Repository navigation
Conversation
![kit-code-agent[bot]](https://avatars.githubusercontent.com/in/4797102?s=60&v=4){kind=small}
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Inspect inline image dimensions even when their encoded field is small, and normalize safely decodable sources to the Responses output limits. Preserve actionable, sanitized local image failures in diagnostics.
Motivation
The existing normalization path only inspected images above the wire-byte limit and applied the output-axis limit before resizing. This rejected bounded long screenshots and let highly compressed oversized dimensions bypass normalization.
Impact
Long images within the existing source safety budgets can continue through retained-history requests without changing durable attachments. Images exceeding those budgets remain rejected with guidance to start a new conversation with a smaller image; removing or replacing an already-retained invalid attachment is not implemented here.
Partial fix for #314.
Technical details
The source limits remain 10 MiB encoded, 16,777,216 pixels, and 64 MiB decoded bytes. Output dimensions are reduced before the existing JPEG byte-budget loop, while fitting images retain their original encoding. WebP chunk ranges, including frame subchunks, are validated against the bounded source before metadata decoding. Diagnostic messages use an exact static allowlist rather than arbitrary provider text.