What you can do today

  • Search indexed text in your private workspace, with matches ranked by relevance.
  • Match all query words within a source line, regardless of their order.
  • Open matching passages with document, page and source-word references.
  • Keep the index aligned with current results and document deletions.
  • See processed-page coverage and unavailable pages alongside search results.

Find a source before making a decision

Search for a supplier or identifier, open the matching PDF page and inspect the original wording before using the extracted data.

Supported scope and limits

  • Search is lexical; it does not understand paraphrases or generate answers.
  • Unprocessed and unreadable pages cannot be searched.
  • No production retrieval speed benchmark is claimed.

Planned extensions

  • Semantic embeddings, metadata filters, hybrid retrieval and reranking.
  • Cited answers with abstention when evidence is missing or conflicting.

How to check the result

  • Check which documents and pages were searched.
  • Read the cited passage in context.
  • Treat no matches as a search result, not proof that a fact is absent.