meshagent-livekit 0.0.20__tar.gz → 0.0.21__tar.gz

{meshagent_livekit-0.0.20 → meshagent_livekit-0.0.21}/CHANGELOG.md

@@ -1,3 +1,6 @@
+## [0.0.21]
+- Stability
+
 ## [0.0.20]
 - Stability
 

{meshagent_livekit-0.0.20/meshagent_livekit.egg-info → meshagent_livekit-0.0.21}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: meshagent-livekit
-Version: 0.0.20
+Version: 0.0.21
 Summary: Livekit support for Meshagent
 License-Expression: Apache-2.0
 Project-URL: Documentation, https://meshagent.com
@@ -20,6 +20,6 @@ Requires-Dist: livekit-plugins-elevenlabs~=1.0.19
 Requires-Dist: livekit-plugins-playai~=1.0.19
 Requires-Dist: livekit-plugins-silero~=1.0.19
 Requires-Dist: livekit-plugins-turn-detector~=1.0.19
-Requires-Dist: meshagent-api~=0.0.20
-Requires-Dist: meshagent-tools~=0.0.20
+Requires-Dist: meshagent-api~=0.0.21
+Requires-Dist: meshagent-tools~=0.0.21
 Dynamic: license-file

{meshagent_livekit-0.0.20 → meshagent_livekit-0.0.21}/meshagent/livekit/agents/voice.py

@@ -1,7 +1,7 @@
 import logging
 import asyncio
 
-from meshagent.api import RoomMessage, ErrorResponse, JsonResponse, FileResponse, Requirement, Participant
+from meshagent.api import RoomMessage, ErrorResponse, JsonResponse, FileResponse, Requirement, Participant, JsonResponse, EmptyResponse, TextResponse
 from meshagent.api.room_server_client import RoomClient
 
 from meshagent.agents import ToolResponseAdapter
@@ -9,12 +9,17 @@ from meshagent.tools import ToolContext, Toolkit
 from livekit.agents import Agent, AgentSession
 from livekit.agents.llm import RawFunctionTool, ToolError, function_tool
 
+from livekit.agents import BackgroundAudioPlayer, AudioConfig, BuiltinAudioClip
+
+from typing import Annotated
 from livekit.plugins import openai, silero
 #from livekit.plugins.turn_detector.multilingual import MultilingualModel
 import uuid
 import asyncio
 import logging
 
+import os
+
 import json 
 
 from typing import Any
@@ -23,7 +28,7 @@ from livekit.plugins import openai
 
 from livekit.plugins import openai, silero
 from livekit import rtc
-from livekit.agents import Agent, AgentSession
+from livekit.agents import Agent, AgentSession, RunContext
 
 from typing import Optional
 
@@ -33,10 +38,40 @@ from meshagent.api.schema_util import merge, prompt_schema
 
 from meshagent.agents import SingleRoomAgent
 
+from livekit.plugins.turn_detector.multilingual import MultilingualModel
+
+import re
 
 logger = logging.getLogger("voice")
 
 
+def _replace_non_matching(text: str, allowed_chars: str, replacement: str) -> str:
+    """
+    Replaces every character in `text` that does not match the given
+    `allowed_chars` regex set with `replacement`.
+    
+    Parameters:
+    -----------
+    text : str
+        The input string on which the replacement is to be done.
+    allowed_chars : str
+        A string defining the set of allowed characters (part of a character set).
+        For example, "a-zA-Z0-9" will keep only letters and digits.
+    replacement : str
+        The string to replace non-matching characters with.
+        
+    Returns:
+    --------
+    str
+        A new string where all characters not in `allowed_chars` are replaced.
+    """
+    # Build a regex that matches any character NOT in allowed_chars
+    pattern = rf"[^{allowed_chars}]"
+    return re.sub(pattern, replacement, text)
+
+def safe_tool_name(name: str):
+    return _replace_non_matching(name, "a-zA-Z0-9_-", "_")
+
 class VoiceConnection:
     def __init__(self, *, room: RoomClient, breakout_room: str):
         self.room = room
@@ -166,13 +201,14 @@ class Voicebot(SingleRoomAgent):
         self, toolkits: list[Toolkit], context: ToolContext, name: str, description: str | None, input_schema: dict
     ) -> RawFunctionTool:
         
+        name = safe_tool_name(name)
         async def _tool_called(raw_arguments: dict) -> Any:
             try:
 
                 tool = None
                 for toolkit in toolkits:
                     for t in toolkit.tools:
-                        if t.name == name:
+                        if safe_tool_name(t.name) == name:
                             tool = t
 
                 if tool is None:
@@ -180,8 +216,12 @@ class Voicebot(SingleRoomAgent):
                     f"Could not find tool {name}"
                     )
 
-                tool_result = await tool.execute(context, **raw_arguments)
-
+                try:
+                    logger.info(f"executing tool {name}: {raw_arguments}")
+                    tool_result = await tool.execute(context=context, **raw_arguments)
+                except Exception as e:
+                    logger.error(f"failed to call tool {tool.name}: {e}")
+                    return ToolError("f{e}")
                 if self.tool_adapter == None:
 
                     if isinstance(tool_result, ErrorResponse):
@@ -189,7 +229,17 @@ class Voicebot(SingleRoomAgent):
 
                     if isinstance(tool_result, JsonResponse):
                         return json.dumps(tool_result.json)
+                    
+                    if isinstance(tool_result, TextResponse):
+                        return tool_result.text
         
+                    if isinstance(tool_result, EmptyResponse):
+                        return "success"
+                    
+                    if tool_result == None:
+                        return "success"
+                    
+
                     raise ToolError(
                         f"Tool '{name}' returned an unexpected result {type(tool_result)}, attach a tool response adapter"
                     )
@@ -197,6 +247,8 @@ class Voicebot(SingleRoomAgent):
                 else:
                     
                     text = await self.tool_adapter.to_plain_text(room=context.room, response=tool_result)
+                    if text == None:
+                        text = "success"
                     return text
 
             except Exception as e:
@@ -206,15 +258,26 @@ class Voicebot(SingleRoomAgent):
 
         return function_tool(
             _tool_called,
-            raw_schema={"name": name, "description": description, "parameters": input_schema},
+            raw_schema={"name": name, "description": description, "strict" : True, "parameters": input_schema},
         )
 
-    async def create_agent(self, *, context: ToolContext):
-        return  Agent(
+    async def create_agent(self, *, context: ToolContext, session: AgentSession):
+        
+        @function_tool
+        async def say(context: RunContext, text: str):
+            "says something out loud to the user"
+            logger.info(f"saying: {text}")
+            session.say(text)
+            return "success"
+
+        return Agent(
             instructions="\n".join(self.rules),
             allow_interruptions=True,
-            tools=await self.make_function_tools(context=context)
-        )     
+            tools=[
+                *await self.make_function_tools(context=context),
+                say
+            ]
+        )
 
         # agent = Agent(
         #    instructions="""
@@ -230,17 +293,14 @@ class Voicebot(SingleRoomAgent):
     def create_session(self) -> AgentSession:
 
         session = AgentSession(
+            max_tool_steps=50,
             allow_interruptions=True,
             vad=silero.VAD.load(),
             stt=openai.STT(),
             tts=openai.TTS(voice="echo"),
-            llm=openai.realtime.RealtimeModel(
-                # it's necessary to turn off turn detection in the Realtime API in order to use
-                # LiveKit's turn detection model
-                voice="alloy",
-                turn_detection=None,
-                input_audio_transcription=None,
-            ),
+            llm=openai.LLM(),
+            #turn_detection=MultilingualModel(),
+
         )
         return session
     
@@ -251,12 +311,23 @@ class Voicebot(SingleRoomAgent):
             
             logger.info("starting voice agent")
 
+            session = self.create_session()
+
             agent = await self.create_agent(context=ToolContext(
                 room=self.room,
                 caller=self.room.local_participant,
                 on_behalf_of=participant
-            ))
-            session = self.create_session()
+            ), session=session)
+           
+            background_audio = BackgroundAudioPlayer(
+                thinking_sound=[
+                    #AudioConfig(
+                    #    os.path.dirname(os.path.abspath(__file__)) +"/sfx/thinking.mp3", volume=0.2),
+                    AudioConfig(BuiltinAudioClip.KEYBOARD_TYPING, volume=0.3),
+                    AudioConfig(BuiltinAudioClip.KEYBOARD_TYPING2, volume=0.4),
+                ],
+            )
+            await background_audio.start(room=connection.livekit_room, agent_session=session)
 
             await session.start(agent=agent, room=connection.livekit_room)
 

meshagent_livekit-0.0.21/meshagent/livekit/version.py

@@ -0,0 +1 @@
+__version__ = "0.0.21"

{meshagent_livekit-0.0.20 → meshagent_livekit-0.0.21/meshagent_livekit.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: meshagent-livekit
-Version: 0.0.20
+Version: 0.0.21
 Summary: Livekit support for Meshagent
 License-Expression: Apache-2.0
 Project-URL: Documentation, https://meshagent.com
@@ -20,6 +20,6 @@ Requires-Dist: livekit-plugins-elevenlabs~=1.0.19
 Requires-Dist: livekit-plugins-playai~=1.0.19
 Requires-Dist: livekit-plugins-silero~=1.0.19
 Requires-Dist: livekit-plugins-turn-detector~=1.0.19
-Requires-Dist: meshagent-api~=0.0.20
-Requires-Dist: meshagent-tools~=0.0.20
+Requires-Dist: meshagent-api~=0.0.21
+Requires-Dist: meshagent-tools~=0.0.21
 Dynamic: license-file

{meshagent_livekit-0.0.20 → meshagent_livekit-0.0.21}/meshagent_livekit.egg-info/requires.txt

@@ -9,5 +9,5 @@ livekit-plugins-elevenlabs~=1.0.19
 livekit-plugins-playai~=1.0.19
 livekit-plugins-silero~=1.0.19
 livekit-plugins-turn-detector~=1.0.19
-meshagent-api~=0.0.20
-meshagent-tools~=0.0.20
+meshagent-api~=0.0.21
+meshagent-tools~=0.0.21

{meshagent_livekit-0.0.20 → meshagent_livekit-0.0.21}/pyproject.toml

@@ -21,8 +21,8 @@ dependencies = [
   "livekit-plugins-playai~=1.0.19",
   "livekit-plugins-silero~=1.0.19",
   "livekit-plugins-turn-detector~=1.0.19",
-  "meshagent-api~=0.0.20",
-  "meshagent-tools~=0.0.20"
+  "meshagent-api~=0.0.21",
+  "meshagent-tools~=0.0.21"
 ]
 
 [project.urls]

meshagent_livekit-0.0.20/meshagent/livekit/version.py

@@ -1 +0,0 @@
-__version__ = "0.0.20"