MindsDB Demo Studio

• Author: Michael Tanenbaum
• License: Apache 2.0

This repository provides a Docker Compose file for launching a MindsDB demo environment that includes MindsDB, RedPanda (Kafka), Cassandra, and Postgres.

The intention of this demo platform is to provide a quick and reliable way to launch a MindsDB demo environment that includes a variety of data sources and sinks.

These demonstrations are intended to be delivered by technical sales/field engineering teams to prospects and customers. This is not intended for production use.

Roadmap

• Add end-to-end examples that include front-end applications so that prospects/customers can understand how projects using MindsDB should be integrated into customer operational environments and workflows. For example,
  • a simple web application that demonstrates how to use MindsDB to deliver RAG capabilities to internal teams.
  • a simple web application that demonstrates incorporating new data coming off a Kafka topic into a RAG model running on MindsDB.
  • a simple web application that demonstrates how to use MindsDB to deliver predictive maintenance capabilities to internal teams.
• Replace the current Ollama container image with one that has `` pre-loaded.

Docker Compose File

This Docker Compose file includes the following services:

• MindsDB
• Ollama
• RedPanda (single broker deployment)
• Cassandra (single node deployment)
• Postgres

Connectivity

Each service has connectivity to each other over a Docker bridge network named demo.

You can confirm this network is created by running docker network ls and looking for the demo network. You can inspect the network by running docker network inspect demo.

FQDN and Endpoints

Where possible, default ports have been preserved.

Each service can be accessed using the following Fully Qualified Domain Names (FQDN) and endpoints:

• MindsDB

  • FQDN: mindsdb.mindsdb-demo-studio_demo
  • Endpoint: http://mindsdb.mindsdb-demo-studio_demo:47334

• Ollama

  • FQDN: ollama.mindsdb-demo-studio_demo
  • Endpoint: http://ollama.mindsdb-demo-studio_demo:11434

• RedPanda

  • FQDN: redpanda.mindsdb-demo-studio_demo
  • Endpoint: http://redpanda.mindsdb-demo-studio_demo:9092

• Cassandra

  • FQDN: cassandra.mindsdb-demo-studio_demo
  • Endpoint: http://cassandra.mindsdb-demo-studio_demo:9042

• Postgres

  • FQDN: postgres.mindsdb-demo-studio_demo
  • Endpoint: http://postgres.mindsdb-demo-studio_demo:5432

Persistence

Because this is designed to be a quick way to launch demos for MindsDB, each service (cluding MindsDB) does not persist anything between container restarts.

Setup and Launch Instructions

1. Clone this repo and navigate to the repo directory
2. Run: docker-compose up -d
3. Navigate to http://localhost:47334

How to uninstall/clean up the MindsDB Docker Desktop Extension

This demo provides its own copy of the MindsDB Studio container, integrated into the Docker bridge network demo that all the other data services in this demo also use.

If you installed the MindsDB Docker Desktop Extension, you will have port conflict with the MindsDB container provided here even if Docker Desktop and docker ps -a show no running containers! Uninstalling the extension from the Docker Desktop leaves a Docker Nework service running that includes a MindsDB container bound to port 47334.

Below are the steps for removing these artifacts.

Steps

1. Uninstall the MindsDB Docker Desktop extension from Docker Desktop UI.
2. Run docker network ls - you should see output similar to the below.

➜  mindsdb-demo-studio git:(main) ✗ docker network ls
NETWORK ID     NAME                                                         DRIVER    SCOPE
f20993b36fb1   bridge                                                       bridge    local
a8de80732472   host                                                         host      local
10253c21557c   mindsdb_mindsdb-docker-extension-desktop-extension_default   bridge    local
8c703939d9fd   none                                                         null      local

3. Next, run docker network inspect <NETWORK ID> with the network ID of the mindsdb_mindsdb-docker-extension-desktop-extension_default
4. Find the ID of the container in the "Containers" section of the Docker network JSON specification for the "mindsdb_service" (in this case, the long string starting 017e below.)

        "Containers": {
            "017e000ca05b61efb19a3e5c4fab59561b5506783091f6ee0ef610d24ba70d2e": {
                "Name": "mindsdb_service",
                "EndpointID": "527a0112704aff45842b56fc83bae430ce3fce6b125eddfc73553a9e77fbc15f",
                "MacAddress": "02:42:ac:15:00:02",
                "IPv4Address": "172.21.0.2/16",
                "IPv6Address": ""
            }
        },

5. Run docker stop <Container ID>
6. Run docker rm <Container ID>
7. Run docker network rm <Network ID>
8. Confirm network was removed. Run: docker network ls.

Manual Setup Instructions

Set up Ollama

1. The Ollama container does not come pre-loaded with a model. Here, we'll use llama3.2. We must first pull the model into the Ollama container. (Note, this will take a while.)

curl http://localhost:11434/api/pull -d '{
  "name": "llama3.2"
}'

2. Next, we'll serve the model from the Ollama container.
3. Last, test that the model is available by running the following command:

curl http://localhost:11434/api/generate -d '{
  "model": "llama3.2",
  "prompt": "Answer in exactly 10 words. Why is the sky blue?",
  "stream": true 
}'

Set up MindsDB

1. Navigate to http://localhost:47334
2. Create a new ML_ENGINE and MODEL using the MindsDB Studio UI and the command below:

CREATE ML_ENGINE ollama_engine
FROM ollama;

CREATE MODEL llama_32
PREDICT completion
USING
   engine = 'ollama_engine',   -- engine name as created via CREATE ML_ENGINE
   model_name = 'llama3.2',  -- model run with 'ollama run model-name'
   ollama_serve_url = 'http://ollama.mindsdb-demo-studio_demo:11434';

3. Check on the progress of the new model's training by running the following command. The model is ready when the status column reads complete.:

select * from mindsdb.models;

4. Test the model by running the following command:

SELECT text, completion
FROM mindsdb.llama_32 
WHERE text = 'Why is the sky blue?'
USING
   prompt_template = 'Answer using maximum 15 words: {{text}}:';

Set up Postgres