GitHub - navicstein/meetbots: A Google Meet & Zoom bots to record audio &am...

This project is far from perfect but only showcases the possibility of recording audio & video streams from a google meeting or zoom call without much users interactivity, kinda like a personal assistant

The idea is, when a user want a bot to join any call, he opens a modal & inputs the url(meet) or meetingId & passcode (zoom) then the server creates an ECS task that runs only on the bot on a docker

Features

• Be able to join meeting on behalf of a user
• Be able to record audio & video streams from meetings
• Ability to scale on AWS fargate (deployment via pulumi)
• S3 bucket upload support (soon)

This project has 3 main services

• Frontend (react app)
• Bot agent
• Graphql server

Screenshots

Pre record UI

Recording meet agent

Recording zoom agent

Running the graphql-server

the server is built with gqlgen, schema first graphql approach

cd into ./api/gql-server & run

$ BOT_MODE=remote go run .

or create an .env file

This should give you a playground url you can interact with the bot without the frontend

DBG server.go:37 > connect to http://localhost:8080/ for GraphQL playground

Testing the bot alone without server

to see the bot in action create an .env in the ./api/bots directory, either of a meet or zoom type for example

# for google Meet
ID="random-id"
NAME="Bot username"
URL=https://meet.google.com/rck-cbnr-smb
TYPE=meet

# for zoom
ID="you-id"
PASSCODE=538Eah
MEETINGID=88975667556
NAME="Bot username"
TYPE=zoom

then from that same api/bot directory, run

$ go run .

This will run quickly send the bot to the meeting & you should see the browser, this is done by ECS RunTask in gql-server on production

basically, the goal is to run a single bot on one container, why? if two bots are recording in the same container & using the same CPU, the audio will appear to be stuttered and affect audio quality

build and check the bot docker locally

from the API directory

$ docker build -t meetbot . -f bot.Dockerfile

Deployment

only deployed for testing!
Deployment is done via Pulumi & can be accessed via ./api/deployment refer to pulumi docs for more info

limitations

• Recorded videos only lives on filesystem (S3 support comming soon)
• performance: the bot may not always work smoothly and may experience delays or errors when joining or recording meetings.
• compatibility: zoom agent doesnt work in certain deployed locations, see zooms supported locations