Agentkit / practical resources

Before your next re-index

What changed in your source export?

Compare two RAG source snapshots. Separate changed text from failed requests and missing pages before deciding what to refresh.

Runs in your browserNo signup or uploadReview only
Missing does not mean deleted. A shorter export can come from a failed crawl, a new filter or lost access. This tool never fetches URLs, updates an index or confirms a deletion.

Use exports from the same crawl scope

Export one record per source URL with its extracted text. Keep URL filters, extraction settings and access permissions consistent. Record the scope below; the tool cannot verify export completeness.

Accepted JSON and JSONL format

Use a JSON array, or one JSON object per line. Each record needs an absolute HTTP(S) url and one string field named content, markdown or text. Optional status is an integer HTTP response code. Other fields are ignored.

[{"url":"https://example.org/help/returns","text":"Returns within 30 days.","status":200}]

Limit: 10 MB and 10,000 records per export. Failed HTTP statuses and empty text are flagged, never treated as successful content changes. URLs use standard URL normalization and ignore fragments. Query strings and trailing slashes remain distinct. Duplicate or invalid URLs stop comparison.

1. Before snapshot
2. After snapshot

Your inputs stay in this tab. Reloading clears them. The downloaded report includes URLs, hashes and your scope notes, but no source text.