I built chat agent using AgentSession now i am storing the instance in memory and want to inject the context into this chat session. how can i do that ?

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

i dont want to update the entire context. actually we are spinning parallel agent calls and wanted to update the context for the subsequent calls mad so that user wont feel any inconsistency when he came back to the chat agent again.

Thanks — I’m responding to your follow-up about updating context for parallel agent calls without replacing the whole

chat_ctx

. You can inject incremental context so subsequent calls see the updates and users don’t experience inconsistency. Options: • Append messages directly to the existing context:

# use the current agent's chat_ctx
ctx = session.current_agent.chat_ctx
ctx.add_message(role="system", content="Background update: user prefers evening slots.")
# then persist/merge the change
await session.current_agent.update_chat_ctx(ctx)

• If parallel workers produce updates, have each worker return its update (message or summary). Merge those into a copy of

chat_ctx

and call

update_chat_ctx()

once with the merged additions to avoid race conditions:

ctx_copy = session.current_agent.chat_ctx.copy()
ctx_copy.add_message(role="assistant", content=worker_summary)
await session.current_agent.update_chat_ctx(ctx_copy)

Guidance to avoid inconsistencies: • Only append finalized/contextual messages (summaries, tool results, user-preference flags). • Batch/serialize updates (collect parallel results, then merge) to prevent conflicting writes. • Don’t replace the entire

chat_ctx

unless you intend to reset history. Would you like a short example that shows collecting updates from multiple parallel tasks and merging them safely? Sources: Pipeline nodes and hooks | LiveKit Documentation | Images | LiveKit Documentation | Agent speech and audio | LiveKit Documentation, with interactive elements