Enhanced from the original yGuy/chatgpt-mattermost-bot

• Experimental support Cohere Command R+ that input is text only and Google Gemini 1.5 Pro API that can input text and image
• Support GPT-4V Vision API
  • Realization of multimodal text and images in conjunction with image plugins. Limitation: The plugins are not possible in threads that have been image attached, Under the 1106 gpt-4-vision-vewview.
• No mention required in Direct Message
• Support Azure OpenAI API
  • Use the original OpenAI for image generation even when using Azure OpenAI API
• Build enhancement (The original is now in TypeScript as well by version 2.0.)
  • Formatted by Prettier
  • Lint by eslint
  • Build by esbuild
  • SWC for debug
• Token-count-based conversation thread management (MessageCollectPlugin may interfere)
• Splitting message that are too long
• Support GitLab AutoDevOps by test dummy

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.

Name	Required	Example Value	Description
MATTERMOST_URL	yes	https://mattermost.server	The URL to the server. This is used for connecting the bot to the Mattermost API
MATTERMOST_TOKEN	yes	abababacdcdcd	The authentication token from the logged in mattermost bot
OPENAI_API_KEY	yes	sk-234234234234234234	The OpenAI API key to authenticate with OpenAI
OPENAI_API_BASE	no	http://example.com:8080/v1	The address of an OpenAI compatible API. Overrides the default base path (https://api.openai.com)
OPENAI_MODEL_NAME	no	gpt-3.5-turbo	The OpenAI or Cohere language model to use, defaults to gpt-3.5-turbo
OPENAI_MAX_TOKENS	no	2000	The maximum number of tokens to pass to the OpenAI API, defaults to 2000
OPENAI_TEMPERATURE	no	0.2	The sampling temperature to use, between 0 and 2, defaults to 1. Higher values like 0.8 will make the output more random, while lower values like 0.2 will make it more focused and deterministic.
MAX_PROMPT_TOKENS	no	2096	Maximum number of prompt tokens, default to 2000. For example on GPT-4 = 8196 - OPENAI_MAX_TOKENS.
OPENAI_VISION_MODEL_NAME	no	gpt-4-vision-preview	The OpenAI VISION model to use, defaults to use OPENAI_MODEL_NAME. Use this when you want to change the model when attaching an image.
OPENAI_IMAGE_MODEL_NAME	no	dall-e-2	The OpenAI IMAGE model to use, defaults to dall-e-3
AZURE_OPENAI_API_KEY	no	0123456789abcdefghijklmno	The Azure OpenAI Service API key to authenticate. If OPENAI_API_KEY is also set, the original OpenAI is used for vision or image generation.
AZURE_OPENAI_API_INSTANCE_NAME	no	example-name	The instance name on the Azure OpenAI Service
AZURE_OPENAI_API_DEPLOYMENT_NAME	no	gpt-35-turbo	The name of the deployed model on the Azure OpenAI Service
AZURE_OPENAI_API_VERSION	no	2024-03-01-preview	The Azure OpenAI version
AZURE_OPENAI_API_VISION_KEY	no	0123456789abcdefghijklmno	The Azure OpenAI Service API key to authoenticate. If this is set, Azure OpenAI is used for vision.
AZURE_OPENAI_API_VISION_INSTANCE_NAME	no	example-name	The instance name on the Azure OpenAI Service for Vision API if specially needed
AZURE_OPENAI_API_VISION_DEPLOYMENT_NAME	no	gpt-4v	The name of the deployed model on the Azure OpenAI Service for Vision API if specially needed
AZURE_OPENAI_API_IMAGE_KEY	no	0123456789abcdefghijklmno	The Azure OpenAI Service API key to authoenticate. If this is set, Azure OpenAI is used for image generation.
AZURE_OPENAI_API_IMAGE_INSTANCE_NAME	no	example-name	The instance name on the Azure OpenAI Service for Image API if specially needed
AZURE_OPENAI_API_IMAGE_DEPLOYMENT_NAME	no	Dalle3	The name of the deployed model on the Azure OpenAI Service for Image API if specially needed. DALL-E 3 require 2023-12-01-preview API Version at least
COHERE_API_KEY	no	0123456789abcdefghijklmno	The Cohere API key to authenticate. If OPENAI_API_KEY is also set, the original OpenAI is used for vision or image generation.
GOOGLE_API_KEY	no	0123456789abcdefghijklmno	The Gemini API key to authenticate. If OPENAI_API_KEY is also set, the original OpenAI is used for vision or image generation. Tested model is only 'gemini-1.5-pro-latest''
YFILES_SERVER_URL	no	http://localhost:3835	The URL to the yFiles graph service for embedding auto-generated diagrams.
NODE_EXTRA_CA_CERTS	no	/file/to/cert.crt	a link to a certificate file to pass to node.js for authenticating self-signed certificates
MATTERMOST_BOTNAME	no	"@chatgpt"	the name of the bot user in Mattermost, defaults to '@chatgpt'
PLUGINS	no	graph-plugin, image-plugin	The enabled plugins of the bot. By default all plugins (grpah-plugin and image-plugin) are enabled.
DEBUG_LEVEL	no	TRACE	a debug level used for logging activity, defaults to INFO
BOT_CONTEXT_MSG	no	15	The number of previous messages which are appended to the conversation with ChatGPT, defaults to 100
BOT_INSTRUCTION	no	Act like Elon Musk	Extra instruction to give your assistance. How should the assistant behave?

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 gitlab.on-o.com/docker/chatgpt-mattermost-bot/release

docker run -d --restart unless-stopped \
  -e MATTERMOST_URL=https://mattermost.server \
  -e MATTERMOST_TOKEN=abababacdcdcd \
  -e OPENAI_API_KEY=234234234234234234 \
  --name chatbot \
  gitlab.on-o.com/docker/chatgpt-mattermost-bot/release:latest

As Azure OpenAI Service case

docker run -d --restart unless-stopped \
  -e MATTERMOST_URL=https://mattermost.server \
  -e MATTERMOST_TOKEN=abababacdcdcd \
  -e AZURE_OPENAI_API_KEY=234234234234234234 \
  -e AZURE_OPENAI_API_INSTANCE_NAME=example-name \
  --name chatbot \
  gitlab.on-o.com/docker/chatgpt-mattermost-bot/release:latest

Building the Docker image manually

First step is to clone this repo.

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

For testing, you could now just run npm ci and npm run start or 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 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 \
  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 \
  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 .env copy from .env-sample.

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 compose.yaml 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, build and start the container service.

Build docker image:

docker compose build

Start container:

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)