0
Chunks
0
Total tokens
0
Avg / chunk
0
Overlap tokens
Chunk map
chunk
overlap
ideal
small
large
Per-chunk breakdown
Cost estimate
$0.00
One-time indexing (embed all chunks)
$0.00
Per query (embed + top-k into generation)
Rough estimate. Token counts use a subword heuristic that tracks real tokenizers within ~10–15% on prose; exact tokenizers vary by model. Costs use the selected preset's embedding / generation rates.
Strategy comparison
Structure-aware
0 chunks
—
Fixed-size
0 chunks
—
Both strategies at the current size. Structure-aware respects headings and paragraphs; fixed-size cuts every N characters. Compare clean cuts vs. mid-sentence splits.
Unlock the Full Edition
Demo limits some controlsThe demo shows structure-aware and fixed-size chunking with live token counts. The Full Edition unlocks the levers that actually tune a production pipeline:
- Overlap control (boundary-spanning facts)
- Cost presets across model tiers
- Top-k query-cost modelling
- Side-by-side strategy comparison
- Export chunk config as JSON
- Vector-DB ready metadata preview