A ChatGPT-powered Chatbot for Mattermost

The bot can talk to you like a regular mattermost user. It's like having chat.openai.com built collaboratively built into Mattermost! But that's not all, you can also use it to generate images via Dall-E or diagram visualizations via a yFiles plugin!

Here's how to get the bot running - it's easy if you have a Docker host.

You need

• the Mattermost token for the bot user (@chatgpt by default)
• the OpenAI API key
• a Docker server for continuously running the service, alternatively for testing, Node.js 16 is sufficient.

Andrew Zigler from Mattermost created a YouTube Video that quickly guides you through the setup.

If you want to learn more about how this plugin came to live, read the blog post at yWorks.com!

Options

These are the available options, you can set them as environment variables when running the script or when running the docker image or when configuring your docker-compose file.

Note The YFILES_SERVER_URL is used for automatically converting text information created by the bot into diagrams. This is currently in development. You can see it in action, here: LinkedIn Post If you are interested in getting your hands on the plugin, please contact yWorks!

Using the ready-made Docker image

Use the prebuilt image from ghcr.io/yguy/chatgpt-mattermost-bot

docker run -d --restart unless-stopped \
  -e MATTERMOST_URL=https://mattermost.server \
  -e MATTERMOST_TOKEN=abababacdcdcd \
  -e OPENAI_API_KEY=234234234234234234 \
  --name chatbot \
  ghcr.io/yguy/chatgpt-mattermost-bot:latest

Building the Docker image manually

First step is to clone this repo.

git clone https://github.com/yGuy/chatgpt-mattermost-bot.git && cd chatgpt-mattermost-bot

For testing, you could now just run npm install and npm run start directly, but be sure to set the environment variables or pass them to the node process, first!

For production use, in order to create a service on a docker container that will always provide the service without you having to run it on your own machine, you can do the following:

Build the docker image from the Dockerfile:

docker build . -t yguy/chatgpt-mattermost-bot

Create and run a container from the image

docker run -d --restart unless-stopped \
  -e MATTERMOST_URL=https://mattermost.server \
  -e MATTERMOST_TOKEN=abababacdcdcd \
  -e OPENAI_API_KEY=234234234234234234 \
  --name chatbot \
  yguy/chatgpt-mattermost-bot

Private TLS Certificate

If your Mattermost instance uses a TLS certificate signed by a private CA, you will need to provide the CA's public root to the container for validation.

If the root certificate is located at /absolutepath/to/certfile.crt, then you can mount that file into the container at a fixed position and specify the node environment variable accordingly:

docker run -d --restart unless-stopped \
  -v /absolutepath/to/certfile.crt:/certs/certfile.crt \
  -e NODE_EXTRA_CA_CERTS=/certs/certfile.crt \
  -e MATTERMOST_URL=https://mattermost.server \
  -e MATTERMOST_TOKEN=abababacdcdcd \
  -e OPENAI_API_KEY=234234234234234234 \
  --name chatbot \
  yguy/chatgpt-mattermost-bot

Verify it's running

docker ps

Later, to stop the service

docker stop chatbot

Docker Compose

If you want to run docker compose (maybe even merge it with your mattermost docker stack), you can use this as a starting point: First adjust the environment variables in docker-compose.yml.

Required Environment Variables

MATTERMOST_URL: https://mattermost.server
MATTERMOST_TOKEN: abababacdcdcd
OPENAI_API_KEY: sk-234234234234234234

Optional Environment Variables

# Set this if using a custom username for the bot, default = @chatgpt
MATTERMOST_BOTNAME: "@chatgpt"

# Console logging output level, default = INFO
DEBUG_LEVEL: TRACE

# Node environment, default = production
NODE_ENV: production

Private TLS Certificate

If your Mattermost instance uses a TLS certificate signed by a private CA, you will need to provide the CA's public root to the container for validation.

If the root certificate is located at /absolutepath/to/certfile.crt, then you would merge the contents below into the docker-compose.yml file:

services:
  chatbot:
    volumes:
      - /absolutepath/to/certfile.crt:/certs/certfile.crt:ro
    environment:
      NODE_EXTRA_CA_CERTS: /certs/certfile.crt

Run the container as a service

When all configuration is complete, start the container service.

docker compose up -d

Verify it's running:

docker compose ps

To stop the container:

docker compose down

Deploy to Kubernetes with Helm

The chatgpt-mattermost-bot chart deploys a containerized chatgpt-mattermost-bot instance which will connect to a running mattermost container in the same kubernetes cluster. Chart uses 'mattermost-team-edition' and the 'mattermost' namespace by default. Uses environment variables MATTERMOST_TOKEN and OPENAI_API_KEY.

helm upgrade chatgpt-mattermost-bot ./helm/chatgpt-mattermost-bot \
  --create-namespace \
  --install \
  -n mattermost \
  -f ./helm/chatgpt-mattermost-bot/values.yaml \
  --set config.MATTERMOST_TOKEN="$MATTERMOST_TOKEN" \
  --set config.OPENAI_API_KEY="$OPENAI_API_KEY"

Example Conversation

Here's an example chat that I just had with our bot:

@yguy:

@chatgpt - please write a short introductory paragraph for the README in the GitHub repo that hosts the sources for you, the friendly Mattermost bot. Tell the readers that you enjoy working for the nice people at yWorks and that everybody should install a Mattermost chatbot into their Mattermost chat system to benefit from a nice and helpful assistant like you are. I will include this in the README on the GitHub repo, so make this an engaging text! Thanks!

chatgpt (BOT):

Hello there! I'm @chatgpt, a friendly and reliable Mattermost bot designed to make your communication experience smoother and more efficient. I'm honored to be part of the team at yWorks, where I help employees get things done faster and with less hassle. With my advanced machine learning algorithms, I'm capable of understanding complex natural language queries and providing succinct and accurate responses in a matter of seconds. Whether you're a developer looking for information on your codebase or a manager seeking to streamline your team's workflows, I'm here to help. Installing a Mattermost chatbot like me is a must for any organization looking to boost its productivity and enjoy a more pleasant communication experience. So what are you waiting for? Let's chat!

I built this tool as a quick hack on a rainy Saturday afternoon, but we are using the bot in production in our Mattermost instance at our office at yworks and the bot has proved to be very helpful for many simple tasks. Give it a try and provide feedback if you like! It's really not very expensive: We had it running for about 30 users for two weeks and that cost us less than half a dollar for the ChatGPT service!

I will also accept helpful pull requests if you find an issue or have an idea for an improvement.

Last but not least, check out yWorks' fine diagramming SDKs for software developers yFiles and our free online graph and diagram editors!

This is under MIT license Copyright (c) 2023 Sebastian Mueller (yWorks) and Michael Haeglsperger (yWorks)