Changelog
Last updated
Was this helpful?
Error 500 changes for 400 when the body of /activate isn't well formated
Add GRPC communication between API proxy and API backend, with fallback on Lambda engine for infinite scalability
Asynchronous and batching engine for use case activation & cache assignments
Add 1mn based HTTP request caching (using body, path, query & headers as cache key)
Add bucket allocation decision logic
Add extra account and environment information on the Decision API response
Add retries on synchronization API to handle too many requests on CDN invalidation
Manage timestamp on context events to handle hit delays for the data collect
Increase DynamoDB timeouts to handle throttling issues
Remove monthly active users quota management as it is done on the API proxy
Fix 1 visitor 1 experiment mode for single campaign calls
Murmur hash allocation improvements in bucketing mode.
When retrieving a single campaign assignment, returns HTTP response code 204: No Content instead of 400: Bad Request if targeting does not match
API-Key is now needed to call the Decision API
Global distribution via eight new regional endpoints in Tokyo, Seoul, Sydney, Singapore, Frankfurt, Ireland, Oregon, Ohio, and Northern Virginia
Error messages format change:
JSON object with “message” field
/activate: if one of the required fields is missing
field: Field is mandatory
/campaigns: if there is a variation assignment error (e.g. traffic allocation does not add up to 100%)
Error on allocation: visitor untracked
/campaigns: If clientId not found
Client %s not found. If the request is malformed (i.e. proto unmarshal fails), returns the proto error
/campaigns: If the campaignId (custom or generated) is not found
The campaign %s is paused or doesn’t exist. Verify your customId or campaignId.
Last updated
Was this helpful?
Was this helpful?

