- cached results of each sequence is zipped
- predictions inside the job folder are copied instead of symbolically linked
- timezone is added to all date time tags
- finished date is added for each finished prediction for multi-sequence jobs
- Better queue priority handling for jobs with difference sizes
- Better sequence parsing to handle bad inputs
- clean old log files periodically
- a database with finished date for each cached prediction is created
TODO:
- deleting outdated cached results
- using server side processing for showing the big table.