Convert BaseDocumentCompressor into a Runnable object so that it can use invok and tracked by LangSmith.

[gbaian10]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

I recently used CohereRerank and found that it cannot track detailed input and output content on LangSmith.

And CohereRerank lacks a score_threshold filtering parameter similar to the one in retrievers (currently it only has top_n).

Motivation

Make objects like CohereRerank that inherit from BaseDocumentCompressor traceable and callable using invoke.

Proposal (If applicable)

I studied the source code and found that its parent class, BaseDocumentCompressor, is not a Runnable but simply a BaseModel.

Since BaseDocumentCompressor already has an abstract method compress_documents, and there is a lot of existing code that inherits from this class, I've thought of a simple solution:

Change BaseDocumentCompressor to inherit from RunnableLambda instead of BaseModel.

Then, let invoke directly execute compress_documents.
Or execute super().__init__ in the __init__ method and pass the compress_documents function to it.

In the current ContextualCompressionRetriever(which is a BaseRetriever), change the compress_documents method to invoke.

This way, the original method remains intact (you can still use compress_documents), while allowing ContextualCompressionRetriever to track content.

Additionally, BaseDocumentCompressor can directly use invoke.

---

If the above modification is acceptable, I can implement this feature and submit a PR.
Please let me know everyone's thoughts on this feature.