Skip to content

A lightweight voice transcription bot for telegram.

License

Notifications You must be signed in to change notification settings

valamistudio/surdobot

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

38 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

GitHub Libraries.io dependency status for GitHub repo

Introduction

This is a lightweight voice transcription bot for telegram. It runs Python 3 in a serverless AWS Lambda instance through Chalice using FFMpeg, pyTelegramBotAPI and Wit.AI.

Running

mkdir ~/.aws
cat >> ~/.aws/config <<EOF
[default]
aws_access_key_id=<YOUR_ACCESS_KEY>
aws_secret_access_key=<YOUR_SECRET_KEY>
region=<YOUR_REGION> (such as us-west-2, us-west-1, etc)
EOF
git clone https://github.com/valamistudio/surdobot.git
cd surdobot/src
mkdir .chalice
cat >> .chalice/config.json <<EOF
{
  "version": "2.0",
  "app_name": "surdobot",
  "automatic_layer": true,
  "layers": ["<YOUR_FFMPEG_LAYER>"],
  "stages": {
    "dev": {
      "api_gateway_stage": "api",
      "environment_variables": {
        "bot_token": "<YOUR_BOT_TOKEN>",
        "wit_token": "<YOUR_WIT_TOKEN>",
        "user_ids": "<OPTIONAL_USER_ID_COMMA_SEPARATED_WHITELIST>"
      }
    }
  },
  "lambda_timeout": 60
}
EOF
python -m pip install pipenv
python -m pipenv install
python -m pipenv shell
chalice deploy

curl -X "POST" "https://api.telegram.org/bot<YOUR_BOT_TOKEN>/setWebhook" -d '{"url": "<REST_API_URL>/webhook"}' -H 'Content-Type: application/json; charset=utf-8'

The REST API URL comes from the chalice deploy command output, so you'll probably want to execute that last command in separate.

.chalice/config.json

  • If user_ids is assigned, the bot will only respond to private chat of users in the whitelist or to groups/supergroups/channels of which any user of the whitelist is a member. Otherwise, the bot will apply no restrictions.
  • The default value for lambda_timeout is 60. You can suppress this attribute if you want to keep it. Jobs that takes long than the set timeout will probably create an infinite message loop. Wit.AI usually takes around 20% of the audio file duration do transcribe it (i.e a 60 second file takes around 12 seconds to transcribe).

Useful links

Infinite message loop

If an operation fails to return "200 OK" (timeout, unhandled exception or whatnot), the bot will try to execute the same operation again, which will probably fail as well. This probably means that the bot will enter a infinite message loop. Apart from always returning 200, which I don't think it'd be the right call, I don't know how to fix this programmatically yet, but here's a command you can run to reset it:

curl -X "POST" "https://api.telegram.org/bot<YOUR_BOT_TOKEN>/setWebhook" -d '{"url": "<REST_API_URL>/webhook", "drop_pending_updates": true}' -H 'Content-Type: application/json; charset=utf-8'

The drop_pending_updates attribute will remove every pending request from the webhook queue. The bot token and REST API URL can be the same as the ones you used in the configuration steps, if they didn't change.