ai-agents-chat-engine-behavior-rag-chat-engine-returns-hardcoded-empty-response-w-feadf5da

THEN Modify the chat engine (e.g., CondensePlusContextChatEngine) to invoke the LLM even when no nodes are found, allowing the model to respond with small talk, clarification requests, or information from the system prompt. A workaround is to set `empty_response_gen=None` or override the `generate` method to fall back to the base chat engine behavior.

# Workaround: Use a custom chat engine that ignores empty retrieval
from llama_index.core.chat_engine import CondensePlusContextChatEngine

class AlwaysRespondChatEngine(CondensePlusContextChatEngine):
    async def agenerate(self, message: str, **kwargs):
        self._current_message = message
        nodes = await self._retriever.aretrieve(message)
        if not nodes:
            return await self._llm.apredict(self._system_prompt, user_message=message)
        return await super().agenerate(message, **kwargs)