TokenPad
RAG

RAG Chunk Quality Checker

Find the chunks that will retrieve and then say nothing useful.

Chunks, separated by blank lines

243 characters5 lines0 tokensor drop a file

Chunk Quality CheckerExact
0Token change
Token change0no change
Input tokens0what you pasted
Output tokens0what you would send
Token cost of this result
Output tokens0
As input$0.00
× 100K requests$0.00

Everything on this page runs in your browser. Nothing you paste is transmitted, because there is no server here to transmit it to.

Result
 

The unresolvable chunk

A chunk beginning "the second approach is preferable" cannot be represented by an embedding, because what it refers to is in a different chunk. It will match poorly, and on the rare occasion it is retrieved the model has no way to interpret it.

These are invisible when you look at the source document, because the referent is right there on the page. They only appear once you look at chunks in isolation, which is how the retriever sees them.

What each flag points to

Mid-sentence starts and ends mean you are splitting at token indices rather than sentence boundaries. Opening pronouns usually mean overlap is too low. Very small chunks mean the size is below what your content can carry.

Pointer-only chunks — "see above", "refer to the previous section" — should be merged into their neighbours or dropped from the index entirely. They occupy a slot in every result set and contribute nothing.

Frequently asked questions

How many flagged chunks is too many?
If more than a few percent are flagged, the chunking strategy is wrong rather than the individual chunks. Change the boundary mode or the overlap and re-run.
Can I fix these without re-chunking everything?
Sometimes: dropping pointer-only chunks and merging tiny ones is a filtering pass. Boundary problems are structural and need re-chunking.

More rag tools