osFoundry / Platform
Erasure
GDPR-style data deletion with retention policies for audit, chunks and memories.
What it is
Erasure is osFoundry's data-deletion and retention control, living under Settings and Security. It gives you a GDPR-style right-to-erasure workflow for the data an AI workspace accumulates as you use it: audit records of what agents and tools did, the indexed chunks that back your knowledge and semantic search, and the memories your agents carry between sessions. Instead of hunting through separate stores, you delete a subject's data or a scoped set of records from one place, and you decide up front how long each class of data is allowed to live.
Inside osFoundry the three data classes are handled distinctly because they behave differently. Audit is your log of actions and runs — useful for review and accountability, so it typically warrants a fixed retention window rather than immediate deletion. Chunks are the fragments your documents and sources get split into for retrieval; erasing a source means its chunks are removed so they can no longer surface in search or feed a model. Memories are the durable notes agents keep about people, projects, and preferences; erasure clears them so an agent stops recalling what you asked it to forget. Retention policies let you set a time-to-live per class, so old data ages out automatically instead of relying on someone remembering to purge it.
This matters because osFoundry is local-first and bring-your-own-key: much of your data can stay on your device and flows to model providers only through keys you control. Erasure completes that privacy story. Deletion applies to the copies osFoundry holds — audit, chunks, and memories in your workspace — so honoring a deletion request or an internal retention rule is a concrete action you can take, not a promise you have to make on a third party's behalf. It does not reach into an external provider's logs, so pairing erasure with your own provider agreements is the honest way to cover the full path.
Key capabilities
- Run GDPR-style right-to-erasure deletions for a data subject or a scoped set of records from Settings and Security
- Set retention windows separately for audit logs, knowledge chunks, and agent memories so each ages out on its own schedule
- Remove indexed chunks when you delete a source, so erased content stops surfacing in semantic search and model context
- Clear agent memories so agents stop recalling people, projects, or details you have asked to forget
- Keep deletion local to the data osFoundry holds in your workspace, consistent with its local-first, BYOK design
Use cases
- Privacy officer: fulfill a customer's right-to-erasure request by deleting their records across audit, chunks, and memories in one pass
- Team admin: set a retention window on audit logs so run history is kept for review but purges automatically after the window
- Knowledge owner: remove a sensitive document and confirm its chunks no longer appear in search or feed any agent's answers
Frequently asked questions
What data does Erasure actually delete?
The data osFoundry keeps in your workspace: audit records of agent and tool activity, the indexed chunks that back knowledge and semantic search, and agent memories. Because osFoundry is BYOK, it does not delete logs held by an external model provider — pair Erasure with your own provider agreements for the full path.
How are retention policies different from a one-time deletion?
A deletion removes data now. Retention policies set a time-to-live per class — audit, chunks, and memories — so data ages out automatically once it passes the window, instead of depending on someone running a manual purge. You can use both: keep audit for a fixed period while deleting a specific subject on request.
If I delete a document, does the AI stop using it?
Yes. Deleting a source removes its chunks, which are the fragments retrieval draws from, so that content can no longer surface in semantic search or be pulled into a model's context. Clearing related agent memories stops an agent from recalling it from past sessions as well.