Build a chatbot with Reinforced Learning Human Feedback (RLHF) and RAG
Who is this for? This template is designed for internal support teams, product specialists, and knowledge managers who want to build an AI-powered knowledge assistant with retrieval-augmented generation (RAG) and reinforcement learning from human feedback (RLHF) via Telegram. What problem is this workflow solving? Manual knowledge management and answering support queries can be time-consuming and error-prone. This solution automates importing and indexing official documentation into MongoDB vector search and enhances AI responses with Telegram-based user feedback to continuously improve answer quality. What these workflows do Workflow 1: Document ingestion & indexing Manually triggered workflow imports product documentation from Google Docs. Documents are split into manageable chunks and embedded using OpenAI embeddings. Embedded document chunks are stored in MongoDB Atlas vector store to enable semantic search. Workflow 2: Telegram chat with RLHF feedback loop Listens for user messages via Telegram bot integration. Uses vector similarity search on MongoDB to retrieve relevant documentation chunks. Generates answers with OpenAI GPT-4o-mini model using retrieval-augmented generation. Sends answers back via Telegram and waits for user feedback (approval or disapproval). Captures feedback, maps it as positive or negative, and stores it with the conversation data for future model improvement. Setup Setting up vector embeddings Authenticate Google Docs and connect your Google Docs URL containing the product documentation you want to index. Authenticate MongoDB Atlas and connect the collection where you want to store the vector embeddings. Create a search index on this collection to support vector similarity queries. Ensure the index name matches the one configured in n8n (data_index). See the example MongoDB search index template below for reference. Setting up chat with Telegram RLHF Create a bot in Telegram with @botFather using the /newbot command. Connect the MongoDB database and search index used for vector search in the previous workflow. Also create two new collections in MongoDB Atlas: one for feedback and one for chat history. Create a search index for feedback, copying the provided template. Configure the AI system prompt in the “Knowledge Base Agent” node, making sure it references all three tools connected (productDocs, feedbackPositive, feedbackNegative) as provided in the template prompt. Make sure Product documentation and feedback collections must connect to the same MongoDB database. There are three distinct MongoDB collections: one for product documentation, one for feedback, and one for chat history (chat history collection can be separate). Telegram API credentials are valid and webhook URLs are correctly set up. MongoDB Search Index Templates Documentation Collection Index { "mappings": { "dynamic": false, "fields": { "_id": { "type": "string" }, "text": { "type": "string" }, "embedding": { "type": "knnVector", "dimensions": 1536, "similarity": "cosine" }, "source": { "type": "string" }, "doc_id": { "type": "string" } } } } Feedback Collection Index { "mappings": { "dynamic": false, "fields": { "prompt": { "type": "string" }, "response": { "type": "string" }, "text": { "type": "string" }, "embedding": { "type": "knnVector", "dimensions": 1536, "similarity": "cosine" }, "feedback": { "type": "token" } } } }
By NovaNodeGet DNS entries of any domain with uProc
Do you want to control the DNS domain entries of your customers or servers? This workflow gets DNS information of any domain using the uProc Get Domain DNS records tool. You can use this workflow to check existing DNS records in real-time to ensure that any domain setup is correct. You need to add your credentials (Email and API Key - real -) located at Integration section to n8n. You can replace "Create Domain Item" with any integration containing a domain, like Google Sheets, MySQL, or Zabbix server. Every "uProc" node returns multiple items with the next fields per every item: type: Contains the DNS record type (A, ALIAS, AAAA, CERT, CNAME, MX, NAPTR, NS, PTR, SOA, SRV, TXT, URL). values: Contains the DNS record values.
By Miquel ColomerGenerate & auto-post tech news AI avatar videos to social media with Heygen and Blotato
This workflow contains community nodes that are only compatible with the self-hosted version of n8n. Description This fully automated AI Avatar Social Media system creates talking head AI clone videos, WITHOUT having to film or edit yourself. It combines n8n, AI agent, HeyGen, and Blotato to research, create, and distribute talking head AI clone videos to every social media platform every single day. This template is ideal for content creators, social media managers, social media agencies, small businesses, and marketers who want to to scale short-form video creation, without manually filming and editing every single video. Overview Trigger: Schedule Configured to run once daily at 10am AI News Research Research viral news from tech-focused forum, Hackernews Fetch the selected news item, plus discussion comments AI Writer AI writes 30-second monologue script AI writes short video caption Create Avatar Video Call Heygen API (requires paid API plan), specifying your avatar ID and voice ID Create avatar video, optionally passing in an image/video background if you have a green screen avatar (matte: true) Get Video Wait awhile, then fetch completed avatar video Upload video to Blotato Publish to Social Media via Blotato Connect your Blotato account Choose your social accounts Either post immediately or schedule for later" 📄 Documentation Full Tutorial Troubleshooting Check your Blotato API Dashboard to see every request, response, and error. Click on a request to see the details. Need Help? In the Blotato web app, click the orange button on the bottom right corner. This opens the Support messenger where I help answer technical questions.
By Sabrina Ramonov 🍄Catch MailChimp subscribe events
Companion workflow for Mailchimp Trigger node docs
By amudhanQwen-Max: journal paper generation from title/abstract
Introduction Generates complete scientific papers from title and abstract using AI. Designed for researchers, automating literature search, content generation, and citation formatting. How It Works Extracts input, searches academic databases (CrossRef, Semantic Scholar, OpenAlex), merges sources, processes citations, generates AI sections (Introduction, Literature Review, Methodology, Results, Discussion, Conclusion), compiles document. Workflow Template Webhook → Extract Data → Search (CrossRef + Semantic Scholar + OpenAlex) → Merge Sources → Process References → Prepare Context → AI Generate (Introduction + Literature Review + Methodology + Results + Discussion + Conclusion via OpenAI) → Merge Sections → Compile Document Workflow Steps Input & Search: Webhook receives title/abstract; searches CrossRef, Semantic Scholar, OpenAlex; merges and processes references AI Generation: OpenAI generates six sections with in-text citations using retrieved references Assembly: Merges sections; compiles formatted document with reference list Setup Instructions Trigger & APIs: Configure webhook URL; add OpenAI API key; customize prompts Databases: Set up CrossRef, Semantic Scholar, OpenAlex API access; configure search parameters Prerequisites OpenAI API, CrossRef API, Semantic Scholar API, OpenAlex API, webhook platform, n8n instance Customization Adjust reference limits, modify prompts for research fields, add citation styles (APA/IEEE), integrate databases (PubMed, arXiv), customize outputs (DOCX/LaTeX/PDF) Benefits Automates paper drafting, comprehensive literature integration, proper citations
By Cheng Siong ChinGenerate personalized sales outreach with GPT across LinkedIn, Email & WhatsApp
Overview This workflow automates your entire sales outreach process across LinkedIn, Email, and WhatsApp using AI to create hyper-personalized messages for each prospect. Instead of spending hours crafting individual messages, the workflow analyzes your lead data and generates customized connection requests, emails, and WhatsApp messages that feel genuinely personal and researched. The workflow includes a built-in approval mechanism, so you can review all AI-generated messages before they're sent, ensuring quality control while still saving massive amounts of time. How It Works The workflow follows a seven-step process: Step 1: Data Collection The workflow starts by reading your lead data from a Google Sheet. Your sheet should contain information about each prospect including their name, title, company, industry, technologies they use, and any other relevant details that can be used for personalization. Step 2: Batch Processing To prevent overwhelming APIs and ensure smooth operation, the workflow processes leads in batches. Each lead's complete data is prepared and formatted for the AI agent to analyze. Step 3: AI Personalization This is where the magic happens. The AI agent receives all the prospect data and generates three distinct messages: A LinkedIn connection request (under 300 characters) that references their specific role, company, or industry A professional HTML email that demonstrates you've researched their business and explains how you can help A casual WhatsApp message that's friendly and approachable The AI is instructed to make these messages sound completely human, never generic or templated. Step 4: Data Cleanup and Storage The AI's output is parsed and cleaned up, then written back to your Google Sheet in separate columns. This creates a permanent record of all generated messages for your review. Step 5: Manual Approval Before anything gets sent, you receive an email asking for your approval. You can review all the generated messages in your Google Sheet, make any edits if needed, and then approve or reject the batch. This ensures you maintain full control over what goes out. Step 6: LinkedIn Automation Once approved, the workflow triggers your Phantombuster agent to send LinkedIn connection requests using the AI-generated messages. Phantombuster handles the actual LinkedIn interaction safely within their platform's limits. Step 7: Email and Notification Delivery Finally, the workflow sends out the personalized emails via Gmail and optionally notifies you via Telegram for each message sent. This happens sequentially to respect rate limits and maintain deliverability. Setup Requirements Before you can use this workflow, you'll need to set up several accounts and gather credentials: Essential Services: An n8n instance (cloud or self-hosted) A Google account with Google Sheets access A Gmail account for sending emails An OpenAI account with API access (for the AI agent) Phantombuster account (for LinkedIn automation) Optional Services: Telegram account and bot (for notifications) Credentials You'll Need: Google Sheets OAuth2 credentials Gmail OAuth2 credentials OpenAI API key Phantombuster API key and agent ID Telegram bot token and chat ID (if using notifications) How to Use This Workflow Initial Setup: Import this workflow into your n8n instance Add all required credentials in n8n's credential manager Create your Google Sheet with the following columns at minimum: First Name, Last Name, Title, Company Name, Personal Email, Industry, Website. Add three additional columns for output: Connection, AI Email, AI Whatsapp Message Copy your Google Sheet ID from the URL and update it in all Google Sheets nodes Open the AI Agent node and update the prompt with your personal information: your name, title, email, and LinkedIn URL Update the email addresses in the Gmail nodes to your actual email addresses Configure your Phantombuster agent for LinkedIn and add the API key and agent ID Running the Workflow: Add your lead data to the Google Sheet (you can start with just 2-3 leads for testing) Click "Execute Workflow" in n8n to start the process Wait for the AI to generate messages (this takes a few seconds per lead) Check your email for the approval request Review the AI-generated messages in your Google Sheet Reply to the approval email with your decision If approved, the workflow will automatically send LinkedIn requests, emails, and WhatsApp messages Best Practices: Start small. Process 5-10 leads at a time initially to test the quality of AI-generated messages and ensure everything works correctly. Once you're confident in the output, you can scale up to larger batches. Monitor your results. Keep track of response rates in your Google Sheet and adjust the AI prompt if certain types of messages aren't performing well. Respect rate limits. Gmail allows 100-500 emails per day depending on your account type, and LinkedIn has strict limits on connection requests (typically 100 per week through automation tools). Stay well within these limits to avoid account restrictions. Customizing This Workflow The workflow is designed to be highly customizable to fit your specific use case: Personalizing the AI Prompt: The most important customization is in the AI Agent node's prompt. You can modify it to: Emphasize different aspects of your value proposition Change the tone from formal to casual or vice versa Include specific pain points relevant to your target industry Add your company's unique selling points Adjust message length and structure Modifying the Output: You can change what the AI generates by editing the prompt. For example, you might want: Different message types (Twitter DMs instead of WhatsApp) Multiple email variations for A/B testing Follow-up message sequences Industry-specific templates Adding Features: The workflow can be extended with additional nodes: Add time delays between sends to appear more natural Include condition checks to segment leads by industry or company size Connect to your CRM to automatically log activities Add sentiment analysis to filter out negative-sounding messages Implement response tracking by monitoring your inbox Changing Tools: If you prefer different services, you can swap out nodes: Replace Phantombuster with other LinkedIn automation tools Use SendGrid or Mailgun instead of Gmail for higher volume Add Slack notifications instead of Telegram Connect to WhatsApp Business API for official messaging Data Source Alternatives: Instead of Google Sheets, you could: Connect directly to your CRM (HubSpot, Salesforce, Pipedrive) Use Airtable as your database Pull data from CSV files uploaded to cloud storage Integrate with lead generation tools like Apollo or Hunter Tips for Success The quality of your AI-generated messages depends heavily on the data you provide. The more information you have about each prospect (their role, company size, technologies used, recent news, pain points), the more personalized and effective the messages will be. Regularly review and refine your AI prompt based on the responses you're getting. If prospects aren't responding, your messages might be too sales-focused or not personal enough. Adjust the prompt to make messages feel more consultative and helpful. Don't send to your entire list at once. Even with approval gates, it's wise to test with small batches, measure results, iterate on your approach, and then scale up gradually. Always comply with email and LinkedIn best practices. Never spam, always provide value in your outreach, respect people's time and privacy, and make it easy for them to opt out if they're not interested. This workflow is a powerful tool that can save you hours of work while actually improving the quality of your outreach through AI-powered personalization. Use it responsibly and watch your response rates improve.
By Aditya MalurAI-driven lead classification & routing with HighLevel and Azure GPT-4o-mini
Description: Streamline your lead management process with this AI-driven n8n automation template. The workflow fetches opportunities from HighLevel (GHL), enriches them with contact details, and uses Azure OpenAI GPT-4o-mini to analyze each lead’s intent (e.g., Demo Request, Support Query, or Partnership Inquiry). It then automatically routes the lead to the right internal team via email, ensuring instant follow-up and zero delays in response time. Perfect for sales, support, and partnership teams who want to save time on manual triage and ensure every inquiry reaches the correct department within seconds. ✅ What This Template Does (Step-by-Step) ⚡ Manual or Scheduled Trigger Run the workflow manually for on-demand classification or schedule it to execute periodically. 📥 Fetch Opportunities from HighLevel Retrieves all opportunities from your GHL CRM, serving as the starting dataset for AI-powered intent detection. 👤 Fetch Detailed Contact Information Enriches each opportunity with full contact details such as name, email, and message notes. 🧠 AI-Powered Lead Classification Uses Azure OpenAI GPT-4o-mini via the LangChain AI Agent to analyze the lead’s message and determine the intent. Possible outputs include: 🎯 Demo Request 🛠️ Support Query 🤝 Partnership Inquiry 🧾 Post-Processing of AI Response JavaScript logic parses and formats the AI’s output into actionable data for conditional routing. 🔀 Intelligent Routing to Relevant Teams Demo Requests → demo@company.com Support Queries → support@company.com Partnership Inquiries → partnership@company.com Each email includes full contact info and original message context. 📧 Instant Team Notifications Sends neatly formatted emails from a centralized sender (noreply@company.com) to ensure smooth handoff and accountability. 🧠 Key Features 🤖 AI intent classification using Azure OpenAI GPT-4o-mini 🔀 Automated lead routing via email 📋 Structured data enrichment from HighLevel ⚙️ Smart conditional logic for 3 lead categories 📩 End-to-end automation from CRM intake to response 💼 Use Cases 📞 Automatically route demo requests to the sales team 🛠️ Send support-related queries directly to helpdesk 🤝 Forward partnership inquiries to business development 💡 Reduce response delays and manual triage errors 📦 Required Integrations HighLevel (GHL) – for opportunity and contact data Azure OpenAI – for AI-driven lead classification SMTP / Gmail – for team routing email notifications 🎯 Why Use This Template? ✅ Automates manual lead sorting and tagging ✅ Ensures every inquiry reaches the right team ✅ Increases response speed and lead conversion ✅ Scalable AI logic adaptable to any organization
By Rahul JoshiReal-time sales quote creation in Odoo via Telegram with Google Gemini AI
Overview This template connects Telegram with Odoo to let your sales team create sales quotes and check product availability in real-time — just by sending chat messages. It’s designed for sales representatives, distributors, and small business owners who want to manage quotes and product information quickly without logging into Odoo. ⚙️ How It Works Once configured, this workflow listens to your Telegram bot for incoming messages. Based on the message text, it performs different actions in Odoo: Product Queries Sales reps can ask about products directly in Telegram: “What’s the price of Product B?” “How many units of Product A are available?” The workflow fetches real-time data from Odoo and replies instantly. Sales Quote Creation Sales reps can also create new sales quotes by typing messages like: “My customer Amazon, his email address is abc@amazon.com wants to buy 10 pcs of Product A and 15 pcs of Product B.” The workflow extracts relevant details, creates a sales quote in Odoo, and sends confirmation back in Telegram. 🧰 Setup Instructions Create a Telegram Bot Go to @BotFather on Telegram. Create a new bot and copy the API Token. Prepare Odoo Enable the Sales and Product modules. Generate an API Key from your Odoo user account. Note your Odoo URL (e.g., https://yourcompany.odoo.com). Import Workflow Open your n8n instance (self-hosted or cloud). Click Import Workflow and upload the provided JSON file. Add Credentials Configure your Telegram credentials (Bot Token). Configure your Odoo credentials (Base URL + API Key). Activate the Workflow Set the workflow to active to start listening for Telegram messages. Send a sample message to your bot to test. 🧠 Use Cases Sales reps capturing orders in the field SMEs managing customer inquiries directly from Telegram Real-time price and stock lookups without opening Odoo Automation of repetitive sales quote tasks 🎛️ Customization Options This workflow can be easily adapted to your business needs: Change trigger platform: Replace Telegram with WhatsApp, Slack, or Discord using the respective n8n nodes. Extend data fields: Add fields like delivery date, salesperson, or payment terms. Auto-confirm orders: Add a node to automatically confirm a Sales Quote once approved. ✅ Requirements Odoo v14 or later (with Sales module enabled) Telegram Bot Token n8n instance (Cloud or self-hosted) 💬 Example Prompts Product Query: “What’s the price of Product B?” “How many units of Product A are available?” Order Entry: “My customer Amazon, his email address is abc@amazon.com wants to buy 10 pcs of Product A and 15 pcs of Product B.”
By Evozard