Taar-lite
A lightweight version of the TAAR service intended for specific deployments where a reduced feature space is available for the recommendation of addons.
Table of Contents (ToC):
How does it work?
Each specific deployment recommendation strategy is implemented through this repo, usually accessible via taar-api-lite. The individual use cases reply on modelling perfromed via use-case-specific ETL jobs hosted in python_mozetl which leverage the Telemetry, data corpora to drive a set fo recommendation choices.
Current Deployments
This is the list of the current deployments of TAAR-lite:
| Model | Description | Conditions |
|---|---|---|
| AMO gui-guid | recommends add-ons based on co-installation rate with other addons. | Sufficient installation rate of requested guid |
ETL workflow AMO guid-guid TAAR-lite
-
taar_amodump.py
- Scheduled to run daily
- Collects all listed addons by callign the AMO public API endpoint
- Applies filter returning only Firefox Web Browser Extensions
- Writes extended_addons_database.json
-
taar_amowhitelisy.py
- Scheduled to run daily, dependent on successful completion of taar_amodump.py
- Filters the addons contained in extended_addons_database.json
- removes legacy addons
- removes Web Extensions with a rating < 3.0
- removes Web Extensions uploaded less than 60 days ago
- removes Firefox Pioneer
- Writes whitelist_addons_database.json
-
taar_lite_guidguid.py
- Computes the coinstallation rate of each whitelisted addon with other whitelisted addons for a sample of Firefox clients
- Removes rare combinations of coinstallations
- writes guid_coinstallation.json
Build and run tests
WIP
