n8n AI Chatbot: Web & WhatsApp
£250-750 GBP
Posted 1 minute ago
•
Ends in 6 days
£250-750 GBP
Paid on delivery
I’m building an AI-powered support assistant orchestrated entirely in n8n. In this first phase the bot must respond on two fronts—our website’s live-chat widget and our WhatsApp number—using the same brain and conversation history regardless of channel. Knowledge base The answers should be generated from: • Content scraped or pulled via API from our public help-center website • A collection of PDFs/manuals I will upload Retrieval, embedding, and prompt-engineering logic can live inside n8n itself or call out to external services (e.g. OpenAI, Pinecone, Supabase, Weaviate); I’m open as long as the workflow remains transparent and editable in n8n. Fallback flow Whenever the model’s confidence drops below an agreed threshold it must hand the dialogue—along with full context—to a human agent. I don’t want the bot asking customers for extra details; it should immediately escalate. Deliverables 1. n8n workflow (export file) covering: ingestion of sources, vector store updates, real-time chat handling for website and WhatsApp, fallback routing. 2. Installation/connection notes for the live-chat widget you choose (if it’s not already native to n8n) and the WhatsApp API (Twilio, 360dialog, or other). 3. Short Loom or markdown walkthrough so my team can extend the flow and add new documents later. Acceptance criteria • Bot answers at least 80 % of a supplied test set correctly within 2 seconds. • Low-confidence queries are routed to the human queue every time, with transcript attached. • Entire solution runs in my self-hosted n8n instance with no proprietary vendor lock-in. If this sounds like a build you can knock out efficiently, let’s talk timelines and any prerequisites you have.
Project ID: 40827395
About the project
55 proposals
Open for bidding
Remote project
Active 3 secs ago
Place your bid
Bid amount Email addressBenefits of bidding on Freelancer
Set your budget and timeframe
It's free to sign up and bid on jobs
55 freelancers are bidding on average £457 GBP for this job


Hi, I am Elias from the sunny state of Florida. I understand you're building an AI-powered support assistant using n8n. My main focus would be ensuring seamless integration of the chatbot with both web and WhatsApp platforms, leveraging n8n’s capabilities for workflow automation. This will involve setting up efficient API integrations and ensuring that your chatbot can handle user queries effectively while maintaining reliability. For the technical approach, I would prioritize the design of workflows that allow for dynamic responses and data handling, ensuring the system is scalable to accommodate future enhancements. Implementing robust error handling and testing strategies will also be essential to prevent downtime and improve user experience. A few questions to clarify the scope: Q1 – What specific functionalities do you envision for the chatbot in terms of user interaction? Q2 – Are there any existing systems or databases that the chatbot needs to integrate with? Q3 – What metrics or KPIs will you use to measure the chatbot's performance? Looking forward to hearing from you.


I’d build this as one n8n support workflow, not two channel-specific bots. A self-hosted Chatwoot widget and WhatsApp inbox can feed the same n8n flow, with conversation history stored against a shared customer ID. Cross-channel continuity will depend on a reliable way to link a website visitor to their WhatsApp number; I’ll define that mapping with you before implementation. For answers, n8n will ingest help-center pages and uploaded PDFs, track source changes, and update a pgvector knowledge store. I’ll use retrieved evidence and a calibrated handoff rule rather than trusting the model’s own confidence score. If evidence is weak, the flow will stop answering and route the full transcript to a human queue immediately—without asking the customer to repeat themselves. You’ll receive editable n8n exports, connection notes, and a walkthrough for adding documents. I’ll also test against your supplied questions, measure the 80% accuracy and 2-second targets, and check every low-confidence case for correct routing. The stack can remain self-hosted and replaceable. To plan the build, please share your n8n version, help-center URL, sample PDFs, test set, and whether you already have a WhatsApp Business API connection or agent inbox.


Hi, Building a unified AI support assistant for both your website live-chat and WhatsApp using a shared n8n pipeline is a smart approach. First, I would check the structure of your help-center sitemap and PDF formats to map out the ingestion pipeline. Architecting custom automation and retrieval systems like this is my daily work. My plan is to set up a self-hosted n8n workflow using an open-source vector store, configure the chunking and embedding logic with OpenAI, and implement a strict confidence threshold that triggers an immediate human escalation routing with the full transcript. For £336, delivery in 1 day includes the complete, exportable n8n workflow file, WhatsApp API setup via Meta Cloud or Twilio, and a clear Loom walkthrough for your team. Shall we confirm whether you prefer Pinecone or a self-hosted vector database for the backend?


Hi, I can build this in n8n as a fully transparent, self-hosted AI support workflow for both website chat and WhatsApp, using one shared knowledge layer and conversation history across channels. I’ll set up ingestion for your help-center content and PDFs/manuals, chunk and embed the data, and store it in a vector database connected through n8n. For the website widget, I’ll integrate a live-chat solution that can send/receive messages through n8n webhooks. For WhatsApp, I’ll connect Twilio, 360dialog, or your preferred provider. I’ll also add confidence checks so low-certainty queries are routed instantly to a human with full transcript and context attached, without asking the customer for more details. I’ll deliver the exported n8n workflow, setup notes, and a clear walkthrough so your team can extend the flow and add documents later. Start the chat so we can discuss the best setup for your stack. Regards, Sai Bhaskar


Hi, You are looking to create an AI-powered support assistant using n8n that operates seamlessly through your website's live-chat widget and WhatsApp, maintaining conversation history across both platforms. My expertise in workflow automation and AI chatbot integration, particularly with n8n, makes this project a strong fit for me. To accomplish this, I would develop an n8n workflow that integrates the scraping of content from your help center and utilizes the PDFs/manuals you provide for answer generation. I would ensure that the chatbot responds efficiently using the system's capabilities while embedding the necessary retrieval and prompt-engineering logic. The fallback flow will be set up to escalate low-confidence responses directly to a human agent without prompting customers for additional information. Additionally, I will deliver the n8n workflow file together with installation notes for the chosen live-chat widget and WhatsApp API, alongside a concise walkthrough for your team to manage and extend the flow later. What specific metrics are you using to determine the bot’s performance against your acceptance criteria? This will guide the setup to ensure a successful implementation.


Hi there, I noticed you’re developing an AI-powered support assistant using n8n, requiring seamless integration with both your website’s live chat and WhatsApp. I have hands-on experience with n8n workflows and can help build the necessary logic for fetching responses from your help-center and PDFs, ensuring a smooth chat interaction. I can create a robust n8n workflow that captures and processes data efficiently, ensuring that low-confidence queries are escalated properly to human agents along with full context. I’ll also include detailed installation notes for the live-chat widget and WhatsApp API, as well as a brief walkthrough for your team to manage future updates on their own. Regards, Ali


Hi there, Building a multi-channel AI support assistant in n8n can often lead to fragmented conversation histories or messy context handoffs between WhatsApp and web widgets if the webhook routing and vector memory are not structured cleanly from the start. I will build a robust, self-hosted n8n workflow that unifies your help-center data and PDFs into a single vector store, handles real-time cross-channel memory, and executes instant, context-aware human escalations whenever confidence drops. Before we lock in the architecture, here are my questions: 1. Which specific vector database and LLM provider do you prefer to host within your self-hosted n8n stack, or should I recommend the most optimal lean setup? 2. Do you already have a preferred live-chat widget and WhatsApp API provider like Twilio or 360dialog in mind for the API bindings? Drop me a message so we can get started right away.


Hi, I've built several n8n automation workflows and integrated multi-channel chatbots before, so I understand the architecture you need—unified conversation state across web and WhatsApp with reliable fallback routing. I'll design the entire workflow in n8n: document ingestion from your help center and PDFs, vector embeddings stored in your choice of Pinecone or Supabase, real-time message handling from both channels, confidence-based escalation to your human queue with full context attached, and no external dependencies beyond the APIs you already control. The workflow will be exported as a single file your team can modify, and I'll include setup notes for your live-chat widget and Twilio or 360dialog integration. To hit your 80% accuracy target and 2-second response window, I'll test against your sample queries before handoff and tune the retrieval and prompt logic based on what actually works with your knowledge base. I'm available to start now and will stay on for any adjustments once it's running in your instance.


Hey, bridging the website chat and WhatsApp through a single n8n brain sounds like a really solid setup. I usually handle the shared memory bit by tying both incoming webhook nodes to the same vector store and conversation table. Getting that confidence score to cleanly drop into a human escalation branch without annoying the user is straightforward once the router node is dialed in. Are you already leaning toward a specific widget for the site, or should we just plug into something lightweight that drops right into n8n?


[N8N AI SUPPORT AUTOMATION] Hi, I'm Fernando from Mexico, the Land of Enchantment. The important part of this project is keeping website chat and WhatsApp connected to the same conversation history, so customers get consistent answers regardless of where they contact you. I can build this around self-hosted n8n, with document ingestion, PDF processing, website content retrieval, embeddings and a shared knowledge base. Both channels will use the same retrieval and response logic, with conversation context preserved. I'd also implement confidence-based escalation so uncertain queries are immediately transferred to a human agent with the full conversation transcript, without asking customers unnecessary questions. I'll deliver the editable n8n workflows, API connection instructions and a clear handover guide. I'd also validate the response quality against your supplied test set and measure actual response times. Do you already have a WhatsApp API provider and preferred live-chat widget? Kind regards, Fernando


I can build your omnichannel AI support assistant directly inside a self-hosted n8n instance to handle both your website live-chat widget and WhatsApp while maintaining unified conversation history. My approach focuses on creating a transparent, editable workflow that ingests your help-center content and PDF manuals into a vector store for accurate RAG-based responses. I will configure the retrieval and prompt logic to ensure rapid query resolution meeting your performance targets. For the fallback flow, I will implement a confidence threshold check that immediately routes low-confidence interactions along with the full transcript to your human agent queue without prompting customers for extra details. You will receive the complete n8n workflow export, clear connection notes for the live-chat widget and WhatsApp API, and a concise markdown walkthrough for your team to easily add new documents in the future. Let us discuss your preferred live-chat widget and timelines to get started on this build.


Hello. [[SHARED N8N AI SUPPORT ASSISTANT]] — you need one consistent support brain across website chat and WhatsApp, with shared conversation history and immediate human escalation when confidence is low. I would build this as transparent, editable n8n workflows: one ingestion flow for help-center content and uploaded PDFs, a retrieval pipeline with a replaceable vector-store integration, and a shared chat flow for both channels. Conversations would use a common customer/session identifier so channel history can be retrieved consistently. For fallback, I would apply an agreed confidence rule and route uncertain queries directly to the human queue with the full conversation context and transcript, without asking customers follow-up questions. I would also provide the exported workflows, connection notes, and a practical walkthrough for maintaining the knowledge base. The 80% accuracy and 2-second response targets need to be measured against your supplied test set and hosting, model, and retrieval setup rather than assumed in advance. Which live-chat widget and WhatsApp provider do you currently use, and where should escalations reach your human agents? Looking forward to working with you.


I’ve set up n8n chatbot flows before, both web widget and WhatsApp via the Business API. For phase one I’d start with the core n8n workflow, intent handling, and API hooks, then layer in WhatsApp once the web side is solid. Can start today, working version in 4 days. The 250-750 range and timeline here are starting estimates based on the post. Once we go over full scope, especially the WhatsApp integration details, numbers may shift. Want me to send a quick scope doc?


Hi there, The main challenge is creating one transparent, self-hosted support workflow that keeps knowledge retrieval and conversation history consistent across live chat and a second customer channel, while reliably escalating low-confidence replies with the complete transcript. I have strong hands-on experience with Automation, Software Architecture, and Engineering, and have solved similar problems involving workflow orchestration, retrieval pipelines, and reliability-focused integrations. My approach would be to: * Review the existing self-hosted setup, source content, channel requirements, escalation process, and acceptance test set. * Identify the root cause of likely latency, context, and confidence-handling risks, determining what can be reused versus what needs to change. * Implement a clean, maintainable solution using editable workflow automation, with Java where service-side logic is genuinely needed. * Test ingestion, document updates, retrieval quality, response latency, conversation continuity, edge cases, security, and human handoff behavior. * Deploy or integrate the workflow without disrupting the existing system, then provide installation notes and a concise maintenance walkthrough. Confidence thresholds should be enforced as a deterministic routing gate, with the original query, retrieved evidence, conversation state, and generated response attached to every escalation rather than relying on the model alone. I prefer fixing the actual root cause rath


Your support assistant needs one brain that answers the same way on your website chat and on WhatsApp, pulls from your help center and manuals, and hands off to a human without making customers repeat themselves. We have built AI chat and RAG applications with OpenAI and Anthropic APIs, PostgreSQL with pgvector, and Node.js/Python services, including document ingestion, chunking, embeddings and semantic search. The same retrieval logic can run inside your n8n instance with editable nodes, so your team keeps full control. Our approach: 1. Ingestion: scheduled n8n flows pull help-center content via API or scrape, plus your uploaded PDFs, then chunk, embed and store them with source metadata. 2. Chat: one n8n workflow handles website and WhatsApp messages, shares conversation history, and retrieves context from the vector store. 3. Confidence and fallback: answers below the agreed threshold are sent to your human queue immediately, with the full transcript and context attached. The bot will not ask customers for extra details. 4. Testing: we will run your test set and report the accuracy and response times against the 80% and 2-second targets. 5. Handover: installation notes for the chat widget and WhatsApp provider (Twilio or 360dialog), plus a markdown walkthrough showing how to add documents. We can start right away and deliver in 14 days. To confirm the plan, could you share your help-center URL, the WhatsApp provider you use, and the live-chat widget you have on the site Thanks, Differenz


⚠️ If you're not happy, you don’t pay. ⚠️ Hi Client, thank you for checking my proposal and sharing the detailed project brief. I can build your AI-powered support assistant using n8n, OpenAI, and Supabase with a scalable and efficient design. I will deliver: • Comprehensive n8n workflow for data ingestion, vector store updates, and real-time chat handling • Integration notes for the chosen live-chat widget and WhatsApp API • A detailed Loom video or markdown guide for your team to extend the workflow • Transparent fallback routing to human agents with full conversation context • Performance metrics ensuring 80% accuracy on test queries within 2 seconds You will also receive ongoing support to ensure smooth transitions as your knowledge base expands. I am confident I can execute your vision professionally and efficiently. Looking forward to discussing timeline and next steps. Best regards, Chirag Pipal


⭐⭐⭐⭐⭐ Hi there, Could you confirm that you need one transparent, self-hosted n8n workflow serving both website chat and WhatsApp with shared conversation history? The key priorities are reliable PDF/help-center ingestion, accurate retrieval within two seconds, and immediate human escalation with the complete transcript when confidence is low. The workflow should remain editable and avoid vendor lock-in. I have built AI automation workflows using Python, OpenAI, Azure, AWS, and API integrations, including knowledge-grounded assistants and structured fallback handling. I can organize the n8n flow so your team can maintain sources and extend it easily. ❓ Which WhatsApp provider and human-agent queue should the workflow connect to? ❓ Do you already have a preferred website chat widget and vector store? The main focus will be connecting ingestion, embeddings, retrieval, channel handling, and confidence-based routing in clear n8n steps. I’m ready to begin immediately and document the installation and extension process. I look forward to your reply and discussing the prerequisites and timeline. Best regards, Phuong Nam D


Hello! We can build your n8n-orchestrated AI support assistant covering both your website chat widget and WhatsApp, sharing one brain and conversation history across channels. Here's how we can help: - Ingestion workflow: scrape or pull your help-center content via API, process uploaded PDFs, chunk and embed into a vector store (Supabase, Pinecone, or Weaviate—your choice). - Real-time chat handling for both channels, with unified history so context carries across web and WhatsApp. - Confidence threshold with immediate human handoff, passing full transcript and context without asking customers for extra details. - Deliverables: exported n8n workflow, connection notes for the chat widget and WhatsApp API (Twilio or 360dialog), and a Loom or markdown walkthrough. - Fully self-hosted in your n8n instance, no vendor lock-in. To clarify, which vector store and LLM do you prefer, and is your n8n instance already self-hosted with credentials ready? Also, do you have a preference between Twilio and 360dialog for WhatsApp, and which live-chat widget do you use? Price is negotiable. Let's map the flow.


Hello! I’d be happy to build your n8n-powered AI support assistant with one shared knowledge base and conversation history across website live chat and WhatsApp. The priority will be consistent answers, fast responses, seamless channel switching, and immediate human escalation whenever the bot cannot answer reliably. I recommend a transparent RAG architecture using n8n, OpenAI embeddings, and a vector database such as Supabase or Qdrant. Website content and uploaded PDFs will be processed into searchable knowledge, keeping the workflow editable and avoiding unnecessary platform lock-in. I will build the ingestion and update workflows, retrieval and response logic, shared conversation storage, website chat integration, WhatsApp API connection, and human handoff with full conversation transcripts. I’ll then test against your supplied questions, measure response times and accuracy, and refine retrieval and prompts against the agreed acceptance criteria. You’ll receive the exported workflows, connection and deployment notes, and a practical handover guide. I’d be happy to review your existing n8n setup and discuss the quickest reliable implementation. Thank you!


One challenge here is keeping the same conversation history across the live-chat widget and WhatsApp without duplicating or losing state, especially when a fallback hands the full context to a human. I would design this as a single n8n execution that receives a normalized message object with a conversation ID, so both channels call the same webhook and the history is stored centrally, which is what makes the 80% accuracy and 2-second response target achievable. The first concrete step I would take is mapping the help-center structure and choosing the scraping or API method that preserves clean text chunks for embedding. This matters because retrieval quality determines both answer accuracy and how often the fallback triggers, so it has to be right before building the chat routing. I have built n8n workflows with vector stores, OpenAI embeddings, and Twilio WhatsApp integration, and I can deliver the export file, connection notes, and a short walkthrough video. The solution will run entirely in your self-hosted n8n with no vendor lock-in.
内容来自 Freelancer News,版权归原作者所有。
阅读原文 ↗