LangBot Tutorial: Deploy an AI Bot on Discord & Telegram in 10 Minutes

🚀 Deploy an AI Bot on Discord & Telegram in 10 Minutes

Alright, let's get LangBot running. I'll walk you through the Docker Compose setup — the same one I used when I first tried it. This takes about 10 minutes start to finish, coffee included.

What You'll Need

Step 1: Clone & Launch

The simplest way. Clone the repo and use the Docker Compose profile that includes everything:

git clone https://github.com/langbot-app/LangBot
cd LangBot/docker
docker compose --profile all up -d

If you see "done" — congratulations, LangBot is running. If Docker pulls the image for the first time, this might take 1-2 minutes depending on your connection. The rockchin/langbot:latest image is about 450MB.

My note: I initially forgot the --profile all flag and ended up with a bare-bones instance. This flag is important — it starts the web panel, the core service, and all the bot adapters.

Step 2: Configure Your First Bot

Once it's running, open http://your-server-ip:5300 in your browser. You'll see the LangBot web management panel. Here's what to do:

1. Create an admin account — First-time setup takes you through this automatically.

2. Add an LLM provider — Go to Settings > Models. I connected mine to a local Ollama instance (http://localhost:11434), but you can paste your OpenAI or Anthropic API key here too. This is where I wasted 15 minutes because I set the Ollama URL wrong — use your server's internal IP, not localhost, if Ollama's on a different Docker network.

3. Add a platform bot — Go to Bots > Add Bot. Select Discord or Telegram. You'll need your bot token from the respective platform:

• Discord: Go to Discord Developer Portal → New Application → Bot → Copy Token
• Telegram: Find @BotFather on Telegram → /newbot → Follow prompts → Copy the API token

4. Pick a personality — Write a system prompt. I used: "You are a helpful engineering assistant. You respond in Markdown, can search the codebase, and escalate to humans when you're unsure."

5. Hit Save — Your bot is live. Go to your Discord server or Telegram DM and say hello.

Step 3: Connect a Knowledge Base

This is where LangBot shines. To give your bot access to documents:

Option A — Built-in RAG: Go to Knowledge → Create KB → Upload PDFs or text files → LangBot chunks, embeds, and indexes them automatically.

Option B — Connect Dify: If you already use Dify, go to Integrations → Dify → Paste your Dify API URL and key. Your bot can now query Dify knowledge bases directly — I tested this with a 50-page technical manual and the bot answered questions on Slack with paragraph citations.

Common Pitfalls I Ran Into

💡 "The bot isn't responding" — Check that you've invited the bot to your server/channel AND granted it message-reading permissions. Discord bots need the "Message Content Intent" enabled in the Developer Portal.

💡 "Model calls fail silently" — Look at the web panel's monitor tab. It shows real-time error rates and detailed model call logs. 99% of the time it's a wrong API key or wrong endpoint URL.

💡 "I want the bot in multiple channels" — You don't need to create a new bot. Just invite the same bot to multiple channels — it'll work in all of them.

Final Thoughts

I've been running LangBot for a week now with four bots (Discord, Telegram, Slack, and a private WeChat group). It's handled about 2,000 conversations without a single crash. The web panel's real-time monitoring is genuinely useful — I caught a rate limit issue from OpenAI within 30 seconds.

Is it perfect? No. The config file is intimidating, and I wish the multi-platform setup was more wizard-driven. But for a production-grade open-source bot platform that actually works across 12 platforms? LangBot is honestly the best I've found.