TAAR API-lite

Instructions for development

The only special thing you will need to do is to export your AWS keys as enviroment variables to run the dev server. The docker-compose.yml file specifies two keys AWS_SECRET_ACCESS_KEY and AWS_ACCESS_KEY_ID. These keys are required to access the S3 buckets which host the data models used by TAARlite.

0. Make sure you have docker and docker-compose
1. export AWS_SECRET_ACCESS_KEY=
2. export AWS_ACCESS_KEY_ID=
3. make up

Instructions for deployment

The target environment for this project follows the dockerflow conventions. In order to run it correctly, a number of environment variables needs to be set up. The full list of variables can be found in the web section of the docker-compose.yml file. From a services standpoint, this project requires:

• an optional Redis cache service, defined by CACHE_URL
• redis cache expiry in seconds defined by TAAR_CACHE_EXPIRY

Updating Taar

The core of taar-service lives in the mozilla-taarlite python package. These are the steps required to deploy a new version of mozilla-taarlite on taar-api-lite using hashin:

From the root of the repository:

hashin mozilla-taarlite==<major>.<minor>.<patch>

Then open a pull request with the changes to requirements.txt. Once it's merged to master, the service will automatically update.

Web API

The dev service is hosted at: htps://taar.dev.mozaws.net

You must include a single GUID as the last segment of a URL encoded path.

The TAARlite service currently exposes a single URL that can be called with HTTP GET on the URL path :

    /taarlite/api/v1/addon_recommendations/<your_guid_here>/

A sample invocation for the GUID {some_addon_guid} using curl would look like this:

curl https://taar.dev.mozaws.net/taarlite/api/v1/addon_recommendations/%7Bsome_addon_guid%7D/

You should get JSON output that appears in this format:

{"results": ["guid1", "guid2", "guid3", "guid4"]}

You may specify an optional normalization mode by passing in a CGI argument normalize using one of the following choices:

• rownorm_sum
• row_count
• row_sum

Generally - you should get the best results using rownorm_sum.

Passing in an invalid normalization type will yield an empty results list.

Example :

curl http://localhost:8000/taarlite/api/v1/addon_recommendations/%7Bsome_addon_guid%7D/?normalize=rownorm_sum