From the moment we began building Maple, we established a strict architectural boundary: data sovereignty belongs entirely to the user. In a market saturated with cloud-first SaaS platforms that quietly absorb user workspace logs, telemetry, and private document indexes to train distant models, we chose a different path. We engineered Maple to operate as a local-first application where your data is stored inside a sandboxed cryptographic vault on your own device.
When you connect Notion workspaces, local markdown vaults, Slack channels, or audio recordings to Maple, the system generates localized vector indexes and relational maps. These do not sit in the cloud. Instead, they reside in a high-performance SQLite database operating in WAL (Write-Ahead Logging) mode, accompanied by highly optimized binary vector logs on your SSD. The decryption key is managed securely via your operating system’s secure hardware store (Apple Keychain or Windows Credential Manager).
How the Local Purging Mechanics Work
If you choose to delete a workspace, Maple doesn't just mark records as 'archived'. We perform a multi-phase hardware-level cleanup. First, memory buffers holding index caches are forcefully flushed to disk. Second, we execute a randomized binary overwrite of the workspace SQLite records. Third, the SQLite VACUUM command is executed to release the storage blocks back to the OS, ensuring that forensics tools cannot recover the discarded records.