v1.1.0

Feat

• s3: read transcripts from an S3 source (claude-agent-sdk SessionStore) (#66)
• cli: add peek subcommand for session introspection (#51) (#59)

Fix

• types: parameterize bare dict/list/tuple for ty 0.0.46 (#78)
• sql_views: cast parent_uuid to VARCHAR so recursive CTEs work (#47) (#58)
• sql_views: read-side dedup for message_trajectory (#46) (#57)

Refactor

• workspace: split into 5 PEP 420 namespace packages under uv workspace (#60)

Perf

• cluster: hand polars numpy arrays directly in the clusters output frame (#74)
• cli: scope peek's UNNEST to one session, not the whole corpus (~5x) (#72)
• workers: project read_all to key column in LLM-worker anti-joins (#70)
• lance: column-project the uuid anti-join scan in get_embedded_uuids (#71)
• kappa: hoist category support out of the Cohen's-kappa bootstrap loop (#73)
• cli: defer lancedb import out of the CLI module-load path (~1.5s) (#77)
• community: zero-copy embedding extraction in session centroids (#68)
• community: segmented numpy mean for session centroids (#65)