Building a KB Manager using MindsDB

Pheonix Coder 🐦‍🔥 — Tue, 24 Jun 2025 05:33:18 +0000

Hey everyone,

I recently participated in Quira Quest 19. For Challenge 1, I built a basic CLI app that connects with MindsDB's Knowledge Bases and Agents.

Here's the repo:
https://github.com/pheonix-coder/kb-manager

It's a small CLI app using Typer. Under the hood, it just runs SQL queries against MindsDB. I kept it modular, so people can reuse the commands and plug them into other apps or scripts. Just enough to show how the pieces fit together in a real system.

You can do the following from the CLI:

Create a knowledge base (KB) with metadata and embedding configs
Ingest CSV data
Run semantic search queries
Summarize the KB using OpenAI (GPT-3.5)
Create a chat agent connected to the KB
Chat with the agent and get responses

All the setup steps are in the README:
https://github.com/pheonix-coder/kb-manager#️-setup

Why I Built It

Mostly to get a feel for how MindsDB actually handles knowledge and agents behind the scenes. The docs give you the high-level picture, but I wanted to run stuff directly and see the full flow: from raw data to intelligent agent responses.

I didn't want to build a fancy frontend. Just CLI + some color output and animations. Something you can fire up in a terminal and see results fast.

What This Is Useful For

You want to test the KB → Embedding → Search pipeline
You’re debugging prompts or trying different embedding models
You want to show someone how this stuff works (dev, investor, teammate)
You want to build something minimal to summarize text datasets
You just want to learn MindsDB quickly and don’t like clicking through dashboards

Realistic Use Cases

1. Internal KB from CSV/Docs
Let’s say you have a bunch of customer quotes or logs. You can just ingest them and start querying with natural language. No dashboard, no API keys to manage — just the CLI.

2. Search Your Data in Plain English
Run query-kb and type something like:
ways to live life
You’ll get results based on semantic similarity. (Like Einstein quotes.)

3. Summarize Big Datasets
Run summarize-kb on a 2,000-row CSV, and get a 2-line summary. Useful if you don’t want to read through the whole thing.

4. Agents That Actually Know Your Data
Create a chatbot that only responds based on the KB content.
So you can ask: “What did Einstein say about imagination?” — and it only pulls from your dataset.

5. Demo or Teaching Tool
This CLI is good for showing how vector search and agents work without spinning up a web app. Great for product teams, new devs, or anyone learning LLMs + vector DBs.

Demo Video

If you want to see it in action, here’s a quick demo:
https://www.youtube.com/watch?v=K-cufogVz0Q

That’s it. Not a big project, but helped me understand a lot.

If this kind of thing interests you, follow me here:
https://x.com/pheonix_coder

DEV Community: Pheonix Coder 🐦‍🔥

Building a KB Manager using MindsDB

Why I Built It

What This Is Useful For

Realistic Use Cases

Demo Video