reload_catalog

Reloads the knowledge catalogue after it changed.

Group: Set up and develop.

Parameters

This tool takes no parameters.

What the model reads

The text below is the tool's own description, exactly as the AI model receives it.

Show the full description

Hot-reload the per-backend KNOWLEDGE dicts from disk.

Closes the gap identified by the
mcp-catalog-staleness-runtime-isolation post-mortem
(2026-06-01): the MCP server normally imports
src/backends/<be>/generators/<physics>.py modules ONCE at
startup and never refreshes them, so catalog edits made
during a long-running session are invisible. Postmortems
in data/postmortems/ are scanned on every request (already
hot), but pitfall dicts are not.

This tool walks every imported `backends.<be>.generators.*`
and `backends.<be>.backend` module, runs importlib.reload
on each, and re-runs load_all_backends() so the registry
re-binds the backend objects to the refreshed module
attributes. After the call, the very next
mcp__openpaso__knowledge call returns the on-disk
catalog without having to restart Claude Code.

Returns a one-line summary of which modules were
successfully reloaded vs which raised, so the caller can
tell when a syntax error in a newly-edited generator
prevented its module from re-importing (in that case the
OLD dict is still served from the previous import).