Defines an SIO and MQ configured_personas_changed event that is emi…

…tted any time the server makes a change to the database

Leaves existing logic untouched to allow clients to request an update from the server at any time
Relates to NeonGeckoCom/neon-llm-core#8

chat_server/blueprints/personas.py

@@ -25,9 +25,12 @@
 # LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
 # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
 # SOFTWARE,  EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+import json
+
 from fastapi import APIRouter
 from starlette.responses import JSONResponse
 
+from chat_server.server_utils.api_dependencies import CurrentUserModel
 from chat_server.server_utils.enums import RequestModelType, UserRoles
 from chat_server.server_utils.http_exceptions import (
     ItemNotFoundException,
@@ -47,7 +50,7 @@
     PersonaData,
 )
 from chat_server.server_utils.api_dependencies.validators import permitted_access
-
+from chat_server.sio.server import sio
 from utils.database_utils.mongo_utils import MongoFilter, MongoLogicalOperators
 from utils.database_utils.mongo_utils.queries.wrapper import MongoDocumentsAPI
 
@@ -61,7 +64,7 @@
 async def list_personas(
     current_user: CurrentUserData,
     request_model: ListPersonasQueryModel = permitted_access(ListPersonasQueryModel),
-):
+) -> JSONResponse:
     """Lists personas matching query params"""
     filters = []
     if request_model.llms:
@@ -112,6 +115,7 @@ async def add_persona(
     if existing_model:
         raise DuplicatedItemException
     MongoDocumentsAPI.PERSONAS.add_item(data=request_model.model_dump())
+    await _notify_personas_changed()
     return KlatAPIResponse.OK
 
 
@@ -131,6 +135,7 @@ async def set_persona(
     MongoDocumentsAPI.PERSONAS.update_item(
         filters=mongo_filter, data=request_model.model_dump()
     )
+    await _notify_personas_changed()
     return KlatAPIResponse.OK
 
 
@@ -140,6 +145,7 @@ async def delete_persona(
 ):
     """Deletes persona"""
     MongoDocumentsAPI.PERSONAS.delete_item(item_id=request_model.persona_id)
+    await _notify_personas_changed()
     return KlatAPIResponse.OK
 
 
@@ -157,4 +163,13 @@ async def toggle_persona_state(
     )
     if updated_data.matched_count == 0:
         raise ItemNotFoundException
+    await _notify_personas_changed()
     return KlatAPIResponse.OK
+
+
+async def _notify_personas_changed():
+    response = await list_personas(CurrentUserModel(_id="", nickname="",
+                                                    first_name="", last_name=""),
+                                   ListPersonasQueryModel(only_enabled=True))
+    enabled_personas = json.loads(response.body.decode())
+    sio.emit("configured_personas_changed", enabled_personas)

services/klatchat_observer/controller.py

@@ -28,6 +28,8 @@
 import json
 import re
 import time
+from typing import Optional
+
 import cachetools.func
 
 from threading import Event, Timer
@@ -304,7 +306,7 @@ def get_neon_service(self, wait_timeout: int = 10) -> None:
         LOG.info("Joining sync consumer")
         sync_consumer.join()
         if not self.neon_service_event.is_set():
-            LOG.warning(f"Failed to get neon_service in {wait_timeout} seconds")
+            LOG.warning(f"Failed to get neon response in {wait_timeout} seconds")
             self.__neon_service_id = ""
 
     def register_sio_handlers(self):
@@ -327,6 +329,8 @@ def register_sio_handlers(self):
             handler=self.request_revoke_submind_ban_from_conversation,
         )
         self._sio.on("auth_expired", handler=self._handle_auth_expired)
+        self._sio.on("configured_personas_changed",
+                     handler=self._handle_personas_changed)
 
     def connect_sio(self):
         """
@@ -396,6 +400,7 @@ def get_neon_request_structure(msg_data: dict):
         if requested_skill == "tts":
             utterance = msg_data.pop("utterance", "") or msg_data.pop("text", "")
             request_dict = {
+                "msg_type": "neon.get_tts",
                 "data": {
                     "utterance": utterance,
                     "text": utterance,
@@ -404,12 +409,14 @@ def get_neon_request_structure(msg_data: dict):
             }
         elif requested_skill == "stt":
             request_dict = {
+                "msg_type": "neon.get_stt",
                 "data": {
                     "audio_data": msg_data.pop("audio_data", msg_data["message_body"]),
                 }
             }
         else:
             request_dict = {
+                "msg_type": "recognizer_loop:utterance",
                 "data": {
                     "utterances": [msg_data["message_body"]],
                 },
@@ -419,13 +426,17 @@ def get_neon_request_structure(msg_data: dict):
         return request_dict
 
     def __handle_neon_recipient(self, recipient_data: dict, msg_data: dict):
+        """
+        Handle a chat message intended for Neon.
+        """
         msg_data.setdefault("message_body", msg_data.pop("messageText", ""))
         msg_data.setdefault("message_id", msg_data.pop("messageID", ""))
         recipient_data.setdefault("context", {})
         pattern = re.compile("Neon", re.IGNORECASE)
         msg_data["message_body"] = (
-            pattern.sub("", msg_data["message_body"], 1).strip("<>@,.:|- ").capitalize()
+            pattern.sub("", msg_data["message_body"], 1).strip("<>@,.:|- \n")
         )
+        # This is really referencing an MQ endpoint (i.e. stt, tts), not a skill
         msg_data.setdefault(
             "requested_skill", recipient_data["context"].pop("service", "recognizer")
         )
@@ -774,6 +785,7 @@ def on_subminds_state(self, body: dict):
 
     @create_mq_callback()
     def on_get_configured_personas(self, body: dict):
+        # Handles request to get all defined personas
         response_data = self._fetch_persona_api(user_id=body.get("user_id"))
         response_data["items"] = [
             item
@@ -791,7 +803,7 @@ def on_get_configured_personas(self, body: dict):
         )
 
     @cachetools.func.ttl_cache(ttl=15)
-    def _fetch_persona_api(self, user_id: str) -> dict:
+    def _fetch_persona_api(self, user_id: Optional[str]) -> dict:
         query_string = self._build_persona_api_query(user_id=user_id)
         url = f"{self.server_url}/personas/list?{query_string}"
         try:
@@ -803,12 +815,25 @@ def _fetch_persona_api(self, user_id: str) -> dict:
             data = {"items": []}
         return data
 
+    def _handle_personas_changed(self, data: dict):
+        """
+        SIO handler called when configured personas are modified. This emits an
+        MQ message to allow any connected listeners to maintain a set of known
+        personas.
+        """
+        self.send_message(
+            request_data=data,
+            vhost=self.get_vhost("llm"),
+            queue="configured_personas_changed",
+            expiration=5000,
+        )
+
     def _refresh_default_persona_llms(self, data):
         for item in data["items"]:
             if default_llm := item.get("default_llm"):
                 self.default_persona_llms[item["id"]] = item["id"] + "_" + default_llm
 
-    def _build_persona_api_query(self, user_id: str) -> str:
+    def _build_persona_api_query(self, user_id: Optional[str]) -> str:
         url_query_params = f"only_enabled=true"
         if user_id:
             url_query_params += f"&user_id={user_id}"