Leave Feedback

What we'd love to hear

Anything is welcome, but these are the kinds of feedback that most directly shape where VectorLens goes. The more specific you are, the better we can act on it.

• An embedding model we don't support yet. Send the exact model information such as the Hugging Face repo name or the API model string (for example OpenAI text-embedding-3-large) and a word on how important it is for your use case.
• A language other than English. Our PII classifiers are trained mainly on English today. Tell us which language you need and, if you can, which embedding model you use it with.
• A vector store or export path that was missing or painful. If your database isn't in the docs or the export was a hassle, we want to smooth it out.
• A new PII category or output format. Different sensitive-data classes, an alternate report shape, CI/CD integration, scriptability or anything else that would make VectorLens better fit your workflow.
• A bug. See the template below so we can reproduce it quickly.
• Inversion attacks. We've built models that invert embeddings back toward their source text, but deliberately withheld that at launch to keep VectorLens a tool just for defenders to start. If you have a legitimate use case, please give us some details.

Filing a good bug report

Paste this into the form and fill it in:

VectorLens version: (ironcore-vector-lens --version)
OS / arch:           (e.g. macOS arm64, Linux x86_64 CUDA)
Model used:
Command run:
Summary table:       (the counts block, if relevant)
Expected vs. actual:

Please don't paste real vectors or live PII into this form. VectorLens is built so your data never leaves your machine, and feedback should hold to the same line. The summary table and the metadata above are all we need.

Prefer email?

If a form isn't your thing, write to us directly at support@ironcorelabs.com. Note that this may be more likely to hit spam filters though.