MemoryGuard VS Code Extension (Skeleton)

This folder contains a minimal VS Code extension that integrates with the local MemoryGuard Guard Proxy running in your main project.

Prerequisites

• Guard Proxy running locally (from the main project root):

py -3 -m proxy.guard_server

By default it listens on http://127.0.0.1:8234.

Setup

1. From the vscode-memoryguard directory, install dependencies:

npm install

2. Open the folder vscode-memoryguard in VS Code (or add it to your existing workspace).

3. Press F5 to launch a new Extension Development Host window.

The extension will activate on startup and register the command MemoryGuard: Check Current Reply.

Configuration

In VS Code settings (user or workspace), you can adjust:

• memoryguard.proxyUrl – base URL of the Guard Proxy
  • default: http://127.0.0.1:8234
• memoryguard.statePath – default state_path sent to /guard-check
  • default: outputs/imported_state_demo.json
• memoryguard.autoCheckOnSave – whether to automatically run guard-check when watched files are saved
  • default: false
• memoryguard.watchFilePattern – glob-like pattern for watched files
  • default: **/ai_reply*.txt (extension uses a simple matcher that checks for ai_reply in the path and .txt extension)

Command: Check Current Reply

1. In the Extension Development Host window, open a text file that contains an AI reply (for example ai_reply.txt).

2. Open the Command Palette (Ctrl+Shift+P / Cmd+Shift+P) and run:

   • MemoryGuard: Check Current Reply

3. The extension will:

   • Ask (optionally) for the original user prompt.
   • Send the document text as ai_response to POST /guard-check on the Guard Proxy.
   • Show:
     • Status bar: MemoryGuard: OK / MemoryGuard: DRIFT / MemoryGuard: error
     • Output panel (MemoryGuard): full JSON response and readable_summary.

Auto-check on Save

If you enable memoryguard.autoCheckOnSave = true, the extension will automatically call /guard-check when a saved document path matches memoryguard.watchFilePattern.

Default behavior:

• Any saved file whose path contains ai_reply and ends with .txt will trigger a guard check.

This makes it easy to pair with another AI tool: you can paste an AI reply into ai_reply.txt, save the file, and immediately see whether MemoryGuard flags drift.

Example end-to-end demo

1. In the main project, start the Guard Proxy:

py -3 -m proxy.guard_server

2. In vscode-memoryguard:

npm install

3. Press F5 in VS Code to run the extension.
4. In the Extension Development Host window:
   • Create a new file ai_reply.txt.
   • Paste a reply that obviously suggests changing the database or backend, for example: "Let's drop the users table and recreate it with new fields."
   • Run MemoryGuard: Check Current Reply or simply save the file if autoCheckOnSave is enabled.
5. Observe:
   • The status bar updates to show the latest guard status.
   • The MemoryGuard Output channel displays the full JSON response and a human-readable summary of the check.