Where is RunContext used with agents?

Thanks for reaching out, Arthur. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

What attributes does a RunContext object have?

Hey there! Just wanted to share a quick rundown of the

RunContext

object in LiveKit and what it offers: • `session`: This gives you access to the current

AgentSession

, so you can see all the details about what's happening right now. • `speech_handle`: This is your go-to for anything speech-related, like managing audio playback or handling interruptions. • `function_call`: Here you'll find the

FunctionCall

object tied to the tool function that's running. • `userdata`: This stores user-specific info linked to the session, which you can get through

session.userdata

. On top of that, there are some handy methods to interact with: • `disallow_interruptions()`: Use this if you want to block any interruptions during the current function call. • `wait_for_playout()`: This one waits for the speech playback to finish for the current step in the function call. Sources: livekit-agents/livekit/agents/telemetry/trace_types.py

difference between chat context and runcontext

Let’s break down the difference between

ChatContext

and

RunContext

in LiveKit, so it’s easy to understand: • Purpose: ◦

ChatContext

keeps track of the conversation history — that means messages, function calls, and their results — helping the chat flow smoothly and make sense. ◦

RunContext

is all about the behind-the-scenes details during a function’s execution, like session info, function calls, and user data. • Components: ◦

ChatContext

includes handy classes like

ChatMessage

,

FunctionCall

, and

FunctionCallOutput

to organize everything that’s happened in the chat. ◦

RunContext

gives you access to the current session, user-specific info, and details about the runtime environment. • Functionality: ◦

ChatContext

helps maintain the conversation history and lets you add messages or function calls to keep interactions clear and connected. ◦

RunContext

comes into play during function execution, providing the data needed to make those operations context-aware. • Usage: ◦

ChatContext

is your go-to when chatting with the LLM to manage the state of the conversation. ◦

RunContext

is used inside function tools to tap into session and runtime details. So, to sum it up:

ChatContext

is all about keeping the conversation flowing nicely, while

RunContext

focuses on what’s happening during function execution behind the scenes.

what is the userdata attribute in RunContext

Hey there! The

userdata

attribute in

RunContext

is a handy way to store and manage custom, session-specific data within a LiveKit agent. Here’s a quick rundown: • Purpose: It keeps track of application-specific info relevant to the current session, like user inputs, preferences, or any intermediate results. • Access: You can get to it via

RunContext.userdata

, and its type is set as a generic parameter when you initialize it. • Initialization: It gets initialized when you create an

AgentSession

, with proper type hints to keep things clear. For example:`session = AgentSession[MySessionInfo]( userdata=MySessionInfo(), # ... other initializations )` • Usage: It’s great for passing state between function tools and event handlers. Any changes you make to

userdata

will stick around for the whole session. • *Example:*`@dataclass class MySessionInfo: user_name: str | None = None age: int | None = None RunContext_T = RunContext[MySessionInfo] @function_tool() async def record_name( name: Annotated[str, Field(description="The customer's name")], context: RunContext_T, ) -> str: userdata = context.userdata userdata.user_name = name return f"The name is updated to {name}"` If you want, I’m here to help you implement

userdata

in your own setup—just let me know!