← All add-ons

MxChat Migration Tool

Version 1.3.2 · Updated 3 weeks ago

Switch embedding models or vector databases without losing your knowledge base. Re-embed everything in managed batches, move between WordPress and Pinecone in either direction, and export your content for backups or external vector stores.

Features

  • Knowledge base migration — move every entry between the WordPress database and Pinecone, in either direction, with metadata preserved.
  • Model switching — re-generate embeddings for your whole knowledge base when changing models: OpenAI (ada-002, 3-small, 3-large), Voyage AI, or Google Gemini.
  • Actions migration — re-embeds your trigger phrases/intents with the new model and automatically adjusts similarity thresholds so they keep firing.
  • Knowledge base export — download your content as Text or JSON for backups or OpenAI vector stores, with automatic splitting for large sets.
  • Batch processing — migrations run in configurable batches to avoid timeouts and API rate limits, with real-time progress and detailed logs.
  • Safe and reversible — choose whether to update your MxChat settings after a migration, and review past runs in the migration history.

What's new

Version 1.3.2

  • New Import tab — restore a JSON backup created by the Export tab. Entries are matched by their source URL and replaced rather than duplicated, so re-importing the same backup is always safe
  • New Imports run in small batches with live progress, and every entry is re-embedded with your current embedding model, so a backup made under a different embedding model restores cleanly
  • New Before anything is written, the import shows a summary: how many entries the file contains, how many already exist on this site and will be replaced, and where the entries will be stored
  • New Role restrictions on entries survive the backup-and-restore round trip, and manually pasted entries keep their identity so they are never duplicated on repeat imports
  • Improved JSON exports now record the site address, the embedding model in use, and a file format version, so future imports can describe exactly where a backup came from
  • Note: Only the JSON export can be re-imported; the text export remains a one-way format for vector-store uploads. Actions entries in a backup file are listed in the import summary but are not imported