WINNOWR

Spec sheet

Purpose

WINNOWR is a CPU-first web-source detection engine for student writing and other long-form text.Read 8 Oct 2026 · github.com

Audience

Longsight presents Winnowr as academic-integrity software for institutions, with a focus on plagiarism detection and reviewable evidence.Read 4 Oct 2026 · longsight.com

Deployment

The README describes building the Go binaries and running the API and worker on a Linux machine.Read 8 Oct 2026 · github.com

Integrations

The production-style runtime uses SQLite, Cloudflare Queues, and Cloudflare R2.Read 8 Oct 2026 · github.com

License

The repository is licensed under GNU AGPLv3.Read 8 Oct 2026 · github.com

Architecture

The documented runtime uses SQLite, Cloudflare Queues, Cloudflare R2, and a local sharded index on disk.Read 4 Oct 2026 · github.com

Company history

Longsight says Sam Ottenhoff and Scott Siddall founded the company in 2004.Read 4 Oct 2026 · longsight.com

Corpus options

Users can build an index from their own JSONL.zst corpus; the repository also describes FineWeb-Edu and Wikipedia workflows.Read 4 Oct 2026 · github.com

Corpus support

Users can build an index from their own JSONL.zst corpus, with url and text fields generally needed and title optional.Read 8 Oct 2026 · github.com

Hosting and support

Longsight says it can host and support its products, and that products with a self-hosted option include a path for institutions to run them.Read 4 Oct 2026 · longsight.com

Limitations

The README says source-page snippets and highlighted spans are best-effort enrichment and may be unavailable when a live page cannot be fetched or parsed.Read 8 Oct 2026 · github.com

Matching

It normalizes text, creates word-shingle fingerprints, searches an index for overlap, and fetches likely source pages for evidence.Read 8 Oct 2026 · github.com

Maturity

The README describes WINNOWR as usable but still actively evolving, with documentation and packaging catching up.Read 8 Oct 2026 · github.com

Reports

Reports include index overlap, source URLs, and highlighted suspicious spans.Read 8 Oct 2026 · github.com

Runtime

The repository describes three main binaries: an index builder, an API server, and a background worker.Read 8 Oct 2026 · github.com

Self-hosting

The project describes itself as understandable and self-hostable, with an on-disk index and local runtime configuration.Read 8 Oct 2026 · github.com

Submission methods

The HTTP API accepts text or file submissions and provides submission, report, purge, and health endpoints.Read 4 Oct 2026 · github.com

Submissions

The API accepts text and file submissions and provides submission, report, text, purge, and health endpoints.Read 8 Oct 2026 · github.com

Where Specifiction places it

• Plagiarism Detection Software#27 of 286.7