Is there a map reduce document summarizer in LlamaIndex? i.e. I want to iterate over several (hundreds) of documents and summarize them together into one

Do function calling agents (OpenAI LLM) manage their history? Keep running into exceeding context length errors with OpenAI due to the history + prompt + tools exceeding 16k tokens, is there any way to avoid this?

Why do Tool Agents randomly abort and return 'None' with no error message? Debugging shows a 200 response from OAI, I tried setting max_iterations=100 but same result, and just after a few steps