What are the benefits passing RunContext vs AgentSession

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

just tell me the difference and what they both do

Let me help clarify the difference between

RunContext

and

AgentSession

in LiveKit for you: • RunContext: Think of this as a temporary context object that’s passed to a function tool while it’s running. It offers: Basically, it serves as a helpful link between one function call and the overall session. ◦ Access to the

AgentSession

, giving you a broader view of the session. ◦ Control over speech playout and interruptions through a

SpeechHandle

. ◦ Information about the current function being executed via a

FunctionCall

object. ◦ A

userData

property where you can keep session-specific data shared with external systems. • AgentSession: This is the big picture — the full runtime environment for a voice agent. It handles: This is created once per user interaction and manages everything from start to finish during the session. ◦ Connecting to a LiveKit room and managing WebRTC media streams. ◦ Detecting turns and tracking user/agent states like "listening" or "speaking". ◦ Keeping the global conversation history with

ChatContext

. ◦ Maintaining a persistent

userData

object throughout the session. ◦ Providing methods to guide the conversation flow, such as

start()

,

generate_reply()

, and

say()

. To sum it up,

AgentSession

is like the container for the whole session, while

RunContext

is a temporary helper for each specific function call within that session. Sources: Agents Core Architecture | livekit/livekit_composite | DeepWiki | tests/test_evals.py | complex-agents/vision/agent.py | agents/src/voice/index.ts