tiledesk-chatbot

🥈Silver

The Tiledesk Chatbot Engine is an open-source Node.js framework for building multichannel chatbots. It enables operations teams to automate customer interactions across various platforms. The engine integrates with the Tiledesk Design Studio for bot management and supports hybrid search and retrieved-augmented-generation for enhanced conversational capabilities.

70400Updated 6d ago

Intermediate30min to implementautomation

Saves ~400 min per use

Quick InstallView Source

git clone https://github.com/Tiledesk/tiledesk-chatbot.git

Works with:

Claude

Overview

About This Skill

How to Use

1. **Set Up the Tiledesk Environment:** - Sign up for a Tiledesk account at [tiledesk.com](https://www.tiledesk.com) and create a new project. - Install the Tiledesk Chatbot Engine locally using `npm install -g tiledesk-chatbot` or deploy it to a cloud provider like Vercel or AWS. 2. **Design the Bot Flow:** - Use the **Tiledesk Design Studio** (drag-and-drop interface) to map out the conversation paths. Start with a greeting message and primary intents (e.g., order status, refunds, product info). - Configure **hybrid search** by connecting your knowledge base (e.g., Notion, Google Drive, or a custom API) to the bot’s search component. Use the `hybridSearch` plugin in the bot’s code to enable this. 3. **Integrate APIs and Data Sources:** - For dynamic data (e.g., order status, product details), set up API endpoints in the bot’s code. Use the `axios` library to fetch data from your CRM, e-commerce platform, or ERP. - Enable **RAG (Retrieval-Augmented Generation)** by configuring the bot to pull context from your knowledge base before generating responses. This improves accuracy for complex queries. 4. **Test and Debug:** - Use Tiledesk’s **testing environment** to simulate conversations and validate the bot’s responses. Check edge cases (e.g., typos, ambiguous queries) and refine the flow. - Monitor the **conversation logs** in Tiledesk to identify patterns in user queries and adjust the bot’s intents or responses accordingly. 5. **Deploy and Monitor:** - Deploy the bot to your desired channels (e.g., WhatsApp, website chat, Facebook Messenger) via the Tiledesk dashboard. - Track performance using **Tiledesk Analytics** to measure resolution rates, escalation triggers, and customer satisfaction. Iterate based on feedback and data. **Pro Tip:** Start with a narrow scope (e.g., handle only order status queries) and expand the bot’s capabilities over time. Use Tiledesk’s **webhook system** to trigger external actions (e.g., sending an email confirmation) when specific intents are detected.

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/Tiledesk/tiledesk-chatbot

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Build a Tiledesk chatbot that handles [SPECIFIC_CUSTOMER_SERVICE_TASK] for [COMPANY_NAME]. Use the Tiledesk Design Studio to create a flow that [DESCRIBE_PRIMARY_FUNCTION, e.g., 'routes billing inquiries to the finance team, processes refund requests, and escalates unresolved issues to a human agent']. Include hybrid search to retrieve relevant FAQ articles from [KNOWLEDGE_BASE_SOURCE] and use RAG to generate context-aware responses. Test the bot in staging and deploy to production channels: [LIST_CHANNELS, e.g., 'WhatsApp, Facebook Messenger, and the company website'].

Example Output

Here’s a Tiledesk chatbot flow designed for **GreenLeaf Organics**, an organic grocery delivery service, to handle customer inquiries about orders and subscriptions:

**Bot Flow Overview:**
1. **Trigger:** Customer initiates chat via WhatsApp or the company website.
2. **Greeting:** The bot welcomes the user and asks, *'Hi! How can I help you today? You can ask about orders, subscriptions, or product details.'*
3. **Intent Detection:** Using hybrid search, the bot queries the **GreenLeaf FAQ database** (hosted on Notion) to match the user’s query. For example, if the user asks *'Why hasn’t my order arrived yet?'*, the bot retrieves the relevant FAQ: *'Orders typically arrive within 3-5 business days. If your order is delayed, check your tracking number or contact support.'*
4. **Action Handling:**
- **Order Status:** If the user provides an order number, the bot queries the **GreenLeaf CRM** (via API) and responds with the latest status (e.g., *'Your order #2024-0543 is out for delivery and will arrive today by 6 PM.'*).
- **Subscription Changes:** For requests like *'How do I pause my subscription?'*, the bot guides the user to the self-service portal or escalates to the support team if needed.
- **Product Questions:** If the user asks about ingredients or allergens, the bot pulls details from the **product catalog API** and responds with: *'Our Organic Quinoa Salad contains quinoa, cherry tomatoes, and a lemon-tahini dressing. It’s vegan and gluten-free.'*
5. **Escalation:** If the user’s query isn’t resolved (e.g., *'I never received my refund!'*), the bot transfers the chat to a human agent with a summary of the conversation and prior interactions.
6. **Fallback:** If no intent is detected, the bot offers to connect the user to a live agent or suggests browsing the FAQ.

**Deployment & Testing:**
- The bot was tested in Tiledesk’s staging environment with 50 simulated user interactions, achieving a 92% resolution rate without human intervention.
- It was deployed to WhatsApp and the company website, with analytics showing a 30% reduction in support ticket volume for common queries within the first week.
- The bot’s hybrid search uses **Tiledesk’s built-in vector database** (powered by Pinecone) to ensure responses are accurate and up-to-date.

**Metrics After 2 Weeks:**
- 1,247 chats handled autonomously (78% of total inquiries).
- Average resolution time dropped from 12 minutes to 2.5 minutes.
- Customer satisfaction (CSAT) scores for bot interactions improved from 78% to 91%.

**Next Steps:**
- Add a **feedback loop** to let users rate responses and improve the FAQ database.
- Integrate with **Slack** for internal team notifications when the bot escalates chats.
- Expand to **Facebook Messenger** and **Instagram DMs** by Q3.