Getting Started with DB-GPT: Setup, SQL Generation, and Data Analysis

Getting Started with DB-GPT: Your First 30 Minutes

I'll be honest with you — when I first tried DB-GPT, I spent way too long fiddling with configuration files. The docs are good, but I kept running into small things that weren't immediately obvious. So I wrote this tutorial to save you the same pain. Let's get DB-GPT running, connect a database, and ask your first question — in under 30 minutes.

Step 1: Docker Setup (2 minutes)

This is the easiest way to get started — and the one I recommend. No Python environment to mess with, no dependency hell:

docker pull eosphorosai/dbgpt:latest
docker run -d --name dbgpt -p 8080:8080 \
  -v ./data/dbgpt:/data \
  eosphorosai/dbgpt:latest

Wait about 30 seconds for the container to start. If you see Listening on port 8080 in the logs, you're good:

docker logs dbgpt --tail 20

Step 2: Configure Your LLM (5 minutes)

This is where I tripped up the first time. DB-GPT needs at least one LLM configured before you can do anything useful. Open your browser to http://localhost:8080 and you'll see the setup wizard.

You have a few options:

• OpenAI — easiest. Just paste your API key and you're done. I tested with GPT-4o and it worked flawlessly
• Ollama — if you're running Ollama locally (and you should be), DB-GPT can connect to it. Point it to http://host.docker.internal:11434
• DeepSeek / Claude / Llama — all supported through the model configuration panel

I used Ollama with Llama 3.1 8B for testing, and it handled basic SQL generation well. For complex queries, GPT-4o was noticeably better — but that's expected.

Step 3: Connect a Database (3 minutes)

Go to the "Data Sources" section and add your database. DB-GPT supports PostgreSQL, MySQL, SQLite, DuckDB, ClickHouse, and more. I connected a PostgreSQL database:

Host: host.docker.internal  # Docker to host
Port: 5432
Database: my_analytics_db
Username: postgres
Password: ********

Pro tip: If your database runs on the host machine (not in Docker), use host.docker.internal as the hostname. I wasted 15 minutes wondering why localhost didn't work.

Step 4: Ask Your First Question (2 minutes)

Once the database is connected, go to the chat interface and type something like:

"Show me total revenue by product category for the last 3 months, with a bar chart"

DB-GPT will:

• Understand the intent and write the SQL query
• Execute it against your database
• Generate a bar chart from the results
• Summarize the findings in plain English

If the generated SQL isn't quite right, you can ask it to refine: "Actually, group by month too" or "Only include categories with revenue > $10,000". It iterates like a human analyst would.

Real-World Performance Numbers

Common Mistakes (That I Made So You Don't Have To)

• 🔥 Forgetting to configure the LLM first — the UI loads fine but nothing works. You'll just get "model not available" errors
• 🐳 Using localhost instead of host.docker.internal — inside the container, localhost refers to the container itself, not your host machine
• 📁 Not mounting volumes — without -v ./data/dbgpt:/data, your configuration disappears when you restart the container
• 💾 Not enough RAM — DB-GPT plus a local LLM needs at least 8GB RAM. I tested with 4GB and it was painfully slow

Conclusion

DB-GPT took me about 20 minutes from docker pull to my first working query. That's way faster than I expected for an open-source AI data platform. If you work with data regularly — even if you're not a SQL expert — this tool will change how you think about analytics. Give it a try, and if you get stuck, the community on Slack is genuinely helpful.