Note 👀

• Configuration methods have changed in version 3 for support multiple bots
• After version 3.1 have backward compatibility for enviromnent variables configurations on single instance.
• Recommend manually rewrite environment variables into config.yaml for multi instance.

Enhanced from the original yGuy/chatgpt-mattermost-bot

• Support for multiple bots in a single process. You no longer need to run multiple docker containers to use multiple LLMs.

  

• 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 gpt-4-vision-preview 1106.

• 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!

Configuration

These are the available options, you can set them in the config.yaml file. The filename can be changed using the CONFIG_FILE environment variable.

see: config-sample.yaml and .env-sample

Example config.yaml

# System default settings
MATTERMOST_URL: 'https://your-mattermost-url.example.com'
BOT_CONTEXT_MSG: 100
PLUGINS: image-plugin
OPENAI_MAX_TOKENS: 2000
OPENAI_TEMPERATURE: 1
MAX_PROMPT_TOKENS: 2000

bots:
  - name: '@OpenAI'
    mattermostUrl: 'https://your-mattermost-url.example.com'
    mattermostToken: 'your-mattermost-token'
    type: 'openai'
    apiKey: 'your-openai-api-key'
    apiBase: 'https://api.openai.com/v1'
    modelName: 'gpt-4o-mini'
    visionModelName: 'gpt-4v'
    imageModelName: 'dall-e-3'
    maxTokens: 16384
    temperature: 1
    maxPromptTokens: 123904
    plugins: 'image-plugin'
  - name: '@ChatGPT'
    mattermostUrl: 'https://your-mattermost-url.example.com'
    mattermostToken: 'your-mattermost-token'
    type: 'azure'
    apiKey: 'your-azure-openai-api-key'
    apiVersion: '2024-10-21'
    instanceName: 'your-azure-instance-name'
    deploymentName: 'gpt-4o-mini'
    visionKey: 'your-azure-openai-vision-key'
    visionInstanceName: 'your-azure-vision-instance-name'
    visionDeploymentName: 'gpt-4-vision-preview'
    imageKey: 'your-azure-openai-image-key'
    imageInstanceName: 'your-azure-image-instance-name'
    imageDeploymentName: 'dall-e-3'
    maxTokens: 16384
    temperature: 1
    maxPromptTokens: 123904
    plugins: 'image-plugin'
  - name: '@Gemini'
    mattermostUrl: 'https://your-mattermost-url.example.com'
    mattermostToken: 'your-mattermost-token'
    type: 'google'
    apiKey: 'your-google-api-key'
    imageModelName: 'dall-e-3'
    maxTokens: 8192
    temperature: 1
    maxPromptTokens: 1048576
    plugins: ''
  - name: '@Cohere'
    mattermostUrl: 'https://your-mattermost-url.example.com'
    mattermostToken: 'your-mattermost-token'
    type: 'cohere'
    apiKey: 'your-cohere-api-key'
    imageModelName: 'dall-e-3'
    maxTokens: 4000
    temperature: 1
    maxPromptTokens: 123904
    plugins: ''
  - name: '@Anthropic'
    mattermostUrl: 'https://your-mattermost-url.example.com'
    mattermostToken: 'your-mattermost-token'
    type: 'anthropic'
    apiKey: 'your-anthropic-api-key'
    imageModelName: 'dall-e-3'
    maxTokens: 4096
    temperature: 1
    maxPromptTokens: 123904
    plugins: ''

# Bot instructions
BOT_INSTRUCTION: "You are a helpful assistant. Whenever users asks you for help you will provide them with succinct answers formatted using Markdown. You know the user's name as it is provided within the meta data of the messages."

Configuration Options

Name	Required	Default Value	Description	Environment Variables
MATTERMOST_URL	yes	none	The URL to the Mattermost server. This is default for all bots.	MATTERMOST_URL
BOT_CONTEXT_MSG	no	100	The number of previous messages which are appended to the conversation with ChatGPT.	BOT_CONTEXT_MSG
PLUGINS	no	'image-plugin graph-plugin'	The enabled plugins of the bot. This is default for all bots. By default, all bot (graph-plugin and image-plugin) are enabled.	PLUGINS
OPENAI_MAX_TOKENS	no	2000	The maximum number of tokens to pass to the LLM's API. This is default for all bots.	OPENAI_MAX_TOKENS
OPENAI_TEMPERATURE	no	1	The sampling temperature to use, between 0 and 2. Higher values make the output more random, while lower values make it more focused and deterministic. This is default for all bots.	OPENAI_TEMPERATURE
MAX_PROMPT_TOKENS	no	2000	Maximum number of prompt tokens. This is default for all bots.	MAX_PROMPT_TOKENS
BOT_INSTRUCTION	no	'You are a helpful assistant...`	Extra instruction to give your assistance. How should the assistant behave? This setting used by all bots.	BOT_INSTRUCTION
Individual bot settings
name	yes	none	The name of the bot.
mattermostUrl	no	none	The URL to the Mattermost server for the bot.	MATTERMOST_URL
mattermostToken	yes	none	The authentication token for the Mattermost bot.	${name}_MATTERMOST_TOKEN, ${type}_MATTERMOST_TOKEN, MATTERMOST_TOKEN
type	yes	none	The type of AI provider (e.g., openai, azure, google, cohere, anthropic).
apiKey	yes	none	The API key for the AI provider.	OPENAI_API_KEY, AZURE_OPENAI_API_KEY, GOOGLE_API_KEY, COHERE_API_KEY, ANTHROPIC_API_KEY
apiBase	no	none	The base URL for the AI provider's API.	OPENAI_API_BASE
modelName	no	'gpt-4o-mini'	The name of the model to use for chat completions.	OPENAI_MODEL_NAME
visionModelName	no	none	The name of the model to use for vision tasks.
imageModelName	no	none	The name of the model to use for image generation.
apiVersion	no	'2024-10-21'	The API version to use for the AI provider.	AZURE_OPENAI_API_VERSION
instanceName	no	none	The instance name for the AI provider (specific to Azure).	AZURE_OPENAI_API_INSTANCE_NAME
deploymentName	no	none	The deployment name for the AI provider (specific to Azure).	AZURE_OPENAI_API_DEPLOYMENT_NAME
visionKey	no	none	The API key for the vision tasks (specific to Azure).	AZURE_OPENAI_API_VISION_KEY
visionInstanceName	no	none	The instance name for the vision tasks (specific to Azure).	AZURE_OPENAI_API_VISION_INSTANCE_NAME
visionDeploymentName	no	none	The deployment name for the vision tasks (specific to Azure).	AZURE_OPENAI_API_VISION_DEPLOYMENT_NAME
imageKey	no	none	The API key for the image generation tasks (specific to Azure).	AZURE_OPENAI_API_IMAGE_KEY
imageInstanceName	no	none	The instance name for the image generation tasks (specific to Azure).	AZURE_OPENAI_API_IMAGE_INSTANCE_NAME
imageDeploymentName	no	none	The deployment name for the image generation tasks (specific to Azure).	AZURE_OPENAI_API_IMAGE_DEPLOYMENT_NAME
maxTokens	no	2000	The maximum number of tokens for the AI provider.	OPENAI_MAX_TOKENS
temperature	no	1	The sampling temperature for the AI provider.	OPENAI_TEMPERATURE
maxPromptTokens	no	2000	The maximum number of prompt tokens for the AI provider.	MAX_PROMPT_TOKENS
plugins	no	'image-plugin graph-plugin'	The enabled plugins for the bot. By default, The bot graph-plugin and image-plugin are enabled.	PLUGINS

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 \
  -v /path/to/config.yaml:/app/config.yaml \
  --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 \
  -v /path/to/config.yaml:/app/config.yaml \
  --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 \
  -v /path/to/config.yaml:/app/config.yaml \
  --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 and Configuration

No environment variables need to be set, but you must provide a config.yaml file.

Optional Environment Variables

# 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)