feat(core): support custom message separator in get_buffer_string()

[chrispy-snps]

Title

feat(core): support custom message separator in get_buffer_string()

Description

Small enhancement to support user-configurable message separator in get_buffer_string().

Issue Addressed

Currently, get_buffer_string() uses a single newline to separate messages. If messages contain complex content with newlines, then the message boundaries become obfuscated in the output:

from langchain_core.messages import SystemMessage, HumanMessage, AIMessage, get_buffer_string

messages = [
    HumanMessage(content="I have a question.\n\nCan you help me?"),
    AIMessage(content="Sure!\n\nI can help you."),
    HumanMessage(content="What is the capital of France?"),
]

print(get_buffer_string(messages))
# Human: I have a question.
#
# Can you help me?
# AI: Sure!
#
# I can help you.
# Human: What is the capital of France?

Changes

• Implemented a message_separator argument for the get_buffer_string() function.
  • The name was chosen to be consistent with other *_separator arguments in the code base.
• Updated unit tests to test the new argument.

Example

print(get_buffer_string(messages, message_separator="\n\n"))
# Human: I have a question.
#
# Can you help me?
#
# AI: Sure!
#
# I can help you.
#
# Human: What is the capital of France?

[codspeed-hq]

CodSpeed Performance Report

Merging #34569 will improve performance by 27.42%

_{Comparing chrispy-snps:feat/buffer-string-message-separator (f507007) with master (a7aad60)}

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Summary

⚡ 10 improvements
✅ 3 untouched
⏩ 21 skipped¹

Benchmarks breakdown

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	WallTime	test_import_time[tool]	666.8 ms	529.9 ms	+25.84%
⚡	WallTime	test_import_time[Runnable]	636.1 ms	499.2 ms	+27.42%
⚡	WallTime	test_import_time[RunnableLambda]	625.8 ms	503.9 ms	+24.21%
⚡	WallTime	test_import_time[HumanMessage]	312.3 ms	264.7 ms	+17.96%
⚡	WallTime	test_import_time[Document]	217.7 ms	189.4 ms	+14.91%
⚡	WallTime	test_import_time[InMemoryVectorStore]	790.7 ms	632.2 ms	+25.08%
⚡	WallTime	test_import_time[LangChainTracer]	563.9 ms	458.2 ms	+23.07%
⚡	WallTime	test_import_time[CallbackManager]	588.6 ms	466.4 ms	+26.2%
⚡	WallTime	test_import_time[BaseChatModel]	661.8 ms	538.6 ms	+22.89%
⚡	WallTime	test_import_time[ChatPromptTemplate]	741.8 ms	608.1 ms	+22%

Footnotes

1. 21 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[chrispy-snps]

Thank you @ccurme!