See the issues too.
- Load starter
- Get channel name automatically
- Build YouTube Manager class
- Create child MySQL class
- Integrate YoutubeMysql class into the YoutubeManager class
- Use the pypi packages I have created instead of the local ones
- Find a better way to change priorities (probably add a function to push everything)
- Create the workflow for the commenter
- Roll the comments for each channel - store comments in sql table?
- Store comments in dropbox
- [Merged] Regularly backup logs files from logs/ to dropbox (for when running on Heroku) + Store errors in sql or dropbox
- Ensure code works without dropbox and emailer modules
- Create the workflow for the accumulator
- Load yt keys from Dropbox
- Add SQL scripts for creating the tables needed (automatically checks and creates on init)
- Different YT env vars for each yml
- Option to set username manually
- Test that everything works properly
- Configure Procfile and circleci config
- Update Readme
- Increase check speed the minutes before o'clock
- Optimize SQL queries
- Load comments & only before the loop to make every loop faster
- Add Video Title
- Recreate the Livestreaming module (In different private repo, will merge them at some point)
- Use multiple accounts (different api keys) to check for new comments
- Instead of retrieving the playlist id everytime, load when we need it
- Slow Mode: Reduce check time in specific times (eg midnight to 6am)
- Add option for channel to only use channel comments (when available)
- Mandatory timeout per channel
- Threading to get batches of new videos in parallel
- Make error catching more specific
- Send me email on fatal error (on later version)
- Email me if there are replies mentioning the word "bot"
- Add seconds late column and update with accumulator
- Make get_video_comments() more efficient by loading 50 comments at a time
- In add_comment() use foreign keys to update the channels table and save time
- Optimize get_next_template_comment()
- Add more tests
- For very fast lookups using Redis would be optimal but an overkill at this point