Fast AI Apps with SQL - MindsDB

So you know some SQL already. And want to build AI Apps fast.

You need to get to know about MindsDB, which helps us to create AI Tools that need realtime data to perform their tasks.

The MindsDB Project

You can find project details and source code at:

• The MindsDB Site
• The MindsDB Source Code at Github
  • The Docker Container to deploy MindsDB
  • Mixed License: ELv2 and MIT ❗

You can give Minds DB a try with Vector DBs, for example with your SelfHosted ChromaDB

MindsDB Integrations - ML & LLMs

There are several ways to integrate MindsDB:

• LLMs: Talking about LLMs, Free Models and locally, we are lucky to have the simplest way
  • Ollama - Setup Ollama locally with Docker
• With VectorDBs:
  • LanceDB, Pinecone, Qdrant…
• Time Series DBs:
  • InfluxDB - F/OSS and you can deploy it with Docker. Plays well with IoT and Grafana.
• And more: https://docs.mindsdb.com/integrations/data-integrations/all-data-integrations

SelfHosting MindsDB with Docker

apt-get update && sudo apt-get upgrade && curl -fsSL https://get.docker.com -o get-docker.sh
sh get-docker.sh && docker version

As per the Docs, we can run it with Docker CLI:

docker run -p 47334:47334 -p 47335:47335 mindsdb/mindsdb

But for proper SelfHosting and Docker Container Management, lets SelfHost mindsdb with docker-compose:

version: '3.9'

services:
  chroma:
    container_name: mindsdb-container
    image: mindsdb/mindsdb
    ports:
      - "47334:47334"
    volumes:
      - mindsdb_data:/mindsdb

volumes:
  mindsdb_data:
    driver: local

Then, just go to: http://localhost:47334

How to use MindsDB

Currently there are +100 Sources to use with MindsDB.

MindsDB - Web Crawler

The primary purpose of a web crawler is to collect data from the internet for various purposes, such as search engine indexing, content scraping, website analysis, and more.

With MindsDB, we can use a Web Crawler and get web data to train models, domain specific chatbots or fine-tune LLMs.

Initialize a web crawler:

CREATE DATABASE my_web 
WITH ENGINE = 'web';

Get content from a Web:

SELECT * 
FROM my_web.crawler 
WHERE url = 'docs.mindsdb.com' 
LIMIT 1;
#LIMIT 10; #10 internal pages

SELECT * 
FROM my_web.crawler 
WHERE url IN ('docs.mindsdb.com', 'docs.python.org') 
LIMIT 1;

SELECT * 
FROM my_web.crawler 
WHERE url = '<link-to-pdf-file>' 
LIMIT 1;

FAQ

Other F/OSS Ways to Check Sentiment Analysis?

• You can have a look the Detoxify Project
  • It uses Pytorch and Transformers
  • Apache v2 Licensed