Talk to ChatGPT via any Matrix client!
Go to file
bertybuttface a03851ab52
Merge pull request #276 from matrixgpt/rc
Roll a release
2024-02-21 14:44:24 +00:00
.github Update release.yml 2023-10-27 19:02:09 +01:00
img Add example screenshot 2022-12-09 21:09:14 +11:00
src Allow threaded query without prefix, fixes #133 2023-11-09 00:53:15 +08:00
.env.example add information about token size parameters 2023-11-07 10:18:30 +01:00
.gitignore Initial commit 2022-12-09 20:41:17 +11:00
.nvmrc Update and simplify Docker 2023-01-08 16:24:49 +00:00
Dockerfile Update Docker node and bot-sdk 2023-10-27 16:00:34 +01:00
LICENSE Initial commit 2022-12-09 20:41:17 +11:00 Update FAQ in 2023-07-15 17:33:36 +02:00
package.json Bump version 2024-02-21 14:43:26 +00:00
tsconfig.json Initial commit 2022-12-09 20:41:17 +11:00
yarn.lock Merge pull request #274 from matrixgpt/dependabot/npm_and_yarn/ip-2.0.1 2024-02-21 14:41:38 +00:00

Matrix ChatGPT Bot

Talk to ChatGPT via any Matrix client!

Screenshot of Element iOS app showing conversation with bot

A Matrix bot that uses waylaidwanderer/node-chatgpt-api to access the official ChatGPT API.

Warning for users upgrading from version 2.x

OpenAI released the official API for ChatGPT. Thus, we no longer have to use any older models or any models which kept on being turned off by OpenAI. This means the bot is now way more stable and way faster. However, please note: The usage of the API is no longer free. If you use this bot, your OpenAI account will be charged! You might want to limit your budget in your account using the OpenAI website. You need to remove the CHATGPT_MODEL variable from your environment, if you changed the value.


  1. Create a room
  2. Add the bot
  3. Start chatting.


  • Shows typing indicator as ChatGPT is thinking!
  • Supports encryption
  • Stores context for ChatGPT conversations


Create a copy of the example .env file

cp .env.example .env

Adjust all required settings in the .env file before running. Optional settings can also be adjusted later.



  • You need a Matrix account on (or any other server) for the bot user.
  • By default, anyone that knows the name of your bot can invite it to rooms or chat with it.
  • When using a self-hosted setup, you could wildcard all your users with MATRIX_WHITELIST=:yourhomeserver.example.

OpenAI / ChatGPT

  • You need to have an account at Please note that the usage of the ChatGPT-API is not free.
  • Create a API Key. Then, set OPENAI_API_KEY in your .env file
  • You can change the chat-model by setting the CHATGPT_API_MODEL in your .env file. ChatGPT is the gpt-3.5-turbo-model which is the default. Please note that depending on the model your OpenAI account will be charged.
  • You can change the API-URL to use another base than the official OpenAI-endpoint. To do so, change the CHATGPT_REVERSE_PROXY in the .env file.


  • Set MATRIX_BOT_PASSWORD (you can remove this later if you want)
  • Run the app using one of the methods below.
  • Copy MATRIX_ACCESS_TOKEN from the output.

Note: Doing any of the following can cause issues with encryption later on:

  • Using this token with any other client.
  • Using an access token extracted via Element.
  • Deleting the storage folder.
  • Switching between environments (e.g. Docker or no Docker)


There are multiple ways to run this bot. The easiest way is to run it within docker.

with Docker

This is the recommended way to run this project. It will use the latest stable release.

docker run -it -v storage:/storage --env-file=./.env --name matrix-chatgpt-bot

or to build locally from the latest unstable release (only do this if you have a good reason):

docker build . -t matrix-chatgpt-bot
docker run -it -v storage:/storage --env-file=./.env --name matrix-chatgpt-bot matrix-chatgpt-bot

Note: Without -it flags in the command above you won't be able to stop the container using Ctrl-C

Note: In order to see the output of your console you need to run docker logs matrix-chatgpt-bot

with Docker Compose

If you prefer you can use a docker-compose file. Copy the content below and save it in a file named docker-compose.yml. Either with a self-build image (run docker build . -t matrix-chatgpt-bot from your local git repo location) or with the latest stable pre-built release from this repo (the recommended way).

The script will look for the .env file in the same folder as the docker-compose.yml. The key storage folder storage will be created in the same folder as well. Adjust the locations to your needs.

  version: '3.7'
      container_name: matrix-chatgpt-bot 
      image: ## change to "matrix-chatgpt-bot" if you want to use your self-build image 
        - ./storage:/storage
        - ./.env

without Docker

Important: It is strongly recommended to run this package under Docker to not need to install various dependencies manually. Nevertheless, you can also run it by using the package manager yarn (get it via apt install -y yarn). You might also need to have a newer version of Node.js and other missing packages.

  • yarn
  • yarn build
  • yarn start

in Development

You only need to do this if you want to contribute code to this package.

  • Run yarn
  • Run yarn build

Good to know

  • By default "storage"-folder contains all your encryption keys. If you delete it, you will loose access to all your encrypted messages.
  • You can use a Keyv storage backend for persistence if you prefer (advanced).
  • The bot uses threads by default, to keep the context you should reply to this thread or the bot will think its a new conversation. "Threads" were previously experimental, you may need to activate them in your client's settings (e.g. in Element in the "lab"-section).
  • There is support to set the context to work at either the:
    • room level
    • thread level
    • both (threads fork the conversation from the main room)


I get "[Error: decryption failed because the room key is missing]"

Encryption works great with this package but can sometimes be a bit sensitive. Following steps can help to solve the "encryption" error

  • Don't use a MATRIX_ACCESS_TOKEN extracted via Element-App, use the generated token from the bot based on your MATRIX_BOT_USERNAME & MATRIX_BOT_PASSWORD set in the envfile. It will be visible in the console at start up if the MATRIX_ACCESS_TOKEN is not already set:
  1. Remove the MATRIX_ACCESS_TOKEN from the env file and make sure MATRIX_BOT_USERNAME & MATRIX_BOT_PASSWORD are set
  2. Re-run the bot
  3. Copy the token from console output to your env file
  4. Restart the bot again.
  • If all fails, you can always reset your key storage. It's important to exercise all of the following steps, because any remaining data could lead to the next encryption error. Once everything is working, make sure to not touch the "storage" folder anymore:
  1. Stop the bot
  2. Delete the "storage" folder
  3. Delete all user data for the matrix bot account (e.g. use Synapse-Admin) or create a fresh bot user account (you can then skip step 4)
  4. Log into your bot account (e.g. via Element) and log out of all sessions
  5. Verify the correctness of your env file and then run the bot setup again (e.g. via docker-compose up if you use docker-compose).

I want to chat with the bot without dealing with encryption problems

  • Set MATRIX_ENCRYPTION=false in your env-file and restart the bot. If it previously was running with encryption switched on, you need to create a new room with the bot as encryption can't be switched off once it was activated.

I get "{ errcode: 'M_NOT_FOUND', error: 'Event not found.' }"

  • So far, its not known to cause issues, you can safely ignore it.

How do I know that the bot is running succesfully?

Once the bot has started succesfully, it will output the following information to your console.

  • [INFO] [index] Starting bot...
  • [INFO] [MatrixClientLite] End-to-end encryption enabled ## this depends on your setup
  • [INFO] [index] Bot started!

I use Docker but I don't see any console output

You most likely need to view the logs by running docker logs matrix-chatgpt-bot

How to set the temperature

Set the temperature by using CHATGPT_TEMPERATURE in your .env file. The default is 0.8.

Here are some guidelines for setting the temperature:

Temperature Values Appropriate Tasks Examples
Below 0.5 (low) Tasks requiring a single correct answer or predictable output Programming
0.5-0.9 (medium) Tasks needing somewhat varied and creative content grounded in reality E-mail response
Above 0.9 (high) Tasks requiring more creative and unpredictable output Story writing

The response I receive is excessively long and gets truncated. Unlike in ChatGPT, my matrix does not have a "Continue generating" button.

You can simply write continue to prompt the bot to continue from its previous answer.

Reporting issues

Please report issues via Github. The chat room is for discussion.

Please use the search on Github and Matrix before asking for support.


Join with any Matrix chat client or on the web!

If you've never set up a Matrix client before you can follow the prompts to get started.