Bugfixes - [12722](https://github.com/rasahq/rasa/issues/12722): Intent names will not be falsely abbreviated in interactive training (fixes OSS-413).
This will also fix a bug where forced user utterances (using the regex matcher) will be reverted even though they are present in the domain. - [12886](https://github.com/rasahq/rasa/issues/12886): Cache `EndpointConfig` session object using `cached_property` decorator instead of recreating this object on every request. Initialize these connection pools for action server and model server endpoints as part of the Sanic `after_server_start` listener. Also close connection pools during Sanic `after_server_stop` listener.
3.6.10
Not secure
Improvements - [12827](https://github.com/rasahq/rasa/issues/12827): Improved handling of last batch during DIET and TED training. The last batch is discarded if it contains less than half a batch size of data. - [12852](https://github.com/rasahq/rasa/issues/12852): Added `username` to the connection parameters for `RedisLockStore` and `RedisTrackerStore` - [1493](https://github.com/rasahq/rasa/issues/1493): Telemetry data is only send for licensed users.
Improved Documentation - [12868](https://github.com/rasahq/rasa/issues/12868): Remove the Playground from docs.
3.6.9
Not secure
Improvements - [12778](https://github.com/rasahq/rasa/issues/12778): Added additional method `fingerprint_addon` to the `GraphComponent` interface to allow inclusion of external data into the fingerprint calculation of a component
Bugfixes - [12790](https://github.com/rasahq/rasa/issues/12790): Fixed `KeyError` which resulted when `domain_responses` doesn't exist as a keyword argument while using a custom action dispatcher with nlg server.
3.6.8
Not secure
No significant changes.
3.6.7
Not secure
Bugfixes - [12768](https://github.com/rasahq/rasa/issues/12768): Updated certifi, cryptography, and scipy packages to address security vulnerabilities.
3.6.6
Not secure
Bugfixes - [12755](https://github.com/rasahq/rasa/issues/12755): Updated setuptools and wheel to address security vulnerabilities.