How Camelia stays grounded in your real projects and files

Camelia's usefulness comes from grounding: she works from your actual projects and files, not a generic model's assumptions. This is why her answers reflect what's really in your project.

Grounded in your data

Camelia runs server-side and is grounded with your real projects and files from the encrypted data store. She can list your projects, list files in a folder, and read a document's full content — so when you ask about your spec, your research, or your design, she's reading the real thing.

Scoped to you, by design

Every action Camelia takes runs against the authenticated account's data, and every stored row is scoped to its owner. Your project data is encrypted at rest with AES-GCM, and the keys never reach the browser. Practically, this means Camelia only ever sees and acts on your own projects — there's no cross-account leakage.

Acting on what she reads

Because she's grounded, Camelia can do real file work: create or overwrite documents, label and move files, and update your project's name or summary. When she drafts something, she can save it straight into the right backbone folder so it becomes part of your project rather than a throwaway chat message.

Per-project context

Camelia also reads per-project directives you can edit (stored in your project's Misc folder), so her behavior adapts to the project she's in. Combined with the live file context, this lets her tailor work to each project's goals.

What this means for you

Keep your files organized and your directives current, and Camelia's output gets sharper. The more accurate your project's real state, the better grounded — and more useful — her work will be.