docs: Update NEXT to reflect some wip

NEXT.md

@@ -11,24 +11,26 @@ Big plans
 
 If you wish to contribute, here is what I have in my head for this project :)
 
-**pytlas** is a library and as such does not provide any end user interface (except the REPL). I've done some projects such as the [server](https://github.com/atlassistant/pytlas-server) and [mobile](https://github.com/atlassistant/pytlas-mobile) to expose this library over websockets and connect to it using a mobile client but there's still many stuff to do.
+**pytlas** is a library and as such does not provide any end user interface (except the REPL). I've done some projects such as the [server](https://github.com/atlassistant/pytlas-server) and [mobile](https://github.com/atlassistant/pytlas-mobile) to expose this library over websockets and connect to it using a mobile client but there's still many stuff to do and is experimental.
 
-The **big plan** is to provide a `pytlas-broker` which will exposes agents via MQTT to provide a better integration with other IoT systems using simple topics listed below.
+The **big plan** is to provide a `pytlas-broker` which will exposes agents via MQTT to provide a better integration with other IoT systems using simple topics listed below. I have started to implement it and we'll release it as soon as it is fully working.
 
-This broker project will expose two entry points: `BrokerServer` and `BrokerClient`. Discovery for the MQTT server should be possible using [avahi](https://gist.github.com/marciogranzotto/20e45b83bbcca11e267708b10507c54a).
+This broker project will expose two entry points: `Server` and `Client`. Discovery for the MQTT server should be possible using [avahi](https://gist.github.com/marciogranzotto/20e45b83bbcca11e267708b10507c54a).
+
+The `Server` will handle agents, training and configurations.
 
 ### Client to Server
 
-- `atlas/<uid>/ping`: Creates an agent if needed
-- `atlas/<uid>/parse`: Ask pytlas to parse a raw message
+- `atlas/<device_id>/<uid>/ping`: Creates an agent if needed
+- `atlas/<device_id>/<uid>/parse`: Ask pytlas to parse a raw message
 
 ### Server to Client
 
-- `atlas/<uid>/pong`: Reply to ping requests when the agent is ready
-- `atlas/<uid>/ask`: Ask something to the user
-- `atlas/<uid>/answer`: Answer something to the user
-- `atlas/<uid>/context`: Switched to a particular context
-- `atlas/<uid>/thinking`: Started processing a message
-- `atlas/<uid>/done`: Stopped processing a message
+- `atlas/<device_id>/<uid>/pong`: Reply to ping requests when the agent is ready
+- `atlas/<device_id>/<uid>/ask`: Ask something to the user
+- `atlas/<device_id>/<uid>/answer`: Answer something to the user
+- `atlas/<device_id>/<uid>/context`: Switched to a particular context
+- `atlas/<device_id>/<uid>/thinking`: Started processing a message
+- `atlas/<device_id>/<uid>/done`: Stopped processing a message
 
-Once that's done, I've got some work in progress for a pod on a RPI which always listen using [snowboy](https://snowboy.kitt.ai/), do some STT using Google Cloud Speech for now (waiting for DeepSpeech...), push the transcript to pytlas and reply back using svox pico. Sources will be available soon, I just need to refactor a little bit.
+Once that's done, I've got some work in progress for a pod on a RPI which always listen using [snowboy](https://snowboy.kitt.ai/), do some STT using Google Cloud Speech for now (waiting for DeepSpeech...), push the transcript to pytlas and reply back using svox pico. Sources will be available soon, I just need to refactor it a little bit.