← All updates
v0.0.23

Choose which provider embeds Insight Memory

October 2026

Related search in Insight Memory now stays on one embedding provider, chosen in Settings → AI Configuration next to your API keys. Adding another chat key, or switching the sidepanel model, does not change which provider embeds new feedback. If you do switch, SupaPM re-embeds saved insights first so older notes still show up as related.

Google chat also moves to Gemini 3.8 Flash. If you were on Gemini 3.7 Flash, SupaPM switches you over and asks you to confirm.

What's new

  • Settings → AI Configuration groups chat models, the optional classifier, and Insight Memory embeddings under separate headings. The embedding provider stays put when other chat keys are added.
  • Switching providers re-embeds the library first, and only saves the new choice if every embed succeeds. Cancel leaves stored insights unchanged.
  • Insights still on an older embedding model can be re-embedded from the same Settings section.
  • Google embeddings use Gemini's current text embedding model. Gemini 3.8 Flash is the Google chat default.

Have feedback or feature ideas? Get in touch.