WINNOWR

What WINNOWR is

WINNOWR is an academic-integrity tool for finding possible web-source matches in student work and other long-form writing. It is designed to run CPU-first and gives reviewers evidence to examine, including index overlap, candidate source links, and suspicious passages highlighted in submitted text. Its matching process applies consistent text normalization and compact word-shingle fingerprints to identify candidate documents. Institutions can create an index from their own JSONL.zst corpus, typically containing URL and text fields, with title optional. The project also documents workflows based on FineWeb-Edu and Wikipedia. Text and files can be submitted through an HTTP API, which includes endpoints for submissions, reports, text, purging, and health checks. The documented system has an index builder, an API server, and a background worker. Its runtime uses SQLite, Cloudflare Queues, Cloudflare R2, and a local sharded disk index. Build and deployment instructions cover the API and worker, including plain Linux machines. WINNOWR is free and available as an API, web service, and self-hosted software. Longsight presents it for institutional plagiarism review and offers hosting and support. The project is still evolving, and source snippets or highlighted spans may not be available if a live page cannot be fetched or parsed. Its GNU AGPLv3 license requires modified deployed versions to make those modifications available under the same license.

Who it suits

WINNOWR suits institutions seeking a self-hostable academic-integrity system that identifies web-source candidates and provides evidence for review. It may also suit teams able to build and maintain a corpus and operate the documented API and worker.

Limits

WINNOWR is free. Source-page snippets and highlighted spans are best-effort and can be unavailable when a live page cannot be fetched or parsed.

Getting started

1. Open the project repository at https://github.com/LongsightGroup/winnowr.
2. Build an index from a JSONL.zst corpus with URL and text fields; title is optional.
3. Follow the repository's build and deployment instructions for the API server and background worker.
4. Run the documented runtime with its SQLite, Cloudflare Queues, Cloudflare R2, and local index components.
5. Submit text or files through the HTTP API and use its report and health endpoints.

Verdict

Choose WINNOWR if an institution wants free, API-based web-source matching with reviewable reports and the option to self-host. Look elsewhere if you need a mature, fixed product or guaranteed page snippets and highlighted evidence; the project is still developing and that enrichment is best-effort.