fix(cli): remove access to private asyncio internals in monitor_stream

[Rakshit-gen]

Description

This PR removes the use of private asyncio internals (_buffer attribute and _maybe_resume_transport() method) from the monitor_stream() function in libs/cli/langgraph_cli/exec.py.

Problem

The current implementation directly accesses asyncio.StreamReader._buffer and calls _maybe_resume_transport() when handling LimitOverrunError. These are private implementation details that:

• Are not guaranteed to exist in future Python versions
• Can change behavior between minor Python releases
• Have no documented API contract
• Require # type: ignore comments due to accessing private members

Solution

Refactored the LimitOverrunError handling to use only public asyncio APIs:

• Replaced _buffer access with stream.read(e.consumed) to read available data
• Removed the _maybe_resume_transport() call (handled automatically by public read methods)
• Implemented proper handling for cases where the separator is not immediately after the consumed bytes
• Added nested LimitOverrunError handling for edge cases

Benefits

• ✅ Future-proof: Won't break when Python's asyncio implementation changes
• ✅ Maintainable: Uses only documented, stable APIs
• ✅ Type-safe: No need for # type: ignore comments
• ✅ Backwards compatible: Maintains identical external behavior

Issue

Fixes potential breakage in future Python versions where asyncio internals may change.

Dependencies

None - uses only standard library asyncio public APIs.

---

Testing

• Verified no references to _buffer or _maybe_resume_transport remain in the codebase
• Maintains same behavior for normal stream reading and buffer overrun scenarios
• No breaking changes to external API

[vercel]

@Rakshit-gen is attempting to deploy a commit to the LangChain Team on Vercel.

A member of the Team first needs to authorize it.