url || '';\n \n // Clean the text - extract actual content if it's wrapped in JSON\n if (typeof text === 'string' && text.trim().startsWith('{\"data\":')) {\n try {\n const parsed = JSON.parse(text);\n text = parsed.data || parsed.text || String(text);\n } catch (e) {\n const match = text.match(/\\{\"data\":\"([\\s\\S]*?)\"\\}/);\n if (match && match[1]) {\n text = match[1].replace(/\\\\n/g, '\\n').replace(/\\\\\"/g, '\"');\n }\n }\n }\n \n // Ensure text is a clean string\n if (typeof text !== 'string') {\n text = String(text || '');\n }\n \n // Limit text length to avoid token limits\n const maxLength = 8000;\n if (text.length > maxLength) {\n text = text.substring(0, maxLength) + '...';\n }\n \n const prompt = `You are an expert company-sector analyst.\\n\\nGiven the following homepage text, classify the company into a business sector and subsector.\\n\\n${glossary}\\n\\nReturn strictly JSON with fields:\\n{\\n \\\"sector\\\": \\\"...\\\",\\n \\\"subsector\\\": \\\"...\\\",\\n \\\"confidence\\\": 0-100,\\n \\\"reason\\\": \\\"short explanation\\\"\\n}\\n\\nHomepage text:\\n${text}\\n\\nRespond with ONLY valid JSON, no additional text. Select the most appropriate sector and subsector from the list above.`;\n \n // Preserve URL and metadata\n const output = { llmPrompt: prompt, url: url };\n if (item.json.urlIndex !== undefined) output.urlIndex = item.json.urlIndex;\n if (item.json.totalUrls !== undefined) output.totalUrls = item.json.totalUrls;\n \n outputItems.push({ json: output });\n}\n\nreturn outputItems;" }, "id": "5279e6c8-5f7a-4e56-8a20-8c0416fe6cbb", "name": "AI Prompt Builder", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ -256, 224 ] }, { "parameters": { "jsCode": "// Prepare JSON body for Ollama API with proper escaping - process ALL items\nconst items = $input.all();\nconst outputItems = [];\n\nfor (const item of items) {\n const prompt = item.json.llmPrompt || '';\n \n // Create the request body object - n8n will serialize this properly\n const requestBody = {\n model: 'qwen3:latest',\n prompt: String(prompt), // Ensure it's a string\n stream: false,\n format: 'json'\n };\n \n // Preserve URL and metadata\n const output = { requestBody: requestBody, url: item.json.url || '' };\n if (item.json.urlIndex !== undefined) output.urlIndex = item.json.urlIndex;\n if (item.json.totalUrls !== undefined) output.totalUrls = item.json.totalUrls;\n \n outputItems.push({ json: output });\n}\n\nreturn outputItems;" }, "id": "dfc8d12d-f0ba-4695-9dda-3d2e7a138775", "name": "Prepare JSON Body", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ -80, 224 ] }, { "parameters": { "method": "POST", "url": "http://host.docker.internal:11434/api/generate", "sendBody": true, "specifyBody": "json", "jsonBody": "={{ $json.requestBody }}", "options": { "response": { "response": { "neverError": true, "responseFormat": "json" } }, "timeout": 120000 } }, "id": "f11e4f95-1a43-4c65-942d-af5b88933974", "name": "AI Call Ollama", "type": "n8n-nodes-base.httpRequest", "typeVersion": 4.2, "position": [ 128, 224 ] }, { "parameters": { "jsCode": "// Parse AI Response - process ALL items\nconst items = $input.all();\nconst outputItems = [];\n\nfor (const item of items) {\n let responseText = item.json.response || item.json.body?.response || item.json.body || '';\n \n // Handle Ollama response format - response might be in different fields\n if (typeof responseText === 'object') {\n responseText = responseText.response || responseText.text || JSON.stringify(responseText);\n }\n \n responseText = String(responseText).replace(/```json/g,'').replace(/```/g,'').trim();\n \n let parsed;\n try { \n parsed = JSON.parse(responseText); \n } catch(e) { \n parsed = { sector:'Unknown', subsector:'Unspecified', confidence:0, reason:'Parse error: ' + e.message };\n }\n \n // Get URL and metadata from current item (passed through workflow)\n const url = item.json.url || '';\n const urlIndex = item.json.urlIndex;\n const totalUrls = item.json.totalUrls;\n \n // Extract company name from URL\n let companyName = '';\n if (url) {\n try {\n const urlObj = new URL(url);\n let hostname = urlObj.hostname;\n // Remove www. prefix\n hostname = hostname.replace(/^www\\./i, '');\n // Extract main domain (e.g., 'stripe.com' -> 'stripe')\n const domainParts = hostname.split('.');\n if (domainParts.length >= 2) {\n // Take the second-to-last part (usually the company name)\n companyName = domainParts[domainParts.length - 2];\n } else {\n companyName = hostname;\n }\n // Capitalize first letter\n companyName = companyName.charAt(0).toUpperCase() + companyName.slice(1);\n } catch(e) {\n // If URL parsing fails, try to extract from string\n const match = url.match(/(?:https?:\\/\\/)?(?:www\\.)?([^.]+)\\./i);\n if (match && match[1]) {\n companyName = match[1].charAt(0).toUpperCase() + match[1].slice(1);\n }\n }\n }\n \n // Add URL and company name to response\n if (url) parsed.url = url;\n if (companyName) parsed['company name'] = companyName;\n \n // Preserve ordering metadata\n if (urlIndex !== undefined) parsed.urlIndex = urlIndex;\n if (totalUrls !== undefined) parsed.totalUrls = totalUrls;\n \n outputItems.push({ json: parsed });\n}\n\nreturn outputItems;" }, "id": "f49d4bdc-f640-424e-8fa1-f1a1c532948c", "name": "Parse AI Response", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ 320, 224 ] }, { "parameters": { "jsCode": "// Aggregate results - collect all classification results\n// This node receives all items from Parse AI Response\nconst allItems = $input.all();\n\nif (allItems.length === 0) {\n return [{ json: { error: 'No results to aggregate' } }];\n}\n\n// Extract json from each item - items should be in standard n8n format { json: {...} }\nconst results = [];\nfor (let index = 0; index < allItems.length; index++) {\n const item = allItems[index];\n \n // Validate item structure\n if (!item || typeof item !== 'object' || Array.isArray(item)) {\n continue; // Skip invalid items\n }\n \n // Get the json property - must be an object\n if (item.json === undefined) {\n continue; // Skip items without json property\n }\n \n // json must be a plain object (not array, not null)\n if (typeof item.json !== 'object' || item.json === null || Array.isArray(item.json)) {\n continue; // Skip items where json is not a plain object\n }\n \n // Valid item - add it to results\n results.push(item.json);\n}\n\nif (results.length === 0) {\n return [{ json: { error: 'No valid results to aggregate', receivedItems: allItems.length, debug: JSON.stringify(allItems[0]) } }];\n}\n\n// Sort by urlIndex if available to maintain order\nif (results.length > 1 && results[0] && results[0].urlIndex !== undefined) {\n results.sort((a, b) => (a.urlIndex || 0) - (b.urlIndex || 0));\n}\n\n// Remove metadata fields from final output\nconst cleanResults = results.map(result => {\n if (result && typeof result === 'object' && !Array.isArray(result)) {\n const { urlIndex, totalUrls, ...cleanResult } = result;\n return cleanResult;\n }\n return result;\n});\n\n// Return results - n8n requires json to be an object, not an array\n// For single result: return the object directly\n// For multiple results: wrap in an object with 'results' property\nlet response;\nif (cleanResults.length === 1) {\n response = cleanResults[0];\n} else {\n // Multiple results - wrap array in object since n8n doesn't allow arrays directly in json\n response = { results: cleanResults };\n}\n\n// Ensure we return in correct n8n format: [{ json: {...} }]\n\nreturn [{ json: response }];" }, "id": "5197637e-5953-468a-a15b-03e0c4723ced", "name": "Aggregate Results", "type": "n8n-nodes-base.code", "typeVersion": 2, "position": [ 496, 224 ] }, { "parameters": { "respondWith": "json", "responseBody": "={{ $json.results || $json }}", "options": { "responseHeaders": { "entries": [ { "name": "Content-Type", "value": "application/json" } ] } } }, "id": "6cf179f3-f0e6-4601-b8bf-7ce7a1aac1cd", "name": "Respond to Webhook", "type": "n8n-nodes-base.respondToWebhook", "typeVersion": 1.1, "position": [ 720, 224 ] }, { "parameters": {}, "id": "260a7a7c-2b20-4aa6-b419-b7ee3810e794", "name": "When clicking ‘Test workflow’", "type": "n8n-nodes-base.manualTrigger", "typeVersion": 1, "position": [ -1376, 80 ] } ], "pinData": {}, "connections": { "Webhook": { "main": [ [ { "node": "Validate URL", "type": "main", "index": 0 } ] ] }, "Validate URL": { "main": [ [ { "node": "Fetch Homepage", "type": "main", "index": 0 } ] ] }, "Fetch Homepage": { "main": [ [ { "node": "Prepare HTML", "type": "main", "index": 0 } ] ] }, "Prepare HTML": { "main": [ [ { "node": "Extract Text", "type": "main", "index": 0 } ] ] }, "Extract Text": { "main": [ [ { "node": "Preserve Metadata", "type": "main", "index": 0 } ] ] }, "Preserve Metadata": { "main": [ [ { "node": "AI Prompt Builder", "type": "main", "index": 0 } ] ] }, "AI Prompt Builder": { "main": [ [ { "node": "Prepare JSON Body", "type": "main", "index": 0 } ] ] }, "Prepare JSON Body": { "main": [ [ { "node": "AI Call Ollama", "type": "main", "index": 0 } ] ] }, "AI Call Ollama": { "main": [ [ { "node": "Parse AI Response", "type": "main", "index": 0 } ] ] }, "Parse AI Response": { "main": [ [ { "node": "Aggregate Results", "type": "main", "index": 0 } ] ] }, "Aggregate Results": { "main": [ [ { "node": "Respond to Webhook", "type": "main", "index": 0 } ] ] }, "When clicking ‘Test workflow’": { "main": [ [] ] } }, "active": false, "settings": { "executionOrder": "v1" }, "versionId": "a3ce6e2c-3af8-486f-9c67-11d19c675da1", "meta": { "templateCredsSetupCompleted": true, "instanceId": "e09d1f0f8b9f78e80ac70b6dc8726dd263b0b6ffc4300b0eee3b58f6da316f29" }, "id": "OORMfVBOjUWwQRzq", "tags": [] }