R2 search endpoint loads all objects into memory

[dotsdl]

Problem

The /api/r2/search route handler calls listAllObjects, which paginates through every S3 object under a given prefix and loads them all into a JS array, then filters by filename in JS:

const objects = await listAllObjects(s3Client, bucket, prefix);
const matches = objects.filter(obj => { /* ... */ });

For a large bucket with many checkpoints, this will be slow, memory-intensive, and could time out on the server.

Suggested fix

• Add a result count limit to listAllObjects (stop paginating after N objects).
• Encourage callers to push the prefix as deep as possible before listing to reduce the search space.
• Consider whether a different S3 access pattern (e.g., constructing known key paths directly rather than searching) could avoid the full listing.