Large File Handling Strategy

[nonothingless]

Problem Statement

Regarding the review strategy for large files and large diffs, it is recommended to implement degradation handling strategies—such as individual review or split review—rather than simply filtering them out and skipping the review entirely.

Proposed Solution

Users can choose whether to process large files via command-line arguments.

Alternatives Considered

No response

Affected Area

Review Agent / LLM interaction

Additional Context

No response

[chacha923]

Hi @lizhengfeng101, I would like to work on this issue. Could you please assign it to me?

For the first PR, I propose keeping the scope focused on large diffs in ocr review:

• add an opt-in CLI strategy for large diffs, while preserving the current skip behavior by default;
• split oversized diffs on complete hunk boundaries and keep each chunk within a safe token budget;
• preserve file headers and line information so generated review comments can still be located correctly;
• handle a single oversized hunk with a bounded line-based fallback;
• add unit tests and CLI documentation for the new behavior.

To keep the first PR reviewable, I plan to leave full-file handling in ocr scan and broader cross-chunk context/deduplication improvements for follow-up work.

If this scope looks good, I will prepare the first PR.

[lizhengfeng101]

@nonothingless @chacha923 Thank you for your suggestion. This feature has a significant impact on changes to the core logic of the system, and we are still considering it. We believe we will invite you to participate when the time is right.