feat(react): wire user-llm-text into usePipecatConversation

[rahulsolanki001]

Follow-up to #211 — depends on it (branched from that fix, needs RTVIEvent.UserLlmText from client-js). Please merge #211 first.

Problem

In chat mode the server emits user-llm-text instead of user-transcription (typed text bypasses STT entirely). Because usePipecatConversation had no handler for RTVIEvent.UserLlmText, user messages were silently dropped and never appeared in the conversation array.

Solution

Handle RTVIEvent.UserLlmText in useConversationEventWiring with a guard that avoids duplicating text in voice mode:

• Voice mode — user-transcription already created an in-progress user message. We skip upsertUserTranscript and only call finalizeLastMessage to mark the turn complete.
• Chat mode — No in-progress user message exists. We create one from the LLM text, then finalize it.

useRTVIClientEvent(
  RTVIEvent.UserLlmText,
  useAtomCallback(
    useCallback((get, set, data) => {
      const messages = get(messagesAtom);
      const lastUserIdx = findLastIndex(
        messages,
        (m: ConversationMessage) => m.role === "user"
      );

      if (lastUserIdx === -1 || messages[lastUserIdx].final) {
        upsertUserTranscript(get, set, data.text ?? "", true);
      }
      finalizeLastMessage(get, set, "user");
    }, [])
  )
);

Changes

File	Change
client-react/src/conversation/useConversationEventWiring.ts	Wire RTVIEvent.UserLlmText with voice/chat mode guard

Test plan

• Chat mode: user messages appear in the conversation array via usePipecatConversation
• Voice mode: no duplicate user text — user-transcription message is finalized, not duplicated
• Existing UserTranscript behavior unchanged
• TypeScript builds cleanly once fix: respect final flag in mergeMessages and handle user-llm-text event #211 is merged (tsc --noEmit)