Better parsing of agent output For some agents, the parsing of LLM output to agent action was pretty brittle A few people called this out, but s/o in particular to brandon69 on discord This should be improved, and can now handle things like multi line return values!

→ View original post on X — @langchain,