alkuaanet

Starting notes of songs. A Telegram bot calling a GraphQL API sitting on top of a DynamoDB table.

How to get things up and running:

1. Build lambda layers for Telegram bot and API
   • run py build_layers.py (see below for a more detailed description)
2. cd infra && terraform apply
3. Set SSM parameters
   • acquire Telegram bot token
   • generate a secret to use as a key for Fernet (see e.g. here)
   • set the above mentioned to env vars BOT_TOKEN and FERNET_KEY and run py set_ssm_parameters.py
4. Generate a superuser token and add it to the DynamoDB table
   • you can use generate_superuser_token.py to create the token. Create record in DynamoDB with pk=token, sk=superuser, value=<key>
   • the type of value in DDB is binary, which only accepts base64 - the above mentioned script takes this into account, and prints out a base64-encoded value
5. Set up a webhook to the bot's function url
   • check the API's URL and set it as an environment variable API_URL and run py set_webhook.py
     • you can do terraform show aws_lambda_function_url.graph_api_lambda_url to get the value
     • you could probably automate this with lambda somehow, too.
   • after this the bot should be functional!

Building lambda layers

Terraform doesn't handle lambda layer building particularly well, so there's a utility script build_layers.py which:

1. Exports dependencies from poetry
2. pip installs the dependencies to gitignored folders
3. zips the folders

This script is handy when doing version upgrades or adding new dependencies. Packaging everything with Docker might be a nicer option, though.

Local testing

Graph API

Run uvicorn graph.lib.schema:app to start a local API instance (against production DB).