kovacKnowledge & retrieval
Answers grounded in your own documents
Upload, connect or crawl your knowledge once. Every search combines meaning and keywords, respects who's allowed to read what, and cites the passage it came from.
59 file formats
PDF, Office documents, email, spreadsheets, code and more, parsed into searchable passages.
Connected sources
Sync from S3-compatible storage, SQL databases, Google Drive, Confluence, SharePoint and websites. Permissions travel with the documents.
Document lifecycle
A re-uploaded document supersedes the old version rather than duplicating it, with its history kept.
Hybrid search with reranking
Meaning-based and keyword search run together, fused into one ranked list, then reordered by a reranker you can swap with no restart.
Multilingual retrieval
A question in one language finds passages written in another.
Implicit filters
The agent works out filters such as a document's department or date straight from the question's wording.
Tables and figures
Tables keep their rows and columns; charts and diagrams get a written description an agent can search.
Chunking tuned on your data
kovac tries different ways of splitting a collection, measures retrieval against your own test set, and recommends a change only when it measurably helps.
Document-level permissions
Each user sees answers only from documents they are allowed to read, enforced in the database.