Concepts — License- and Privacy-Aware Dataset Curation¶
Knovaryn is designed for license-aware and privacy-aware training-data curation: it tracks the rights on your sources, flags sensitive content, and refuses to publish anything it cannot show is permitted. This page is the orientation summary; the operational checklist is in Privacy & Licensing.
License awareness¶
- Every source document carries a declared license in the source-license registry.
- The license gate blocks any sample whose sources are not permitted to be used for training.
- Publication is dry-run by default and gated on license approval — nothing is pushed anywhere without explicit authorization.
- Code licensing (Apache-2.0) is kept separate from dataset licensing; dataset licensing is governed by the source-license registry.
Privacy awareness¶
- Untrusted-in, controlled-out: every source is treated as untrusted at intake; nothing in a source can drive code execution, path traversal, or prompt injection.
- Secrets from the environment only — never hard-coded, never logged, redacted at the display boundary.
- Server-side privacy reports classify sensitive content per policy; the privacy gate flags failures rather than exporting them.
- Local by default: SQLite + filesystem; scale to a team deployment (PostgreSQL + S3-compatible) is a config change, not a fork.
When you move to production¶
The built-in controls reduce risk; they are not a certification. Whether a given corpus may be used for training is a legal determination you make for your jurisdiction, your contract terms, and the provenance of each document. See Privacy & Licensing and the deployment guide.
Related¶
- Deep dive — Privacy & Licensing, hardening.
- Concepts — quality gates, dataset provenance.
- Security model — architecture/security.