Feature Request: Native Cursor Ingestion Support

🚀 The Problem

Currently, MemPalace relies on standard text/JSON exports or manual mining of directories. However, Cursor (the AI code editor) stores its most valuable conversational "DNA" (including agent reasoning and composer history) inside internal SQLite databases (state.vscdb) located in obfuscated workspace storage directories. Users cannot easily "mine" this memory into MemPalace without manual extraction.

💡 The Proposal

I propose adding a native Cursor ingestion mode to the mempalace mine command. This would allow MemPalace to automatically:

1. Locate Cursor's workspaceStorage directory (platform-specific).
2. Iterate through all workspace SQLite databases.
3. Extract conversational data from the ItemTable (specifically the composer.composerData key).
4. File these sessions verbatim into appropriate Rooms/Wings in the Memory Palace.

🛠️ Technical Details

I have already developed a working prototype script that achieves this on Windows:

• Path: %APPDATA%\Cursor\User\workspaceStorage
• Query: SELECT value FROM ItemTable WHERE key = 'composer.composerData'
• Data Format: The value is a JSON string containing an allComposers array with full chat transcripts.

By integrating this logic into mempalace/importers/cursor.py, we can provide a zero-config way for Cursor users to back up their entire coding history into MemPalace.

✅ Alignment with MemPalace Principles

• Verbatim First: Extracts raw transcripts exactly as they appear in the Cursor DB.
• Local First: No cloud API or export required; works entirely on the local filesystem.
• Zero API: Uses standard Python sqlite3 and json modules.

🧪 Proof of Concept

I successfully mined 160 sessions from a local Cursor installation using this method today.

I am happy to submit a PR if this approach aligns with the roadmap!