Customer Service Bot

In this tutorial you will build a Telegram-based customer service assistant that answers questions from a structured FAQ knowledge base, greets new conversations automatically, and maintains a professional tone throughout.

Time to complete: ~20 minutes

Goal

By the end of this tutorial you will have:

• A Telegram group dedicated to customer support
• A knowledge base loaded with your FAQ documents
• A professional persona with a custom system prompt
• An automated greeting message sent to every new conversation
• A tested bot that answers real customer questions accurately

Prerequisites

• NanoGemClaw is installed and running (npm run dev)
• You have a Telegram account and the bot token configured
• The dashboard is accessible at http://localhost:3000

No prior bot experience needed

This tutorial walks through every step in the UI. No code is required.

Step 1 — Register a group for customer service

First, create a dedicated Telegram group for customer support, then register it with NanoGemClaw.

Create the Telegram group:

1. Open Telegram and tap the compose icon.
2. Select New Group, add your bot as a member.
3. Name the group (e.g., Acme Support).
4. Open Group Info → Administrators and grant the bot admin rights so it can read all messages.

Get the group's chat ID:

• Add @userinfobot to the group temporarily. It will reply with the group's chat ID — a negative number like -1001234567890.
• Remove @userinfobot after noting the ID.

Register the group in the dashboard:

1. Open the dashboard at http://localhost:3000.
2. Go to the Overview page and click Add Group.
3. Paste the chat ID and set the display name to Customer Support.
4. Click Register.

The group now appears in the Overview list with a green connected status.

Step 2 — Upload FAQ documents to the Knowledge Base

The knowledge base uses SQLite FTS5 full-text search to retrieve relevant content before each Gemini call, giving the bot accurate, grounded answers.

Prepare your FAQ document:

Create a plain text file named faq.md with content like the following:

# Frequently Asked Questions

## What are your business hours?
We are open Monday to Friday, 9 AM to 6 PM EST. We are closed on public holidays.

## How do I reset my password?
Visit https://example.com/reset and enter your email address. You will receive a reset link within 5 minutes.

## What is your refund policy?
We offer a 30-day money-back guarantee on all purchases. Contact support@example.com to initiate a refund.

## How long does shipping take?
Standard shipping takes 5–7 business days. Express shipping (2 business days) is available at checkout.

## How do I track my order?
Log in to your account and go to Order History. Each order shows a tracking link once shipped.

Upload to the knowledge base:

1. In the dashboard, go to the Knowledge page.
2. Select Customer Support from the group dropdown.
3. Click Upload Document and select your faq.md file.
4. Wait for the indexing confirmation message.

Supported formats

The knowledge base accepts .txt, .md, and .pdf files. Upload multiple documents — each is indexed separately and searched together.

Test that retrieval works:

In your Telegram group, send:

@YourBotName search the knowledge base for refund policy

The bot should return a snippet from your FAQ document.

Step 3 — Configure a professional persona

A well-written system prompt is the difference between a generic assistant and one that feels like part of your brand.

Open Memory Studio:

1. In the dashboard, go to Memory Studio.
2. Click the folder icon and select Customer Support.
3. You will see the GEMINI.md file open in the Monaco editor.

Replace the content with a custom system prompt:

# Customer Support Assistant — Acme Corp

You are a helpful and professional customer service representative for Acme Corp.

## Your role
- Answer customer questions accurately using the knowledge base
- Be polite, empathetic, and concise
- If you cannot find the answer in the knowledge base, say: "I don't have that information right now. Please email support@example.com and our team will help within 24 hours."
- Never invent information or make promises you are not certain about

## Tone
- Professional but warm
- Use the customer's name if they provide it
- Keep responses under 3 short paragraphs
- Use bullet points for lists of steps

## What you can help with
- Business hours and contact information
- Password resets and account issues
- Refund and return policy
- Shipping and order tracking

## What you cannot help with
- Billing disputes (direct to billing@example.com)
- Technical bugs (direct to bugs@example.com)

Click Save in Memory Studio.

Keep the system prompt focused

A focused persona performs better than a long list of rules. Start minimal and add detail only when the bot gives wrong answers.

Step 4 — Set up an auto-greeting with a scheduled task

A welcoming message sent once a day during business hours makes the support channel feel responsive even when no agent is active.

Create the scheduled task:

1. In the dashboard, go to the Tasks page.
2. Click New Task.
3. Fill in the form:
   • Group: Customer Support
   • Schedule: 0 9 * * 1-5 (9 AM every weekday)
   • Prompt: Send a friendly opening message to let customers know support is available today. Mention business hours and how to ask a question.
4. Click Save.

Set your timezone so the task fires at the right wall-clock time. Add this to your .env file:

TZ=America/New_York

Restart the bot after changing .env:

# Stop the running process (Ctrl+C), then:
npm run dev

One-time greeting for testing

To test immediately without waiting for the cron schedule, set the schedule to in 1 minute and watch the Telegram group. Change it back to the cron expression afterward.

Step 5 — Test with sample customer questions

With the knowledge base loaded and the persona set, test the bot with realistic customer questions.

In the Telegram group, send each of the following:

@YourBotName What are your business hours?

@YourBotName I need to return something I bought last week

@YourBotName My order hasn't arrived. How do I track it?

@YourBotName Can you help me fix a bug in your app?

What to verify:

Question	Expected behavior
Business hours	Answers from the FAQ
Return request	Explains 30-day policy, mentions email
Order tracking	Explains the Order History method
Bug report	Politely deflects to bugs@example.com

Check the Logs page

The dashboard Logs page shows each incoming message, the knowledge base snippets retrieved, and the full Gemini response in real time. Use it to understand why the bot answered the way it did.

Tips for improvement

Add more documents. Upload your full product documentation, pricing sheets, or policy PDFs. The knowledge base handles multiple documents and searches all of them simultaneously.

Use custom tools. If your CRM has an API, write a plugin that adds an get_order_status Gemini tool. The bot will call it automatically when a customer asks about a specific order. See Weather Plugin for a complete tool example.

Tune the persona over time. After a week of real conversations, review the Logs page for cases where the bot gave a wrong or unhelpful answer. Add a specific rule to GEMINI.md to address each failure pattern.

Rate limiting. The default rate limit is 20 requests per 5-minute window per group. For a busy support channel, raise this in .env:

RATE_LIMIT_MAX=50
RATE_LIMIT_WINDOW=5

Multiple support tiers. Register separate Telegram groups (e.g., General Support and Premium Support) with different knowledge bases and personas. Each group is configured independently.