[AsyncEngine Refactor 1/N] define MultimodalProcessor to handle multimodal data processing

[lvhan028]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily receiving feedbacks. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

Please describe the motivation of this PR and the goal you want to achieve through this PR.

Modification

Please briefly describe what modification is made in this PR.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[Copilot]

Pull request overview

This PR introduces the MultimodalProcessor class as the first step in refactoring the AsyncEngine architecture. The goal is to extract and consolidate prompt processing logic (both text-only and multimodal) into a dedicated processor component.

Key Changes:

• Introduced MultimodalProcessor class to handle prompt preprocessing, message content merging, and multimodal processing
• Refactored VLAsyncEngine and AsyncEngine to delegate prompt processing to MultimodalProcessor
• Added Response.extend() method to replace the standalone _append_response() function
• Converted GenOut to use a to_response() method instead of a standalone conversion function

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
lmdeploy/serve/multimodal_processor.py	New file introducing the MultimodalProcessor class that consolidates prompt preprocessing logic for both text and multimodal inputs
lmdeploy/serve/vl_async_engine.py	Refactored to delegate prompt processing to MultimodalProcessor and removed duplicate image conversion logic
lmdeploy/serve/async_engine.py	Removed prompt processing logic and helper functions; now uses MultimodalProcessor for all prompt handling and updated to use new Response.extend() and GenOut.to_response() methods
lmdeploy/messages.py	Added Response.extend() method to enable fluent chaining when accumulating response data

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CUHKSZzxy]

LGTM