The /api/_dump endpoint currently tries to fetch the entire database into memory at once. If a user stores thousands of keys, this is going to hit Vercel's serverless function memory limits and timeout. We need to refactor the dump logic to use Redis SCAN with a cursor, and update the dashboard UI to lazy-load or paginate the results as the user scrolls.
The /api/_dump endpoint currently tries to fetch the entire database into memory at once. If a user stores thousands of keys, this is going to hit Vercel's serverless function memory limits and timeout. We need to refactor the dump logic to use Redis SCAN with a cursor, and update the dashboard UI to lazy-load or paginate the results as the user scrolls.