langchain - 💡(How to fix) Fix convert_to_openai_messages mutates the caller's input message in place [1 pull requests]

convert_to_openai_messages rewrites the caller's own message dict while converting it. For an OpenAI file block with no filename, it does block["file"]["filename"] = "LC_AUTOGENERATED" on the passed-in object (libs/core/langchain_core/messages/utils.py, ~L1778) instead of building a new block. So after calling the function, the original message you passed in is silently modified, a side effect in what looks like a pure converter.

Repro: build a HumanMessage with a file content block that has no filename, call convert_to_openai_messages([msg]), then inspect the original block. Its file.filename is now "LC_AUTOGENERATED". Re-using the same message list (e.g. across providers) carries the mutation forward.

Proposed fix: construct a new block instead of mutating in place. I have a small fix plus a unit test that asserts the caller's input is unchanged, and I'd be happy to open the PR. Could a maintainer assign this to me?