Skip to content

SLayer 0.6.8: refresh of memory embeddings during ingest

Choose a tag to compare

View all tags
@ZmeiGorynych ZmeiGorynych released this 14 May 16:28
· 290 commits to main since this release
v0.6.8
1245dde
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

SLayer 0.6.8

A small follow-up release that keeps your memory embeddings in sync.

If you save memories and then change your data (or set up OPENAI_API_KEY after the fact), you no longer have to do anything special to get search working. Every slayer ingest, and every --ingest-on-startup boot pass, now re-embeds the memories rooted at the datasource being ingested. A stale embeddings.db (created before credentials were available, or after a hand-edit of memories.yaml) is repaired by the next ingest. If a particular memory fails to embed, the startup log attributes it as memory:<id> so you can tell it apart from a model or datasource-doc failure.

One small naming rule change worth flagging: : is now rejected in column, measure, model, datasource, and query names. The character is reserved for the aggregation suffix (revenue:sum) and was already implicitly off-limits; the validators now make that explicit. If you have an existing name with a colon in it, rename it before upgrading.

Assets 2
Loading