v1.6.1

Sling v1.6.1 (2026-09-08T12:50:56Z)

New Features

• Build execution status sync: Platform status reporting for build runs and per-model progress (BuildStatus / BuildModelStatus), including model log fetch via ExecBuildModels.

• UUID column typing: column_typing.uuid.as_text maps UUID columns to varchar(36) instead of unbounded text.

• Databricks merge_update: Uses MERGE INTO ... WHEN MATCHED THEN UPDATE instead of UPDATE ... FROM.

• Assist report composition: Compose redacted issue reports from collected parts without a local log snapshot; guard contact-form URL size under Cloudflare limits.

• MCP compact dataset output: Shared pipe-delimited compact helpers for discover / query / file-list assistant payloads.

• sling build subcommands: Split into run, compile, and list (replacing flag-based --compile / --list). Adds --path and --recursive for discovering sling_build.yml in subdirectories.

• Build pipeline / hook command: and env:: build: steps support command: run|test|compile|list (with test: true as an alias) and an env: overlay applied for the step.

Bug Fixes

• MSSQL uniqueidentifier transforms: Auto parse_ms_uuid no longer drops or corrupts user-defined staged transforms.

• DuckDB process death: Clearer mid-query death errors (including OOM-killer hint) and reconnect after the process exits.

• Unsupported API auth types: Return an explicit error for unknown authentication types instead of failing unclearly.

• Incremental where parentheses: incremental_where_cond syntax wraps conditions in parentheses.

• Parallel build logs: Per-model log buffers so parallel worker output is attributed to the correct model.

• Replication state locking: NewReplicationState owns the mutex so concurrent stream state updates are safe.