Does this create embeddings?
No. It uses deterministic local term scoring, so exploration has no model or API cost.
Retrieval before embeddings
See what your RAG system would actually retrieve.
Split real source text, ask a real question, and inspect the passages a deterministic local retriever ranks highest. Tune chunk size and overlap before paying for embeddings.
Runs in this browserNo uploadNo embeddingsNothing saved
01 / Source
Test your content
Paste text or load a local TXT/Markdown file up to 5 MB.
02 / Retrieval test
ReadyWhat surfaced
Chunks0
Average words0
Repeated overlap0%
Matching results0
Top retrieval matches
What to review
All chunks
This lexical ranking is a fast diagnostic, not a promise of embedding-model performance. Validate representative questions with your production retrieval stack before launch.
Testing limits
Useful before—not instead of—production evaluation.
What should I paste?
Use representative policies, help articles or documentation and ask questions your users genuinely ask.
What is a good chunk size?
There is no universal number. Prefer the smallest chunks that preserve the complete answer and retrieve consistently.