Thanks — great question. Regarding extracting a call transcript from `llm.ChatContext`: there isn’t a built-in helper that returns the entire transcript as a single serialized string. You’ll need to read the stored messages from the

ChatContext

and construct the transcript yourself (including role, content, and timestamps if you want them). A simple approach: 1. Iterate the messages in the

ChatContext

in chronological order. 2. For each message, append the role (user/assistant/system), timestamp (if present), and content into your preferred text format. 3. Join the pieces into one string and pass that into your

llm.ChatContext

or LLM prompt. If you’d like, tell me: • which fields your messages include (e.g.,

role

,

content

,

timestamp

), and • the exact format you want for the transcript (simple "role: content" lines, or timestamped lines), and I can provide a short code snippet that builds the string in that format. Sources: External data and RAG | LiveKit Documentation | Pipeline nodes and hooks | LiveKit Documentation